[ { "path": ".gitignore", "content": "models/lora/*\nmodels/Realistic-LCM-LoRA/*\n" }, { "path": "README.md", "content": "# \"title\" Multi-LoRA Composition for Image Generation\n

\n \n \n \n \"Open\n

\n\n๐Ÿ–‹ **Authors:** [Ming Zhong](https://maszhongming.github.io/), [Yelong Shen](https://scholar.google.com/citations?user=S6OFEFEAAAAJ&hl=en), [Shuohang Wang](https://www.microsoft.com/en-us/research/people/shuowa/), [Yadong Lu](https://adamlu123.github.io/), [Yizhu Jiao](https://yzjiao.github.io/), [Siru Ouyang](https://ozyyshr.github.io/), [Donghan Yu](https://plusross.github.io/), [Jiawei Han](https://hanj.cs.illinois.edu/), [Weizhu Chen](https://www.microsoft.com/en-us/research/people/wzchen/)\n\n## ๐Ÿ“œ Overview\n\nLow-Rank Adaptation (LoRA) is extensively utilized in text-to-image models for the accurate rendition of specific elements like distinct characters or unique styles in generated images.\n\nOur project presents two training-free methods: **LoRA Switch** and **LoRA Composite** for integrating any number of elements in an image through multi-LoRA composition.\n\nThe figure below illustrates differences between the traditional LoRA Merge approach and our newly proposed techniques:\n\n

\n \"intro_case\"\n

\n\n## ๐Ÿš€ Getting Started\n\n### Setting Up the Environment\nTo begin, set up your environment with the necessary packages:\n```bash\nconda create --name multi-lora python=3.10\nconda activate multi-lora\npip install -r requirements.txt\n```\n\n### Downloading Pre-trained LoRAs\nOur **ComposLoRA** testbed collects 22 pre-trained LoRAs, spanning characters, clothing, styles, backgrounds, and objects. Download `ComposLoRA.zip` from [this link](https://drive.google.com/file/d/1SuwRgV1LtEud8dfjftnw-zxBMgzSCwIT/view?usp=sharing), put it in the [models](./models) folder, and unzip it.\n\n## ๐Ÿ–ผ๏ธ Image Generation with Multi-LoRA Composition\n\nTo compose multiple LoRAs using different methods during image generation, follow these steps:\n\nFirst, load the base model:\n\n```python\nfrom diffusers import DiffusionPipeline\n\npipeline = DiffusionPipeline.from_pretrained(\n 'SG161222/Realistic_Vision_V5.1_noVAE',\n custom_pipeline=\"MingZhong/StableDiffusionPipeline-with-LoRA-C\",\n use_safetensors=True\n).to(\"cuda\")\n```\n\nThis model from Hugging Face is selected for realistic-style image generation. Additionally, our custom pipeline integrates the LoRA composite method into the standard Stable Diffusion pipeline.\n\nNext, choose a character LoRA and a clothing LoRA from ComposLoRA for composition:\n\n```python\n# Load LoRAs\nlora_path = 'models/lora/reality'\npipeline.load_lora_weights(lora_path, weight_name=\"character_2.safetensors\", adapter_name=\"character\")\npipeline.load_lora_weights(lora_path, weight_name=\"clothing_2.safetensors\", adapter_name=\"clothing\")\n\n# List of LoRAs to be composed\ncur_loras = [\"character\", \"clothing\"]\n```\nSelect a composition method. \"switch\" and \"composite\" are our new proposals, offering alternatives to the traditional \"merge\" method:\n\n```python\nfrom callbacks import make_callback\n\nmethod = 'switch'\n\n# Initialize based on the selected composition method\nif method == \"merge\":\n pipeline.set_adapters(cur_loras)\n switch_callback = None\nelif method == \"switch\":\n pipeline.set_adapters([cur_loras[0]])\n switch_callback = make_callback(switch_step=args.switch_step, loras=cur_loras)\nelse:\n pipeline.set_adapters(cur_loras)\n switch_callback = None\n```\n\nFinally, set your prompt and generate the image:.\n\n```python\n# Set the prompts for image generation\nprompt = \"RAW photo, subject, 8k uhd, dslr, high quality, Fujifilm XT3, half-length portrait from knees up, scarlett, short red hair, blue eyes, school uniform, white shirt, red tie, blue pleated microskirt\"\nnegative_prompt = \"extra heads, nsfw, deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, text, cropped, out of frame, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck\"\n\n# Generate and save the image\ngenerator = torch.manual_seed(11)\nimage = pipeline(\n prompt=prompt, \n negative_prompt=negative_prompt,\n height=1024,\n width=768,\n num_inference_steps=100,\n guidance_scale=7,\n generator=generator,\n cross_attention_kwargs={\"scale\": 0.8},\n callback_on_step_end=switch_callback,\n lora_composite=True if method == \"composite\" else False\n).images[0]\nimage.save('example.png')\n```\n\nRefer to `example.py` for the full code, and adjust the following command to see results from different composition methods:\n\n```bash\npython example.py --method switch\n```\n\nImages generated by each of the three methods are showcased below:\n\n| Merge | Switch | Composite |\n|-------|--------|-----------|\n|

\"merge_example\"

|

\"switch_example\"

|

\"composite_example\"

|\n\n### Examples for LCM and LCM-LoRA\n\nOur methods can seamlessly integrate with both [LCM and LCM-LoRA](https://github.com/luosiallen/latent-consistency-model), significantly accelerating the image generation process by reducing the need for denoising steps to just 2-8. Below are example code and generated images:\n\n```bash\n# LCM\npython lcm_example.py --method switch\n```\n\n| Merge | Switch | Composite |\n|-------|--------|-----------|\n|

\"merge_lcm_example\"

|

\"switch_lcm_example\"

|

\"composite_lcm_example\"

|\n\n```bash\n# LCM-LoRA\npython lcm_lora_example.py --method composite\n```\n\n| Merge | Switch | Composite |\n|-------|--------|-----------|\n|

\"merge_lcm_lora_example\"

|

\"switch_lcm_lora_example\"

|

\"composite_lcm_lora_example\"

|\n\n### Example for SDXL\n\nIn addition to SD1.5-based backbones, our method can also be applied to SDXL. The following example demonstrates how to combine [Pikachu](https://huggingface.co/TheLastBen/Pikachu_SDXL) and [Vision Pro](https://huggingface.co/fofr/sdxl-vision-pro) in an image:\n\n```bash\npython sdxl_example.py --method switch\n```\n\n| Merge | Switch | Composite |\n|-------|--------|-----------|\n|

\"merge_sdxl_example\"

|

\"switch_sdxl_example\"

|

\"composite_sdxl_example\"

|\n\n## ๐ŸŽจ Experiments on ComposLoRA\n\n**ComposLoRA** features 22 LoRAs and 480 different composition sets, allowing for the generation of images with any composition of 2-5 LoRAs, including at least one character LoRA.\n\n### Image Generation\nTo generate anime-style images incorporating 2 LoRAs using LoRA Composite method, use the following command:\n\n```bash\nexport CUDA_VISIBLE_DEVICES=0\n\npython compose_lora.py \\\n --method composite \\\n --compos_num 2 \\\n --save_path output \\\n --lora_scale 0.8 \\\n --image_style anime \\\n --denoise_steps 200 \\\n --cfg_scale 10 \\\n```\n\nAdjust the parameters in `compos_reality.sh` and `compose_anime.sh` for different compositions.\n\n### Comparative Evaluation with GPT-4V\n\nFor comparative evaluation on composition efficacy and image quality, we use GPT-4V. Set your OpenAI API key first:\n\n```bash\nexport OPENAI_API_KEY='your_openai_api_key_here'\n```\n\nThen, compare the composite and merge methods with this command:\n\n```bash\npython evaluate.py \\\n --base_method merge \\\n --comp_method composite \\\n --compos_num 2 \\\n --image_style anime \\\n --image_path output \\\n --save_path eval_result \\\n```\n\nModify `eval.sh` for comparative evaluation under different conditions. Note the position bias of GPT-4V as mentioned in our paper, making it essential to input images in both orders and average the scores for a fair final assessment.\n\n## Human Evaluation\nWe also conduct human evaluations on 120 generated images to assess composition and image quality from a human perspective. These evaluations offer additional insights into the performance of our Multi-LoRA Composition methods and metrics. For detailed information on the evaluation process and results, please visit the [human_eval](./human_eval) folder.\n\n## ๐Ÿ“š Citation\nIf you find this work useful, please kindly cite our paper:\n```\n@article{zhong2024multi,\n title={Multi-LoRA Composition for Image Generation},\n author={Zhong, Ming and Shen, Yelong and Wang, Shuohang and Lu, Yadong and Jiao, Yizhu and Ouyang, Siru and Yu, Donghan and Han, Jiawei and Chen, Weizhu},\n journal={arXiv preprint arXiv:2402.16843},\n year={2024}\n}\n```\n\n" }, { "path": "analyze.py", "content": "import json\nfrom os.path import join\n\nimage_style = \"anime\"\ncompos_num = 2\nmethod_1 = 'merge'\nmethod_2 = 'composite'\n\ndef calculate_averages(evals):\n # Initialize sum and count dictionaries\n sum_scores = {'image 1': {'composition quality': 0, 'image quality': 0},\n 'image 2': {'composition quality': 0, 'image quality': 0}}\n count = {'image 1': 0, 'image 2': 0}\n\n # Sum up scores and count entries\n for entry in evals:\n for image, image_scores in entry['scores'].items():\n for dimension, value in image_scores.items():\n sum_scores[image][dimension] += value\n count[image] += 1\n\n # Calculate averages\n avg_scores = {}\n for image, image_scores in sum_scores.items():\n avg_scores[image] = {dimension: value / (count[image] // 2) for dimension, value in image_scores.items()}\n\n return avg_scores\n\ndef compare_methods(evals):\n comparison = {'composition quality': {'wins': 0, 'ties': 0, 'losses': 0},\n 'image quality': {'wins': 0, 'ties': 0, 'losses': 0}}\n total_comparisons = 0\n\n for entry in evals:\n total_comparisons += 1\n for dimension in ['composition quality', 'image quality']:\n if entry['scores']['image 2'][dimension] > entry['scores']['image 1'][dimension]:\n comparison[dimension]['wins'] += 1\n elif entry['scores']['image 2'][dimension] < entry['scores']['image 1'][dimension]:\n comparison[dimension]['losses'] += 1\n else:\n comparison[dimension]['ties'] += 1\n\n # Convert wins/losses/ties counts to win rates\n win_rates = {dim: {outcome: count / total_comparisons for outcome, count in outcomes.items()}\n for dim, outcomes in comparison.items()}\n\n return win_rates\n\nwith open(join(f'eval_result/{image_style}_{compos_num}_elements_{method_1}_vs_{method_2}.json')) as f:\n eval_results = json.loads(f.read())\n\n# Total numbers\nprint(f'Total {len(eval_results)} evaluation pairs!')\n\n# Calculate averages\naverage_scores = calculate_averages(eval_results)\nprint(\"Average Scores:\")\nfor method, dimensions in average_scores.items():\n if method == \"image 1\":\n print(f\" {method_1}:\")\n else:\n print(f\" {method_2}:\")\n for dimension, score in dimensions.items():\n print(f\" {dimension}: {score:.2f}\")\n\n# Compare methods (Method 2 vs Method 1)\nmethod_comparison = compare_methods(eval_results)\nprint(f\"\\nMethod Comparison ({method_2} vs {method_1}):\")\nfor dimension, outcomes in method_comparison.items():\n print(f\" {dimension.capitalize()}:\")\n for outcome, rate in outcomes.items():\n print(f\" {outcome.capitalize()} Rate: {rate:.2f}\")\n\n\n" }, { "path": "anime_lora_info.json", "content": "{\n \"character\": [\n {\n \"id\": \"character_1\",\n \"name\": \"Kamado Nezuko\",\n \"trigger\": [\n \"kamado nezuko\",\n \"black hair\",\n \"pink eyes\",\n \"forehead\"\n ],\n \"url\": \"https://civitai.com/models/20346/nezuko-demon-slayer-kimetsu-no-yaiba-lora?modelVersionId=24191\"\n },\n {\n \"id\": \"character_2\",\n \"name\": \"Texas the Omertosa in Arknights\",\n \"trigger\": [\n \"omertosa\",\n \"1girl\",\n \"wolf ears\",\n \"long hair\"\n ],\n \"url\": \"https://civitai.com/models/6779/arknights-texas-the-omertosa?modelVersionId=7974\"\n },\n {\n \"id\": \"character_3\",\n \"name\": \"Son Goku\",\n \"trigger\": [\n \"son goku\",\n \"spiked hair\",\n \"muscular male\",\n \"wristband\"\n ],\n \"url\": \"https://civitai.com/models/18279/son-goku-dragon-ball-all-series-lora?modelVersionId=21654\"\n }\n ],\n \"clothing\": [\n {\n \"id\": \"clothing_1\",\n \"name\": \"Garreg Mach Monastery Uniform\",\n \"trigger\": [\n \"gmuniform\",\n \"blue thighhighs\",\n \"long sleeves\"\n ],\n \"url\": \"https://civitai.com/models/151330/garreg-mach-monastery-uniform-fire-emblem-three-houses-lora-or-2-variants?modelVersionId=169217\"\n },\n {\n \"id\": \"clothing_2\",\n \"name\": \"Zero Suit (Metroid)\",\n \"trigger\": [\n \"zero suit\",\n \"blue gloves\",\n \"high heels\"\n ],\n \"url\": \"https://civitai.com/models/82304/zero-suit-metroid-outfit-lora?modelVersionId=87396\"\n }\n ],\n \"style\": [\n {\n \"id\": \"style_1\",\n \"name\": \"Hand-drawn Style\",\n \"trigger\": [\n \"lineart\",\n \"hand-drawn style\"\n ],\n \"url\": \"https://civitai.com/models/143532/handdrawn\"\n },\n {\n \"id\": \"style_2\",\n \"name\": \"Shuimobysim (Chinese Ink Wash Style)\",\n \"trigger\": [\n \"shuimobysim\",\n \"traditional chinese ink painting\"\n ],\n \"url\": \"https://civitai.com/models/12597?modelVersionId=14856\"\n }\n ],\n \"background\": [\n {\n \"id\": \"background_1\",\n \"name\": \"Bamboolight Background\",\n \"trigger\": [\n \"bamboolight\",\n \"outdoors\",\n \"bamboo\"\n ],\n \"url\": \"https://civitai.com/models/113381/lora-bamboolight-concept-with-dropout-and-noise-version?modelVersionId=122477\"\n },\n {\n \"id\": \"background_2\",\n \"name\": \"Auroral Background\",\n \"trigger\": [\n \"auroral\",\n \"starry sky\",\n \"outdoors\"\n ],\n \"url\": \"https://civitai.com/models/85026?modelVersionId=156828\"\n }\n ],\n \"object\": [\n {\n \"id\": \"object_1\",\n \"name\": \"Huge Two-Handed Burger\",\n \"trigger\": [\n \"two-handed burger\",\n \"holding a huge burger with both hands\"\n ],\n \"url\": \"https://civitai.com/models/27858/huge-two-handed-burger-lora?modelVersionId=34071\"\n },\n {\n \"id\": \"object_2\",\n \"name\": \"Toast\",\n \"trigger\": [\n \"toast\",\n \"toast in mouth\"\n ],\n \"url\": \"https://civitai.com/models/26945/toast-in-mouth-lora?modelVersionId=32252\"\n }\n ]\n}" }, { "path": "callbacks.py", "content": "import json\n\ndef make_callback(switch_step, loras):\n def switch_callback(pipeline, step_index, timestep, callback_kwargs):\n callback_outputs = {}\n if step_index > 0 and step_index % switch_step == 0:\n for cur_lora_index, lora in enumerate(loras):\n if lora in pipeline.get_active_adapters():\n next_lora_index = (cur_lora_index + 1) % len(loras)\n pipeline.set_adapters(loras[next_lora_index])\n break\n return callback_outputs\n return switch_callback" }, { "path": "compose_anime.sh", "content": "export CUDA_VISIBLE_DEVICES=0\n\npython compose_lora.py \\\n --method composite \\\n --compos_num 2 \\\n --save_path output \\\n --lora_scale 0.8 \\\n --image_style anime \\\n --denoise_steps 200 \\\n --cfg_scale 10 \\\n" }, { "path": "compose_lora.py", "content": "import os\nimport torch\nimport argparse\nfrom tqdm import tqdm\nfrom os.path import join, exists\nfrom diffusers import DiffusionPipeline, AutoencoderKL\nfrom diffusers import DPMSolverMultistepScheduler\n\nfrom utils import load_lora_info, generate_combinations\nfrom utils import get_prompt\nfrom callbacks import make_callback\n\ndef main(args):\n\n # set path based on the image style\n args.save_path = args.save_path + \"_\" + args.image_style\n args.lora_path = join(args.lora_path, args.image_style)\n\n # load all the information of LoRAs\n lora_info = load_lora_info(args.image_style, args.lora_info_path)\n\n # set base model based on the image style\n if args.image_style == 'anime':\n model_name = 'gsdf/Counterfeit-V2.5'\n else:\n model_name = 'SG161222/Realistic_Vision_V5.1_noVAE'\n\n pipeline = DiffusionPipeline.from_pretrained(\n model_name,\n custom_pipeline=\"MingZhong/StableDiffusionPipeline-with-LoRA-C\",\n # torch_dtype=torch.float16,\n use_safetensors=True\n ).to(\"cuda\")\n\n # set vae\n if args.image_style == \"reality\":\n vae = AutoencoderKL.from_pretrained(\n \"stabilityai/sd-vae-ft-mse\",\n # torch_dtype=torch.float16\n ).to(\"cuda\")\n pipeline.vae = vae\n\n # set scheduler\n schedule_config = dict(pipeline.scheduler.config)\n schedule_config[\"algorithm_type\"] = \"dpmsolver++\"\n pipeline.scheduler = DPMSolverMultistepScheduler.from_config(schedule_config)\n\n # initialize LoRAs\n for element in list(lora_info.keys()):\n for lora in lora_info[element]:\n pipeline.load_lora_weights(\n args.lora_path,\n weight_name=lora['id'] + '.safetensors',\n adapter_name=lora['id']\n )\n\n # generate all combinations that can be composed\n combinations = generate_combinations(lora_info, args.compos_num)\n \n # prompt initialization\n init_prompt, negative_prompt = get_prompt(args.image_style)\n\n # generate images for each combination based on LoRAs\n for combo in tqdm(combinations):\n cur_loras = [lora['id'] for lora in combo]\n\n # set prompt\n triggers = [trigger for lora in combo for trigger in lora['trigger']]\n prompt = init_prompt + ', ' + ', '.join(triggers)\n \n # set LoRAs\n if args.method == \"switch\":\n pipeline.set_adapters([cur_loras[0]])\n switch_callback = make_callback(args.switch_step,\n cur_loras)\n elif args.method == \"merge\":\n pipeline.set_adapters(cur_loras)\n switch_callback = None\n else:\n pipeline.set_adapters(cur_loras)\n switch_callback = None\n\n # generate images\n image = pipeline(\n prompt=prompt, \n negative_prompt=negative_prompt,\n height=args.height,\n width=args.width,\n num_inference_steps=args.denoise_steps,\n guidance_scale=args.cfg_scale,\n generator=args.generator,\n cross_attention_kwargs={\"scale\": args.lora_scale},\n callback_on_step_end=switch_callback,\n lora_composite=True if args.method == \"composite\" else False\n ).images[0]\n \n # save image\n save_path = join(args.save_path, f'{args.compos_num}_elements')\n if not exists(save_path):\n os.makedirs(save_path)\n file_name = args.method + '_' + '_'.join([lora['id'] for lora in combo]) + '.png'\n image.save(join(save_path, file_name))\n\nif __name__ == \"__main__\":\n parser = argparse.ArgumentParser(\n description='Given LoRAs in the ComposLoRA benchmark, generate images with arbitrary combinations based on LoRAs'\n )\n\n # Arguments for composing LoRAs\n parser.add_argument('--compos_num', default=2,\n help='number of elements to be combined in a single image', type=int)\n parser.add_argument('--method', default='switch',\n choices=['merge', 'switch', 'composite'],\n help='methods for combining LoRAs', type=str)\n parser.add_argument('--save_path', default='output',\n help='path to save the generated image', type=str)\n parser.add_argument('--lora_path', default='models/lora',\n help='path to store all LoRA models', type=str)\n parser.add_argument('--lora_info_path', default='lora_info.json',\n help='path to stroe all LoRA information', type=str)\n parser.add_argument('--lora_scale', default=0.8,\n help='scale of each LoRA when generating images', type=float)\n parser.add_argument('--switch_step', default=5,\n help='number of steps to switch LoRA during denoising, applicable only in the switch method', type=int)\n\n # Arguments for generating images\n parser.add_argument('--height', default=512,\n help='height of the generated images', type=int)\n parser.add_argument('--width', default=512,\n help='width of the generated images', type=int)\n parser.add_argument('--denoise_steps', default=200,\n help='number of the denoising steps', type=int)\n parser.add_argument('--cfg_scale', default=10,\n help='scale for classifier-free guidance', type=float)\n parser.add_argument('--seed', default=111,\n help='seed for generating images', type=int)\n parser.add_argument('--image_style', default='anime',\n choices=['anime', 'reality'],\n help='sytles of the generated images', type=str)\n\n args = parser.parse_args()\n args.generator = torch.manual_seed(args.seed)\n\n main(args)" }, { "path": "compose_reality.sh", "content": "export CUDA_VISIBLE_DEVICES=0\n\npython compose_lora.py \\\n --method switch \\\n --compos_num 2 \\\n --save_path output \\\n --lora_scale 0.8 \\\n --image_style reality \\\n --denoise_steps 100 \\\n --cfg_scale 7 \\\n --height 1024 \\\n --width 768 \\\n" }, { "path": "docs/index.html", "content": "\n\n\n \n \n Multi-LoRA Composition for Image Generation\n\n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n \n \n\n\n\n\n\n\n
\n
\n
\n
\n
\n

Multi-LoRA Composition for Image Generation

\n
\n \n Ming Zhong1,\n \n Yelong Shen2,\n \n Shuohang Wang2,\n \n \n Yadong Lu2,\n \n \n Yizhu Jiao1,\n \n \n Siru Ouyang1,\n \n \n Donghan Yu2\n \n \n Jiawei Han1\n \n \n Weizhu Chen2\n \n
\n\n
\n 1University of Illinois Urbana-Champaign,\n 2Microsoft Corporation\n
\n\n
\n
\n \n \n \n \n \n \n Paper\n \n \n \n \n \n \n \n \n \n \n \n Code\n \n \n \n \n \n \n \n \n ComposLoRA\n \n \n \n \n \n \n \n \n Twitter\n \n \n
\n\n
\n
\n
\n
\n
\n
\n\n
\n
\n
\n
\n This project explores new methods for text-to-image generation, with a focus on the integration of multiple Low-Rank Adaptations (LoRAs) to create highly customized and detailed images. We present LoRA Switch and LoRA Composite, approaches that aim to surpass traditional techniques in terms of accuracy and image quality, especially in complex compositions.\n

Project Features: \n
    \n
  • ๐Ÿš€ Training-Free Methods\n
      \n
    • LoRA Switch and LoRA Composite enable dynamic and precise integration of multiple LoRAs without fine-tuning.
      • \n
    • Unlike methods that merge LoRA weights, ours focuses on the decoding process, keeping all LoRA weights intact.
      • \n
    \n
    • \n
  • ๐Ÿ“Š ComposLoRA Testbed\n
      \n
    • A new comprehensive platform, featuring 480 composition sets and 22 pre-trained LoRAs across six categories.
      • \n
    • ComposLoRA is designed for the quantitative evaluation of LoRA-based composable image generation tasks.
      • \n
    \n
    • \n
  • ๐Ÿ“ GPT-4V-based Evaluator\n
      \n
    • We propose using GPT-4V as an evaluator to assess the efficacy of compositions and the quality of images.
      • \n
    • This evaluator has demonstrated a better correlation with human judgments.
      • \n
    \n
    • \n
  • ๐Ÿ† Superior Performance\n
      \n
    • Both automated and human evaluations show that our approaches substantially outperform the prevalent LoRA Merge.
      • \n
    • Our methods exhibit a more significant advantage when generating complex compositions.
      • \n
    \n
    • \n
  • ๐Ÿ•ต๏ธโ€โ™‚๏ธ Detailed Analysis\n
      \n
    • We delve deeply into the scenarios where each method excels.
      • \n
    • We explore the potential bias associated with using GPT-4V for evaluation.
      • \n
    \n
    • \n
\n
\n
\n
\n
\n\n
\n
\n
\n \n \n

\n Multi-LoRA composition techniques effectively blend different elements into a cohesive image. Unlike the conventional LoRA Merge approach, which can lead to detail loss and image distortion as more LoRAs are added, our methods retain the accuracy of each element and the overall image quality.\n

\n
\n
\n
\n\n
\n
\n \n
\n
\n

Methods of Multi-LoRA Composition

\n

\n \n

\n
\n
    \n
  • LoRA Merge:\n
      \n
    • Prevalent approach to integrating multiple elements in a unified way in an image.
      • \n
    • It is realized by linearly combining multiple LoRAs to synthesize a unified LoRA, subsequently plugged into the text-to-image model.
      • \n
    • LoAR Merge completely overlooks the interaction with the diffusion model during the generative process, resulting in the deformation of the hamburger and fingers in the Figure.
      • \n
    \n
    • \n
  • LoRA Switch (LoRA-S):\n
      \n
    • To explore activating a single LoRA in each denoising step, we propose LoRA Switch.
      • \n
    • This method introduces a dynamic adaptation mechanism within diffusion models by sequentially activating individual LoRAs at designated intervals throughout the decoding process.
      • \n
    • As illustrated in the Figure, each LoRA is represented by a unique color corresponding to a specific element, with only one LoRA engaged per denoising step.
      • \n
    \n
    • \n
  • LoRA Composite (LoRA-C):\n
      \n
    • To explore incorporating all LoRAs at each timestep without merging weight matrices, we propose LoRA Composite.
      • \n
    • It involves calculating both unconditional and conditional score estimates for each LoRA individually at each step.
      • \n
    • By aggregating these scores, the technique ensures balanced guidance throughout the image generation process, facilitating the cohesive integration of all elements represented by different LoRAs.
      • \n
    \n
    • \n
\n
\n
\n
\n
\n
\n\n
\n
\n \n
\n
\n

GPT-4V-based Evaluator

\n

\n \n

\n
\n
    \n
  • While existing metrics can calculate the alignment between text and images, they fall short in assessing the intricacies of specific elements within an image and the quality of their composition.
    • \n
  • We employ a comparative evaluation method, utilizing GPT-4V to rate generated images across two dimensions: composition quality and image quality.
    • \n
\n
\n
\n
\n
\n
\n\n
\n
\n \n
\n
\n

Experimental Results

\n

\n \n

\n
\n
    \n
  • Our proposed method consistently outperforms LoRA Merge across all configurations and in both dimensions, with the margin of superiority increasing as the number of LoRAs grows.
    • \n
  • LoRA Switch shows superior performance in composition quality, whereas LoRA Composite excels in image quality.
    • \n
  • The task of compositional image generation remains highly challenging, especially as the number of elements to be composed increases.
    • \n
\n
\n

\n \n

\n
\n
    \n
  • Human evaluations aligh with GPT-4V's findings.
    • \n
  • GPT-4V-based evaluator we adopt shows substantially higher correlations with human judgments, affirming the validity of our evaluation framework.
    • \n
\n
\n
\n
\n
\n
\n\n
\n
\n \n
\n
\n

Analysis

\n

\n \n

\n
\n
    \n
  • LoRA Switch is more adept at composing elements in realistic-style images.
    • \n
  • LoRA Composite shows a stronger performance in anime-style imagery.
    • \n
\n
\n

\n \n

\n
\n
    \n
  • The efficiency of the LoRA Switch improves progressively with increased step size, reaching peak performance at 5.
    • \n
  • The initial choice of LoRA in the activation sequence clearly influences overall performance, while alterations in the subsequent order have minimal impact.
    • \n
\n
\n

\n \n

\n
\n
    \n
  • GPT-4V exhibits significant positional bias in comparative evaluation.
    • \n
  • This bias varies depending on the input position of the image and the dimension of the evaluation.
    • \n
\n
\n
\n
\n
\n
\n\n
\n
\n

BibTeX

\n 
@article{zhong2024multi,\n      title={Multi-LoRA Composition for Image Generation},\n      author={Zhong, Ming and Shen, Yelong and Wang, Shuohang and Lu, Yadong and Jiao, Yizhu and Ouyang, Siru and Yu, Donghan and Han, Jiawei and Chen, Weizhu},\n      journal={arXiv preprint arXiv:2402.16843},\n      year={2024}\n}
\n
\n
\n\n\n\n\n\n\n" }, { "path": "docs/static/css/bulma.css.map.txt", "content": "{\"version\":3,\"sources\":[\"../bulma.sass\",\"../sass/utilities/_all.sass\",\"../sass/utilities/animations.sass\",\"bulma.css\",\"../sass/utilities/mixins.sass\",\"../sass/utilities/initial-variables.sass\",\"../sass/utilities/controls.sass\",\"../sass/base/_all.sass\",\"../sass/base/minireset.sass\",\"../sass/base/generic.sass\",\"../sass/utilities/derived-variables.sass\",\"../sass/elements/_all.sass\",\"../sass/elements/box.sass\",\"../sass/elements/button.sass\",\"../sass/utilities/functions.sass\",\"../sass/elements/container.sass\",\"../sass/elements/content.sass\",\"../sass/elements/icon.sass\",\"../sass/elements/image.sass\",\"../sass/elements/notification.sass\",\"../sass/elements/progress.sass\",\"../sass/elements/table.sass\",\"../sass/elements/tag.sass\",\"../sass/elements/title.sass\",\"../sass/elements/other.sass\",\"../sass/form/_all.sass\",\"../sass/form/shared.sass\",\"../sass/form/input-textarea.sass\",\"../sass/form/checkbox-radio.sass\",\"../sass/form/select.sass\",\"../sass/form/file.sass\",\"../sass/form/tools.sass\",\"../sass/components/_all.sass\",\"../sass/components/breadcrumb.sass\",\"../sass/components/card.sass\",\"../sass/components/dropdown.sass\",\"../sass/components/level.sass\",\"../sass/components/media.sass\",\"../sass/components/menu.sass\",\"../sass/components/message.sass\",\"../sass/components/modal.sass\",\"../sass/components/navbar.sass\",\"../sass/components/pagination.sass\",\"../sass/components/panel.sass\",\"../sass/components/tabs.sass\",\"../sass/grid/_all.sass\",\"../sass/grid/columns.sass\",\"../sass/grid/tiles.sass\",\"../sass/helpers/_all.sass\",\"../sass/helpers/color.sass\",\"../sass/helpers/flexbox.sass\",\"../sass/helpers/float.sass\",\"../sass/helpers/other.sass\",\"../sass/helpers/overflow.sass\",\"../sass/helpers/position.sass\",\"../sass/helpers/spacing.sass\",\"../sass/helpers/typography.sass\",\"../sass/helpers/visibility.sass\",\"../sass/layout/_all.sass\",\"../sass/layout/hero.sass\",\"../sass/layout/section.sass\",\"../sass/layout/footer.sass\"],\"names\":[],\"mappings\":\"AACA,6DAAA;ACDA,oBAAA;ACAA;EACE;IACE,uBAAuB;ECGzB;EDFA;IACE,yBAAyB;ECI3B;AACF;ADTA;EACE;IACE,uBAAuB;ECGzB;EDFA;IACE,yBAAyB;ECI3B;AACF;;AC0JA;;;;EANE,2BAA2B;EAC3B,yBAAyB;EACzB,sBAAsB;EACtB,qBAAqB;EACrB,iBAAiB;AD7InB;;ACkKA;EAfE,6BAD8B;EAE9B,kBAAkB;EAClB,eAAe;EACf,aAAa;EACb,YAAY;EACZ,cAAc;EACd,eAAe;EACf,qBAAqB;EACrB,oBAAoB;EACpB,kBAAkB;EAClB,QAAQ;EACR,yBAAyB;EACzB,wBAAwB;EACxB,cAAc;AD/IhB;;ACqJE;;EACE,qBC3IkB;AFNtB;;ACwNA;EAhEE,qBAAqB;EACrB,wBAAwB;EACxB,uCClM2B;EDmM3B,YAAY;EACZ,uBC/HuB;EDgIvB,eAAe;EACf,oBAAoB;EACpB,qBAAqB;EACrB,YAAY;EACZ,cAAc;EACd,YAAY;EACZ,YAAY;EACZ,gBAAgB;EAChB,eAAe;EACf,gBAAgB;EAChB,eAAe;EACf,aAAa;EACb,kBAAkB;EAClB,mBAAmB;EACnB,WAAW;ADpJb;;ACqJE;EAEE,uBCzM2B;ED0M3B,WAAW;EACX,cAAc;EACd,SAAS;EACT,kBAAkB;EAClB,QAAQ;EACR,0DAA0D;EAC1D,+BAA+B;ADnJnC;;ACoJE;EACE,WAAW;EACX,UAAU;ADjJd;;ACkJE;EACE,WAAW;EACX,UAAU;AD/Id;;ACgJE;EAEE,uCCtOyB;AFwF7B;;AC+IE;EACE,uCCxOyB;AF4F7B;;AC8IE;EACE,YAAY;EACZ,gBAAgB;EAChB,eAAe;EACf,gBAAgB;EAChB,eAAe;EACf,WAAW;AD3If;;AC4IE;EACE,YAAY;EACZ,gBAAgB;EAChB,eAAe;EACf,gBAAgB;EAChB,eAAe;EACf,WAAW;ADzIf;;AC0IE;EACE,YAAY;EACZ,gBAAgB;EAChB,eAAe;EACf,gBAAgB;EAChB,eAAe;EACf,WAAW;ADvIf;;ACwJA;EAXE,mDAA2C;UAA3C,2CAA2C;EAC3C,yBC7P4B;ED8P5B,uBCjMuB;EDkMvB,+BAA+B;EAC/B,6BAA6B;EAC7B,WAAW;EACX,cAAc;EACd,WAAW;EACX,kBAAkB;EAClB,UAAU;ADzIZ;;ACqJA;;;;;;;;;;;;;;;;;EANE,SADuB;EAEvB,OAFuB;EAGvB,kBAAkB;EAClB,QAJuB;EAKvB,MALuB;ADtHzB;;AGvHA;;;;;EA3BE,qBAAqB;EACrB,wBAAwB;EACxB,mBAAmB;EACnB,6BAA+C;EAC/C,kBDqDU;ECpDV,gBAAgB;EAChB,oBAAoB;EACpB,eDkBW;ECjBX,aAfoB;EAgBpB,2BAA2B;EAC3B,gBAhBuB;EAiBvB,iCAf+D;EAgB/D,gCAfkE;EAgBlE,iCAhBkE;EAiBlE,8BAlB+D;EAmB/D,kBAAkB;EAClB,mBAAmB;AH0JrB;;AGxJE;;;;;;;;;;;;;;;;;EAIE,aAAa;AHwKjB;;AGvKE;;;;;;;;;;;;;;;;EAEE,mBAAmB;AHwLvB;;AI7NA,eAAA;ACAA,0EAAA;AAEA;;;;;;;;;;;;;;;;;;;;;;;EAuBE,SAAS;EACT,UAAU;ALgOZ;;AK7NA;;;;;;EAME,eAAe;EACf,mBAAmB;ALgOrB;;AK7NA;EACE,gBAAgB;ALgOlB;;AK7NA;;;;EAIE,SAAS;ALgOX;;AK7NA;EACE,sBAAsB;ALgOxB;;AK9NA;EAII,mBAAmB;AL8NvB;;AK3NA;;EAEE,YAAY;EACZ,eAAe;AL8NjB;;AK3NA;EACE,SAAS;AL8NX;;AK3NA;EACE,yBAAyB;EACzB,iBAAiB;AL8NnB;;AK5NA;;EAEE,UAAU;AL+NZ;;AKjOA;;EAII,mBAAmB;ALkOvB;;AK9PA;EClBE,uBJjB6B;EIkB7B,eAhCc;EAiCd,kCAAkC;EAClC,mCAAmC;EACnC,gBAlCoB;EAmCpB,kBAhCsB;EAiCtB,kBAhCsB;EAiCtB,kCApCiC;EAqCjC,8BAAsB;KAAtB,2BAAsB;MAAtB,0BAAsB;UAAtB,sBAAsB;ANoRxB;;AMlRA;;;;;;;EAOE,cAAc;ANqRhB;;AMnRA;;;;;;EAME,oLJ7ByL;AFmT3L;;AMpRA;;EAEE,6BAA6B;EAC7B,4BAA4B;EAC5B,sBJlC0B;AFyT5B;;AMrRA;EACE,cJ3D4B;EI4D5B,cA1DkB;EA2DlB,gBJ3BiB;EI4BjB,gBA1DoB;ANkVtB;;AMpRA;EACE,cJpDgC;EIqDhC,eAAe;EACf,qBAAqB;ANuRvB;;AM1RA;EAKI,mBAAmB;ANyRvB;;AM9RA;EAOI,cJ1E0B;AFqW9B;;AMzRA;EACE,4BJtE4B;EIuE5B,cCpBsB;EDqBtB,kBArEiB;EAsEjB,mBAvEkB;EAwElB,4BAzEgC;ANqWlC;;AM1RA;EACE,4BJ7E4B;EI8E5B,YAAY;EACZ,cAAc;EACd,WAxEa;EAyEb,gBAxEkB;ANqWpB;;AM3RA;EACE,YAAY;EACZ,eAAe;AN8RjB;;AM5RA;;EAEE,wBAAwB;AN+R1B;;AM7RA;EACE,kBAvFuB;ANuXzB;;AM9RA;EACE,mBAAmB;EACnB,oBAAoB;ANiStB;;AM/RA;EACE,cJ1G4B;EI2G5B,gBJrEe;AFuWjB;;AM9RA;EACE,YAAY;ANiSd;;AM/RA;EL1DE,iCAAiC;EK4DjC,4BJ7G4B;EI8G5B,cJpH4B;EIqH5B,kBAjGqB;EAkGrB,gBAAgB;EAChB,uBAlG0B;EAmG1B,gBAAgB;EAChB,iBAAiB;ANkSnB;;AM1SA;EAUI,6BAA6B;EAC7B,mBAAmB;EACnB,cAvGoB;EAwGpB,UAAU;ANoSd;;AMlSA;;EAGI,mBAAmB;ANoSvB;;AMvSA;;EAKM,mBAAmB;ANuSzB;;AM5SA;EAOI,cJxI0B;AFib9B;;AQvbA,mBAAA;ACSA;EAEE,uBPI6B;EOH7B,kBP0DgB;EOzDhB,0FPX2B;EOY3B,cPP4B;EOQ5B,cAAc;EACd,gBAZmB;AT6brB;;AS/aA;EAGI,yEPC8B;AF+alC;;ASnbA;EAKI,oEPD8B;AFmblC;;AUzZA;EAGE,uBRpC6B;EQqC7B,qBR1C4B;EQ2C5B,iBPlDwB;EOmDxB,cRhD4B;EQiD5B,eAAe;EAGf,uBAAuB;EACvB,iCApD6D;EAqD7D,iBApD6B;EAqD7B,kBArD6B;EAsD7B,8BAvD6D;EAwD7D,kBAAkB;EAClB,mBAAmB;AVwZrB;;AUxaA;EAkBI,cAAc;AV0ZlB;;AU5aA;EAwBM,aAAa;EACb,YAAY;AVwZlB;;AUjbA;ETgGI,+BSrEwG;ETqExG,oBSpEgE;AV0ZpE;;AUtbA;ETgGI,mBSlEgE;ETkEhE,gCSjEwG;AV4Z5G;;AU3bA;EAiCM,+BAAmF;EACnF,gCAAoF;AV8Z1F;;AUhcA;EAsCI,qBR7E0B;EQ8E1B,cRjF0B;AF+e9B;;AUrcA;EA0CI,qBRpE8B;EQqE9B,cRrF0B;AFof9B;;AU1cA;EA6CM,kDRvE4B;AFwelC;;AU9cA;EAgDI,qBRzF0B;EQ0F1B,cR3F0B;AF6f9B;;AUndA;EAoDI,6BAA6B;EAC7B,yBAAyB;EACzB,cR/F0B;EQgG1B,0BAjF8B;AVoflC;;AU1dA;EA4DM,4BR/FwB;EQgGxB,cRvGwB;AFygB9B;;AU/dA;EAgEM,yBCH2B;EDI3B,cR3GwB;AF8gB9B;;AUpeA;;EAoEM,6BAA6B;EAC7B,yBAAyB;EACzB,gBAAgB;AVqatB;;AU3eA;EA2EM,uBR5GyB;EQ6GzB,yBAAyB;EACzB,cR3HuB;AF+hB7B;;AUjfA;EAgFQ,yBCnByB;EDoBzB,yBAAyB;EACzB,cRhIqB;AFqiB7B;;AUvfA;EAqFQ,yBAAyB;EACzB,cRpIqB;AF0iB7B;;AU5fA;EAwFU,mDRzHqB;AFiiB/B;;AUhgBA;EA2FQ,yBC9ByB;ED+BzB,yBAAyB;EACzB,cR3IqB;AFojB7B;;AUtgBA;;EAgGQ,uBRjIuB;EQkIvB,yBAAyB;EACzB,gBAAgB;AV2axB;;AU7gBA;EAoGQ,yBRlJqB;EQmJrB,YRtIuB;AFmjB/B;;AUlhBA;EAwGU,uBC3CuB;AXydjC;;AUthBA;;EA2GU,yBRzJmB;EQ0JnB,yBAAyB;EACzB,gBAAgB;EAChB,YR/IqB;AF+jB/B;;AU9hBA;EAiHU,gEAA4E;AVibtF;;AUliBA;EAmHQ,6BAA6B;EAC7B,mBRrJuB;EQsJvB,YRtJuB;AFykB/B;;AUxiBA;EA0HU,uBR3JqB;EQ4JrB,mBR5JqB;EQ6JrB,cR1KmB;AF4lB7B;;AU9iBA;EA+HY,4DAA8D;AVmb1E;;AUljBA;EAqIc,gEAA4E;AVib1F;;AUtjBA;;EAwIU,6BAA6B;EAC7B,mBR1KqB;EQ2KrB,gBAAgB;EAChB,YR5KqB;AF+lB/B;;AU9jBA;EA6IQ,6BAA6B;EAC7B,qBR5LqB;EQ6LrB,cR7LqB;AFknB7B;;AUpkBA;EAoJU,yBRlMmB;EQmMnB,YRtLqB;AF0mB/B;;AUzkBA;EA4Jc,4DAA8D;AVib5E;;AU7kBA;;EA+JU,6BAA6B;EAC7B,qBR9MmB;EQ+MnB,gBAAgB;EAChB,cRhNmB;AFmoB7B;;AUrlBA;EA2EM,yBRzHuB;EQ0HvB,yBAAyB;EACzB,YR9GyB;AF4nB/B;;AU3lBA;EAgFQ,yBCnByB;EDoBzB,yBAAyB;EACzB,YRnHuB;AFkoB/B;;AUjmBA;EAqFQ,yBAAyB;EACzB,YRvHuB;AFuoB/B;;AUtmBA;EAwFU,gDRtImB;AFwpB7B;;AU1mBA;EA2FQ,uBC9ByB;ED+BzB,yBAAyB;EACzB,YR9HuB;AFipB/B;;AUhnBA;;EAgGQ,yBR9IqB;EQ+IrB,yBAAyB;EACzB,gBAAgB;AVqhBxB;;AUvnBA;EAoGQ,uBRrIuB;EQsIvB,cRnJqB;AF0qB7B;;AU5nBA;EAwGU,yBC3CuB;AXmkBjC;;AUhoBA;;EA2GU,uBR5IqB;EQ6IrB,yBAAyB;EACzB,gBAAgB;EAChB,cR5JmB;AFsrB7B;;AUxoBA;EAiHU,4DAA4E;AV2hBtF;;AU5oBA;EAmHQ,6BAA6B;EAC7B,qBRlKqB;EQmKrB,cRnKqB;AFgsB7B;;AUlpBA;EA0HU,yBRxKmB;EQyKnB,qBRzKmB;EQ0KnB,YR7JqB;AFyrB/B;;AUxpBA;EA+HY,gEAA8D;AV6hB1E;;AU5pBA;EAqIc,4DAA4E;AV2hB1F;;AUhqBA;;EAwIU,6BAA6B;EAC7B,qBRvLmB;EQwLnB,gBAAgB;EAChB,cRzLmB;AFstB7B;;AUxqBA;EA6IQ,6BAA6B;EAC7B,mBR/KuB;EQgLvB,YRhLuB;AF+sB/B;;AU9qBA;EAoJU,uBRrLqB;EQsLrB,cRnMmB;AFiuB7B;;AUnrBA;EA4Jc,gEAA8D;AV2hB5E;;AUvrBA;;EA+JU,6BAA6B;EAC7B,mBRjMqB;EQkMrB,gBAAgB;EAChB,YRnMqB;AFguB/B;;AU/rBA;EA2EM,4BR9GwB;EQ+GxB,yBAAyB;EACzB,yBC7Ce;AXqqBrB;;AUrsBA;EAgFQ,yBCnByB;EDoBzB,yBAAyB;EACzB,yBClDa;AX2qBrB;;AU3sBA;EAqFQ,yBAAyB;EACzB,yBCtDa;AXgrBrB;;AUhtBA;EAwFU,mDR3HoB;AFuvB9B;;AUptBA;EA2FQ,yBC9ByB;ED+BzB,yBAAyB;EACzB,yBC7Da;AX0rBrB;;AU1tBA;;EAgGQ,4BRnIsB;EQoItB,yBAAyB;EACzB,gBAAgB;AV+nBxB;;AUjuBA;EAoGQ,oCCpEa;EDqEb,iBRxIsB;AFywB9B;;AUtuBA;EAwGU,oCC3CuB;AX6qBjC;;AU1uBA;;EA2GU,oCC3EW;ED4EX,yBAAyB;EACzB,gBAAgB;EAChB,iBRjJoB;AFqxB9B;;AUlvBA;EAiHU,sFAA4E;AVqoBtF;;AUtvBA;EAmHQ,6BAA6B;EAC7B,wBRvJsB;EQwJtB,iBRxJsB;AF+xB9B;;AU5vBA;EA0HU,4BR7JoB;EQ8JpB,wBR9JoB;EQ+JpB,yBC5FW;AXkuBrB;;AUlwBA;EA+HY,sEAA8D;AVuoB1E;;AUtwBA;EAqIc,sFAA4E;AVqoB1F;;AU1wBA;;EAwIU,6BAA6B;EAC7B,wBR5KoB;EQ6KpB,gBAAgB;EAChB,iBR9KoB;AFqzB9B;;AUlxBA;EA6IQ,6BAA6B;EAC7B,gCC9Ga;ED+Gb,yBC/Ga;AXwvBrB;;AUxxBA;EAoJU,oCCpHW;EDqHX,iBRxLoB;AFg0B9B;;AU7xBA;EA4Jc,sEAA8D;AVqoB5E;;AUjyBA;;EA+JU,6BAA6B;EAC7B,gCChIW;EDiIX,gBAAgB;EAChB,yBClIW;AXywBrB;;AUzyBA;EA2EM,yBRrHwB;EQsHxB,yBAAyB;EACzB,WC3CU;AX6wBhB;;AU/yBA;EAgFQ,yBCnByB;EDoBzB,yBAAyB;EACzB,WChDQ;AXmxBhB;;AUrzBA;EAqFQ,yBAAyB;EACzB,WCpDQ;AXwxBhB;;AU1zBA;EAwFU,gDRlIoB;AFw2B9B;;AU9zBA;EA2FQ,yBC9ByB;ED+BzB,yBAAyB;EACzB,WC3DQ;AXkyBhB;;AUp0BA;;EAgGQ,yBR1IsB;EQ2ItB,yBAAyB;EACzB,gBAAgB;AVyuBxB;;AU30BA;EAoGQ,sBClEQ;EDmER,cR/IsB;AF03B9B;;AUh1BA;EAwGU,yBC3CuB;AXuxBjC;;AUp1BA;;EA2GU,sBCzEM;ED0EN,yBAAyB;EACzB,gBAAgB;EAChB,cRxJoB;AFs4B9B;;AU51BA;EAiHU,0DAA4E;AV+uBtF;;AUh2BA;EAmHQ,6BAA6B;EAC7B,qBR9JsB;EQ+JtB,cR/JsB;AFg5B9B;;AUt2BA;EA0HU,yBRpKoB;EQqKpB,qBRrKoB;EQsKpB,WC1FM;AX00BhB;;AU52BA;EA+HY,gEAA8D;AVivB1E;;AUh3BA;EAqIc,0DAA4E;AV+uB1F;;AUp3BA;;EAwIU,6BAA6B;EAC7B,qBRnLoB;EQoLpB,gBAAgB;EAChB,cRrLoB;AFs6B9B;;AU53BA;EA6IQ,6BAA6B;EAC7B,kBC5GQ;ED6GR,WC7GQ;AXg2BhB;;AUl4BA;EAoJU,sBClHM;EDmHN,cR/LoB;AFi7B9B;;AUv4BA;EA4Jc,gEAA8D;AV+uB5E;;AU34BA;;EA+JU,6BAA6B;EAC7B,kBC9HM;ED+HN,gBAAgB;EAChB,WChIM;AXi3BhB;;AUn5BA;EA2EM,yBRvG4B;EQwG5B,yBAAyB;EACzB,WC3CU;AXu3BhB;;AUz5BA;EAgFQ,yBCnByB;EDoBzB,yBAAyB;EACzB,WChDQ;AX63BhB;;AU/5BA;EAqFQ,yBAAyB;EACzB,WCpDQ;AXk4BhB;;AUp6BA;EAwFU,iDRpHwB;AFo8BlC;;AUx6BA;EA2FQ,yBC9ByB;ED+BzB,yBAAyB;EACzB,WC3DQ;AX44BhB;;AU96BA;;EAgGQ,yBR5H0B;EQ6H1B,yBAAyB;EACzB,gBAAgB;AVm1BxB;;AUr7BA;EAoGQ,sBClEQ;EDmER,cRjI0B;AFs9BlC;;AU17BA;EAwGU,yBC3CuB;AXi4BjC;;AU97BA;;EA2GU,sBCzEM;ED0EN,yBAAyB;EACzB,gBAAgB;EAChB,cR1IwB;AFk+BlC;;AUt8BA;EAiHU,0DAA4E;AVy1BtF;;AU18BA;EAmHQ,6BAA6B;EAC7B,qBRhJ0B;EQiJ1B,cRjJ0B;AF4+BlC;;AUh9BA;EA0HU,yBRtJwB;EQuJxB,qBRvJwB;EQwJxB,WC1FM;AXo7BhB;;AUt9BA;EA+HY,gEAA8D;AV21B1E;;AU19BA;EAqIc,0DAA4E;AVy1B1F;;AU99BA;;EAwIU,6BAA6B;EAC7B,qBRrKwB;EQsKxB,gBAAgB;EAChB,cRvKwB;AFkgClC;;AUt+BA;EA6IQ,6BAA6B;EAC7B,kBC5GQ;ED6GR,WC7GQ;AX08BhB;;AU5+BA;EAoJU,sBClHM;EDmHN,cRjLwB;AF6gClC;;AUj/BA;EA4Jc,gEAA8D;AVy1B5E;;AUr/BA;;EA+JU,6BAA6B;EAC7B,kBC9HM;ED+HN,gBAAgB;EAChB,WChIM;AX29BhB;;AU7/BA;EAwKU,yBC/HsC;EDgItC,cCvH2D;AXg9BrE;;AUlgCA;EA4KY,yBC/GqB;EDgHrB,yBAAyB;EACzB,cC5HyD;AXs9BrE;;AUxgCA;EAiLY,yBCpHqB;EDqHrB,yBAAyB;EACzB,cCjIyD;AX49BrE;;AU9gCA;EA2EM,yBRrG4B;EQsG5B,yBAAyB;EACzB,WC3CU;AXk/BhB;;AUphCA;EAgFQ,yBCnByB;EDoBzB,yBAAyB;EACzB,WChDQ;AXw/BhB;;AU1hCA;EAqFQ,yBAAyB;EACzB,WCpDQ;AX6/BhB;;AU/hCA;EAwFU,kDRlHwB;AF6jClC;;AUniCA;EA2FQ,yBC9ByB;ED+BzB,yBAAyB;EACzB,WC3DQ;AXugChB;;AUziCA;;EAgGQ,yBR1H0B;EQ2H1B,yBAAyB;EACzB,gBAAgB;AV88BxB;;AUhjCA;EAoGQ,sBClEQ;EDmER,cR/H0B;AF+kClC;;AUrjCA;EAwGU,yBC3CuB;AX4/BjC;;AUzjCA;;EA2GU,sBCzEM;ED0EN,yBAAyB;EACzB,gBAAgB;EAChB,cRxIwB;AF2lClC;;AUjkCA;EAiHU,0DAA4E;AVo9BtF;;AUrkCA;EAmHQ,6BAA6B;EAC7B,qBR9I0B;EQ+I1B,cR/I0B;AFqmClC;;AU3kCA;EA0HU,yBRpJwB;EQqJxB,qBRrJwB;EQsJxB,WC1FM;AX+iChB;;AUjlCA;EA+HY,gEAA8D;AVs9B1E;;AUrlCA;EAqIc,0DAA4E;AVo9B1F;;AUzlCA;;EAwIU,6BAA6B;EAC7B,qBRnKwB;EQoKxB,gBAAgB;EAChB,cRrKwB;AF2nClC;;AUjmCA;EA6IQ,6BAA6B;EAC7B,kBC5GQ;ED6GR,WC7GQ;AXqkChB;;AUvmCA;EAoJU,sBClHM;EDmHN,cR/KwB;AFsoClC;;AU5mCA;EA4Jc,gEAA8D;AVo9B5E;;AUhnCA;;EA+JU,6BAA6B;EAC7B,kBC9HM;ED+HN,gBAAgB;EAChB,WChIM;AXslChB;;AUxnCA;EAwKU,yBC/HsC;EDgItC,cCvH2D;AX2kCrE;;AU7nCA;EA4KY,yBC/GqB;EDgHrB,yBAAyB;EACzB,cC5HyD;AXilCrE;;AUnoCA;EAiLY,yBCpHqB;EDqHrB,yBAAyB;EACzB,cCjIyD;AXulCrE;;AUzoCA;EA2EM,yBRtG4B;EQuG5B,yBAAyB;EACzB,WC3CU;AX6mChB;;AU/oCA;EAgFQ,yBCnByB;EDoBzB,yBAAyB;EACzB,WChDQ;AXmnChB;;AUrpCA;EAqFQ,yBAAyB;EACzB,WCpDQ;AXwnChB;;AU1pCA;EAwFU,kDRnHwB;AFyrClC;;AU9pCA;EA2FQ,yBC9ByB;ED+BzB,yBAAyB;EACzB,WC3DQ;AXkoChB;;AUpqCA;;EAgGQ,yBR3H0B;EQ4H1B,yBAAyB;EACzB,gBAAgB;AVykCxB;;AU3qCA;EAoGQ,sBClEQ;EDmER,cRhI0B;AF2sClC;;AUhrCA;EAwGU,yBC3CuB;AXunCjC;;AUprCA;;EA2GU,sBCzEM;ED0EN,yBAAyB;EACzB,gBAAgB;EAChB,cRzIwB;AFutClC;;AU5rCA;EAiHU,0DAA4E;AV+kCtF;;AUhsCA;EAmHQ,6BAA6B;EAC7B,qBR/I0B;EQgJ1B,cRhJ0B;AFiuClC;;AUtsCA;EA0HU,yBRrJwB;EQsJxB,qBRtJwB;EQuJxB,WC1FM;AX0qChB;;AU5sCA;EA+HY,gEAA8D;AVilC1E;;AUhtCA;EAqIc,0DAA4E;AV+kC1F;;AUptCA;;EAwIU,6BAA6B;EAC7B,qBRpKwB;EQqKxB,gBAAgB;EAChB,cRtKwB;AFuvClC;;AU5tCA;EA6IQ,6BAA6B;EAC7B,kBC5GQ;ED6GR,WC7GQ;AXgsChB;;AUluCA;EAoJU,sBClHM;EDmHN,cRhLwB;AFkwClC;;AUvuCA;EA4Jc,gEAA8D;AV+kC5E;;AU3uCA;;EA+JU,6BAA6B;EAC7B,kBC9HM;ED+HN,gBAAgB;EAChB,WChIM;AXitChB;;AUnvCA;EAwKU,yBC/HsC;EDgItC,cCvH2D;AXssCrE;;AUxvCA;EA4KY,yBC/GqB;EDgHrB,yBAAyB;EACzB,cC5HyD;AX4sCrE;;AU9vCA;EAiLY,yBCpHqB;EDqHrB,yBAAyB;EACzB,cCjIyD;AXktCrE;;AUpwCA;EA2EM,yBRxG4B;EQyG5B,yBAAyB;EACzB,WC3CU;AXwuChB;;AU1wCA;EAgFQ,yBCnByB;EDoBzB,yBAAyB;EACzB,WChDQ;AX8uChB;;AUhxCA;EAqFQ,yBAAyB;EACzB,WCpDQ;AXmvChB;;AUrxCA;EAwFU,kDRrHwB;AFszClC;;AUzxCA;EA2FQ,yBC9ByB;ED+BzB,yBAAyB;EACzB,WC3DQ;AX6vChB;;AU/xCA;;EAgGQ,yBR7H0B;EQ8H1B,yBAAyB;EACzB,gBAAgB;AVosCxB;;AUtyCA;EAoGQ,sBClEQ;EDmER,cRlI0B;AFw0ClC;;AU3yCA;EAwGU,yBC3CuB;AXkvCjC;;AU/yCA;;EA2GU,sBCzEM;ED0EN,yBAAyB;EACzB,gBAAgB;EAChB,cR3IwB;AFo1ClC;;AUvzCA;EAiHU,0DAA4E;AV0sCtF;;AU3zCA;EAmHQ,6BAA6B;EAC7B,qBRjJ0B;EQkJ1B,cRlJ0B;AF81ClC;;AUj0CA;EA0HU,yBRvJwB;EQwJxB,qBRxJwB;EQyJxB,WC1FM;AXqyChB;;AUv0CA;EA+HY,gEAA8D;AV4sC1E;;AU30CA;EAqIc,0DAA4E;AV0sC1F;;AU/0CA;;EAwIU,6BAA6B;EAC7B,qBRtKwB;EQuKxB,gBAAgB;EAChB,cRxKwB;AFo3ClC;;AUv1CA;EA6IQ,6BAA6B;EAC7B,kBC5GQ;ED6GR,WC7GQ;AX2zChB;;AU71CA;EAoJU,sBClHM;EDmHN,cRlLwB;AF+3ClC;;AUl2CA;EA4Jc,gEAA8D;AV0sC5E;;AUt2CA;;EA+JU,6BAA6B;EAC7B,kBC9HM;ED+HN,gBAAgB;EAChB,WChIM;AX40ChB;;AU92CA;EAwKU,yBC/HsC;EDgItC,cCvH2D;AXi0CrE;;AUn3CA;EA4KY,yBC/GqB;EDgHrB,yBAAyB;EACzB,cC5HyD;AXu0CrE;;AUz3CA;EAiLY,yBCpHqB;EDqHrB,yBAAyB;EACzB,cCjIyD;AX60CrE;;AU/3CA;EA2EM,yBRzG4B;EQ0G5B,yBAAyB;EACzB,yBC7Ce;AXq2CrB;;AUr4CA;EAgFQ,yBCnByB;EDoBzB,yBAAyB;EACzB,yBClDa;AX22CrB;;AU34CA;EAqFQ,yBAAyB;EACzB,yBCtDa;AXg3CrB;;AUh5CA;EAwFU,kDRtHwB;AFk7ClC;;AUp5CA;EA2FQ,yBC9ByB;ED+BzB,yBAAyB;EACzB,yBC7Da;AX03CrB;;AU15CA;;EAgGQ,yBR9H0B;EQ+H1B,yBAAyB;EACzB,gBAAgB;AV+zCxB;;AUj6CA;EAoGQ,oCCpEa;EDqEb,cRnI0B;AFo8ClC;;AUt6CA;EAwGU,oCC3CuB;AX62CjC;;AU16CA;;EA2GU,oCC3EW;ED4EX,yBAAyB;EACzB,gBAAgB;EAChB,cR5IwB;AFg9ClC;;AUl7CA;EAiHU,sFAA4E;AVq0CtF;;AUt7CA;EAmHQ,6BAA6B;EAC7B,qBRlJ0B;EQmJ1B,cRnJ0B;AF09ClC;;AU57CA;EA0HU,yBRxJwB;EQyJxB,qBRzJwB;EQ0JxB,yBC5FW;AXk6CrB;;AUl8CA;EA+HY,gEAA8D;AVu0C1E;;AUt8CA;EAqIc,sFAA4E;AVq0C1F;;AU18CA;;EAwIU,6BAA6B;EAC7B,qBRvKwB;EQwKxB,gBAAgB;EAChB,cRzKwB;AFg/ClC;;AUl9CA;EA6IQ,6BAA6B;EAC7B,gCC9Ga;ED+Gb,yBC/Ga;AXw7CrB;;AUx9CA;EAoJU,oCCpHW;EDqHX,cRnLwB;AF2/ClC;;AU79CA;EA4Jc,gEAA8D;AVq0C5E;;AUj+CA;;EA+JU,6BAA6B;EAC7B,gCChIW;EDiIX,gBAAgB;EAChB,yBClIW;AXy8CrB;;AUz+CA;EAwKU,yBC/HsC;EDgItC,cCvH2D;AX47CrE;;AU9+CA;EA4KY,yBC/GqB;EDgHrB,yBAAyB;EACzB,cC5HyD;AXk8CrE;;AUp/CA;EAiLY,yBCpHqB;EDqHrB,yBAAyB;EACzB,cCjIyD;AXw8CrE;;AU1/CA;EA2EM,yBRnG2B;EQoG3B,yBAAyB;EACzB,WC3CU;AX89ChB;;AUhgDA;EAgFQ,yBCnByB;EDoBzB,yBAAyB;EACzB,WChDQ;AXo+ChB;;AUtgDA;EAqFQ,yBAAyB;EACzB,WCpDQ;AXy+ChB;;AU3gDA;EAwFU,kDRhHuB;AFuiDjC;;AU/gDA;EA2FQ,yBC9ByB;ED+BzB,yBAAyB;EACzB,WC3DQ;AXm/ChB;;AUrhDA;;EAgGQ,yBRxHyB;EQyHzB,yBAAyB;EACzB,gBAAgB;AV07CxB;;AU5hDA;EAoGQ,sBClEQ;EDmER,cR7HyB;AFyjDjC;;AUjiDA;EAwGU,yBC3CuB;AXw+CjC;;AUriDA;;EA2GU,sBCzEM;ED0EN,yBAAyB;EACzB,gBAAgB;EAChB,cRtIuB;AFqkDjC;;AU7iDA;EAiHU,0DAA4E;AVg8CtF;;AUjjDA;EAmHQ,6BAA6B;EAC7B,qBR5IyB;EQ6IzB,cR7IyB;AF+kDjC;;AUvjDA;EA0HU,yBRlJuB;EQmJvB,qBRnJuB;EQoJvB,WC1FM;AX2hDhB;;AU7jDA;EA+HY,gEAA8D;AVk8C1E;;AUjkDA;EAqIc,0DAA4E;AVg8C1F;;AUrkDA;;EAwIU,6BAA6B;EAC7B,qBRjKuB;EQkKvB,gBAAgB;EAChB,cRnKuB;AFqmDjC;;AU7kDA;EA6IQ,6BAA6B;EAC7B,kBC5GQ;ED6GR,WC7GQ;AXijDhB;;AUnlDA;EAoJU,sBClHM;EDmHN,cR7KuB;AFgnDjC;;AUxlDA;EA4Jc,gEAA8D;AVg8C5E;;AU5lDA;;EA+JU,6BAA6B;EAC7B,kBC9HM;ED+HN,gBAAgB;EAChB,WChIM;AXkkDhB;;AUpmDA;EAwKU,yBC/HsC;EDgItC,cCvH2D;AXujDrE;;AUzmDA;EA4KY,yBC/GqB;EDgHrB,yBAAyB;EACzB,cC5HyD;AX6jDrE;;AU/mDA;EAiLY,yBCpHqB;EDqHrB,yBAAyB;EACzB,cCjIyD;AXmkDrE;;AUrnDA;EATE,kBR6BgB;EQ5BhB,kBRFc;AFooDhB;;AU1nDA;EANE,eRLW;AFyoDb;;AU9nDA;EAJE,kBRRc;AF8oDhB;;AUloDA;EAFE,iBRXa;AFmpDf;;AUtoDA;;EAgMI,uBRjO2B;EQkO3B,qBRvO0B;EQwO1B,gBAtNyB;EAuNzB,YAtNyB;AViqD7B;;AU9oDA;EAqMI,aAAa;EACb,WAAW;AV68Cf;;AUnpDA;EAwMI,6BAA6B;EAC7B,oBAAoB;AV+8CxB;;AUxpDA;ETvCE,kBAAkB;EAKhB,2BAAiC;EACjC,0BAAgC;ES8O9B,6BAA6B;AVk9CnC;;AU/pDA;EA+MI,4BRlP0B;EQmP1B,qBRtP0B;EQuP1B,cRzP0B;EQ0P1B,gBAAgB;EAChB,oBAAoB;AVo9CxB;;AUvqDA;EAqNI,uBR9LqB;EQ+LrB,gCAA0D;EAC1D,iCAA2D;AVs9C/D;;AUp9CA;EACE,mBAAmB;EACnB,aAAa;EACb,eAAe;EACf,2BAA2B;AVu9C7B;;AU39CA;EAMI,qBAAqB;AVy9CzB;;AU/9CA;ETzHI,oBSiIwC;AV29C5C;;AUn+CA;EAUI,sBAAsB;AV69C1B;;AUv+CA;EAYI,mBAAmB;AV+9CvB;;AU3+CA;EAlOE,kBR6BgB;EQ5BhB,kBRFc;AFmtDhB;;AUh/CA;EA7NE,kBRRc;AFytDhB;;AUp/CA;EA3NE,iBRXa;AF8tDf;;AUx/CA;EA0BQ,4BAA4B;EAC5B,yBAAyB;AVk+CjC;;AU7/CA;EA6BQ,6BAA6B;EAC7B,0BAA0B;ETvJ9B,kBSwJwC;AVo+C5C;;AUngDA;ETzHI,eS0JqC;AVs+CzC;;AUvgDA;EAoCQ,UAAU;AVu+ClB;;AU3gDA;EA0CQ,UAAU;AVq+ClB;;AU/gDA;EA4CU,UAAU;AVu+CpB;;AUnhDA;EA8CQ,YAAY;EACZ,cAAc;AVy+CtB;;AUxhDA;EAiDI,uBAAuB;AV2+C3B;;AU5hDA;EAoDQ,oBAAoB;EACpB,qBAAqB;AV4+C7B;;AUjiDA;EAuDI,yBAAyB;AV8+C7B;;AUriDA;EA0DQ,oBAAoB;EACpB,qBAAqB;AV++C7B;;AYhzDA;EACE,YAAY;EACZ,cAAc;EACd,kBAAkB;EAClB,WAAW;AZmzDb;;AYvzDA;EAMI,0BAA0B;EAC1B,kBV2CM;EU1CN,mBV0CM;EUzCN,WAAW;AZqzDf;;AChuDE;EW9FF;IAWI,gBAAuC;EZwzDzC;AACF;;AC5tDI;EWxGJ;IAcM,iBAAqE;EZ2zDzE;AACF;;ACntDI;EWvHJ;IAiBM,iBAAiE;EZ8zDrE;AACF;;ACnuDI;EW7GJ;IAoBM,iBAAqE;EZi0DzE;AACF;;AC1tDI;EW5HJ;IAuBM,iBAAiE;EZo0DrE;AACF;;Aa50DA;EAII,kBAAkB;Ab40DtB;;Aah1DA;;;;;;;EAcM,kBAAkB;Ab40DxB;;Aa11DA;;;;;;EAqBI,cXlC0B;EWmC1B,gBXEiB;EWDjB,kBAxC+B;Abs3DnC;;Aar2DA;EAyBI,cAAc;EACd,oBAAoB;Abg1DxB;;Aa12DA;EA4BM,eAAe;Abk1DrB;;Aa92DA;EA8BI,iBAAiB;EACjB,uBAAuB;Abo1D3B;;Aan3DA;EAiCM,oBAAoB;Abs1D1B;;Aav3DA;EAmCI,gBAAgB;EAChB,uBAAuB;Abw1D3B;;Aa53DA;EAsCM,oBAAoB;Ab01D1B;;Aah4DA;EAwCI,iBAAiB;EACjB,oBAAoB;Ab41DxB;;Aar4DA;EA2CI,kBAAkB;EAClB,uBAAuB;Ab81D3B;;Aa14DA;EA8CI,cAAc;EACd,kBAAkB;Abg2DtB;;Aa/4DA;EAiDI,4BXvD0B;EDmI1B,8BCtI0B;EW4D1B,qBAhEqC;Abk6DzC;;Aar5DA;EAqDI,4BAA4B;EZwE5B,gBYvEmC;EACnC,eAAe;Abo2DnB;;Aa35DA;EAyDM,wBAAwB;Abs2D9B;;Aa/5DA;EA2DQ,4BAA4B;Abw2DpC;;Aan6DA;EA6DQ,4BAA4B;Ab02DpC;;Aav6DA;EA+DQ,4BAA4B;Ab42DpC;;Aa36DA;EAiEQ,4BAA4B;Ab82DpC;;Aa/6DA;EAmEI,wBAAwB;EZ0DxB,gBYzDmC;EACnC,eAAe;Abg3DnB;;Aar7DA;EAuEM,uBAAuB;EACvB,iBAAiB;Abk3DvB;;Aa17DA;EA0EQ,uBAAuB;Abo3D/B;;Aa97DA;EZ6HI,gBYjDmC;Abs3DvC;;Aal8DA;EA8EI,gBAAgB;EAChB,iBAAiB;EACjB,kBAAkB;Abw3DtB;;Aax8DA;EAkFM,eAAe;Ab03DrB;;Aa58DA;EAoFM,kBAAkB;Ab43DxB;;Aah9DA;EAsFM,qBAAqB;Ab83D3B;;Aap9DA;EAwFM,kBAAkB;Abg4DxB;;Aax9DA;EZ2CE,iCAAiC;EYgD/B,gBAAgB;EAChB,qBAvG8B;EAwG9B,gBAAgB;EAChB,iBAAiB;Abk4DrB;;Aah+DA;;EAiGI,cAAc;Abo4DlB;;Aar+DA;EAmGI,WAAW;Abs4Df;;Aaz+DA;;EAsGM,yBX/GwB;EWgHxB,qBA/GmC;EAgHnC,qBA/GmC;EAgHnC,mBAAmB;Abw4DzB;;Aaj/DA;EA2GM,cXxHwB;AFkgE9B;;Aar/DA;EA6GQ,mBAAmB;Ab44D3B;;Aaz/DA;;EAiHQ,qBAtHsC;EAuHtC,cX/HsB;AF4gE9B;;Aa//DA;;EAsHQ,qBAzHsC;EA0HtC,cXpIsB;AFkhE9B;;AargEA;;EA6HY,sBAAsB;Ab64DlC;;Aa1gEA;EAgIM,aAAa;Ab84DnB;;Aa9gEA;EAmII,kBXhHY;AF+/DhB;;AalhEA;EAqII,kBXpHY;AFqgEhB;;AathEA;EAuII,iBXvHW;AF0gEf;;AcxiEA;EACE,mBAAmB;EACnB,oBAAoB;EACpB,uBAAuB;EACvB,cATsB;EAUtB,aAVsB;AdqjExB;;AchjEA;EAQI,YAZwB;EAaxB,WAbwB;AdyjE5B;;AcrjEA;EAWI,YAdyB;EAezB,WAfyB;Ad6jE7B;;Ac1jEA;EAcI,YAhBwB;EAiBxB,WAjBwB;AdikE5B;;AelkEA;EACE,cAAc;EACd,kBAAkB;AfqkEpB;;AevkEA;EAII,cAAc;EACd,YAAY;EACZ,WAAW;AfukEf;;Ae7kEA;EAQM,uBb6DmB;AF4gEzB;;AejlEA;EAUI,WAAW;Af2kEf;;AerlEA;;;;;;;;;;;;;;;;;EA+BM,YAAY;EACZ,WAAW;Af0kEjB;;Ae1mEA;EAmCI,iBAAiB;Af2kErB;;Ae9mEA;EAqCI,gBAAgB;Af6kEpB;;AelnEA;EAuCI,gBAAgB;Af+kEpB;;AetnEA;EAyCI,qBAAqB;AfilEzB;;Ae1nEA;EA2CI,gBAAgB;AfmlEpB;;Ae9nEA;EA6CI,mBAAmB;AfqlEvB;;AeloEA;EA+CI,gBAAgB;AfulEpB;;AetoEA;EAiDI,qBAAqB;AfylEzB;;Ae1oEA;EAmDI,iBAAiB;Af2lErB;;Ae9oEA;EAqDI,sBAAsB;Af6lE1B;;AelpEA;EAuDI,iBAAiB;Af+lErB;;AetpEA;EAyDI,sBAAsB;AfimE1B;;Ae1pEA;EA2DI,sBAAsB;AfmmE1B;;Ae9pEA;EA6DI,iBAAiB;AfqmErB;;AelqEA;EA+DI,iBAAiB;AfumErB;;AetqEA;EAmEM,YAAwB;EACxB,WAAuB;AfumE7B;;Ae3qEA;EAmEM,YAAwB;EACxB,WAAuB;Af4mE7B;;AehrEA;EAmEM,YAAwB;EACxB,WAAuB;AfinE7B;;AerrEA;EAmEM,YAAwB;EACxB,WAAuB;AfsnE7B;;Ae1rEA;EAmEM,YAAwB;EACxB,WAAuB;Af2nE7B;;Ae/rEA;EAmEM,YAAwB;EACxB,WAAuB;AfgoE7B;;AepsEA;EAmEM,aAAwB;EACxB,YAAuB;AfqoE7B;;AgBlsEA;EAEE,4BdE4B;EcD5B,kBdyDU;EcxDV,kBAAkB;EAEhB,sCAXoD;AhB8sExD;;AgBzsEA;EAUI,mBAAmB;EACnB,0BAA0B;AhBmsE9B;;AgB9sEA;EAaI,mBAAmB;AhBqsEvB;;AgBltEA;;EAgBI,iBdV2B;AFitE/B;;AgBvtEA;EAkBI,uBAAuB;AhBysE3B;;AgB3tEA;Ef+II,ae3H4B;EAC5B,kBAAkB;EAClB,WAAW;AhB2sEf;;AgBjuEA;;;EA0BI,mBAAmB;AhB6sEvB;;AgBvuEA;EAgCM,uBd1ByB;Ec2BzB,cdxCuB;AFmvE7B;;AgB5uEA;EAgCM,yBdvCuB;EcwCvB,Yd3ByB;AF2uE/B;;AgBjvEA;EAgCM,4Bd5BwB;Ec6BxB,yBLsCe;AX+qErB;;AgBtvEA;EAgCM,yBdnCwB;EcoCxB,WLwCU;AXkrEhB;;AgB3vEA;EAgCM,yBdrB4B;EcsB5B,WLwCU;AXurEhB;;AgBhwEA;EAuCU,yBLyCsC;EKxCtC,cLiD2D;AX4qErE;;AgBrwEA;EAgCM,yBdnB4B;EcoB5B,WLwCU;AXisEhB;;AgB1wEA;EAuCU,yBLyCsC;EKxCtC,cLiD2D;AXsrErE;;AgB/wEA;EAgCM,yBdpB4B;EcqB5B,WLwCU;AX2sEhB;;AgBpxEA;EAuCU,yBLyCsC;EKxCtC,cLiD2D;AXgsErE;;AgBzxEA;EAgCM,yBdtB4B;EcuB5B,WLwCU;AXqtEhB;;AgB9xEA;EAuCU,yBLyCsC;EKxCtC,cLiD2D;AX0sErE;;AgBnyEA;EAgCM,yBdvB4B;EcwB5B,yBLsCe;AXiuErB;;AgBxyEA;EAuCU,yBLyCsC;EKxCtC,cLiD2D;AXotErE;;AgB7yEA;EAgCM,yBdjB2B;EckB3B,WLwCU;AXyuEhB;;AgBlzEA;EAuCU,yBLyCsC;EKxCtC,cLiD2D;AX8tErE;;AiBxzEA;EAEE,qBAAqB;EACrB,wBAAwB;EACxB,YAAY;EACZ,uBf0DuB;EezDvB,cAAc;EACd,YfsBW;EerBX,gBAAgB;EAChB,UAAU;EACV,WAAW;AjB0zEb;;AiBp0EA;EAYI,yBfT2B;AFq0E/B;;AiBx0EA;EAcI,yBff0B;AF60E9B;;AiB50EA;EAgBI,yBfjB0B;AFi1E9B;;AiBh1EA;EAkBI,yBfnB0B;EeoB1B,YAAY;AjBk0EhB;;AiBr1EA;EAyBQ,uBflBuB;AFk1E/B;;AiBz1EA;EA2BQ,uBfpBuB;AFs1E/B;;AiB71EA;EA6BQ,uBftBuB;AF01E/B;;AiBj2EA;EA+BQ,mEAA2F;AjBs0EnG;;AiBr2EA;EAyBQ,yBf/BqB;AF+2E7B;;AiBz2EA;EA2BQ,yBfjCqB;AFm3E7B;;AiB72EA;EA6BQ,yBfnCqB;AFu3E7B;;AiBj3EA;EA+BQ,qEAA2F;AjBs1EnG;;AiBr3EA;EAyBQ,4BfpBsB;AFo3E9B;;AiBz3EA;EA2BQ,4BftBsB;AFw3E9B;;AiB73EA;EA6BQ,4BfxBsB;AF43E9B;;AiBj4EA;EA+BQ,wEAA2F;AjBs2EnG;;AiBr4EA;EAyBQ,yBf3BsB;AF24E9B;;AiBz4EA;EA2BQ,yBf7BsB;AF+4E9B;;AiB74EA;EA6BQ,yBf/BsB;AFm5E9B;;AiBj5EA;EA+BQ,qEAA2F;AjBs3EnG;;AiBr5EA;EAyBQ,yBfb0B;AF64ElC;;AiBz5EA;EA2BQ,yBff0B;AFi5ElC;;AiB75EA;EA6BQ,yBfjB0B;AFq5ElC;;AiBj6EA;EA+BQ,qEAA2F;AjBs4EnG;;AiBr6EA;EAyBQ,yBfX0B;AF25ElC;;AiBz6EA;EA2BQ,yBfb0B;AF+5ElC;;AiB76EA;EA6BQ,yBff0B;AFm6ElC;;AiBj7EA;EA+BQ,qEAA2F;AjBs5EnG;;AiBr7EA;EAyBQ,yBfZ0B;AF46ElC;;AiBz7EA;EA2BQ,yBfd0B;AFg7ElC;;AiB77EA;EA6BQ,yBfhB0B;AFo7ElC;;AiBj8EA;EA+BQ,qEAA2F;AjBs6EnG;;AiBr8EA;EAyBQ,yBfd0B;AF87ElC;;AiBz8EA;EA2BQ,yBfhB0B;AFk8ElC;;AiB78EA;EA6BQ,yBflB0B;AFs8ElC;;AiBj9EA;EA+BQ,qEAA2F;AjBs7EnG;;AiBr9EA;EAyBQ,yBff0B;AF+8ElC;;AiBz9EA;EA2BQ,yBfjB0B;AFm9ElC;;AiB79EA;EA6BQ,yBfnB0B;AFu9ElC;;AiBj+EA;EA+BQ,qEAA2F;AjBs8EnG;;AiBr+EA;EAyBQ,yBfTyB;AFy9EjC;;AiBz+EA;EA2BQ,yBfXyB;AF69EjC;;AiB7+EA;EA6BQ,yBfbyB;AFi+EjC;;AiBj/EA;EA+BQ,qEAA2F;AjBs9EnG;;AiBr/EA;EAkCI,gCAtCkC;UAsClC,wBAtCkC;EAuClC,2CAAmC;UAAnC,mCAAmC;EACnC,yCAAiC;UAAjC,iCAAiC;EACjC,yCAAiC;UAAjC,iCAAiC;EACjC,yBfnC2B;EeoC3B,qEAA0F;EAC1F,6BAA6B;EAC7B,4BAA4B;EAC5B,0BAA0B;AjBu9E9B;;AiBjgFA;EA4CM,6BAA6B;AjBy9EnC;;AiBrgFA;EA8CM,6BAA6B;AjB29EnC;;AiBzgFA;EAgDM,oBAAoB;AjB69E1B;;AiB7gFA;EAoDI,eftBY;AFm/EhB;;AiBjhFA;EAsDI,ef1BY;AFy/EhB;;AiBrhFA;EAwDI,cf7BW;AF8/Ef;;AiB/9EA;EACE;IACE,2BAA2B;EjBk+E7B;EiBj+EA;IACE,4BAA4B;EjBm+E9B;AACF;;AiBx+EA;EACE;IACE,2BAA2B;EjBk+E7B;EiBj+EA;IACE,4BAA4B;EjBm+E9B;AACF;;AkB/gFA;EAEE,uBhBd6B;EgBe7B,chBxB4B;AFyiF9B;;AkBphFA;;EAMI,yBhBvB0B;EgBwB1B,qBA9B6B;EA+B7B,qBA9B6B;EA+B7B,mBAAmB;AlBmhFvB;;AkB5hFA;;EAeQ,uBhB3BuB;EgB4BvB,mBhB5BuB;EgB6BvB,chB1CqB;AF4jF7B;;AkBniFA;;EAeQ,yBhBxCqB;EgByCrB,qBhBzCqB;EgB0CrB,YhB7BuB;AFsjF/B;;AkB1iFA;;EAeQ,4BhB7BsB;EgB8BtB,wBhB9BsB;EgB+BtB,yBPoCa;AX4/ErB;;AkBjjFA;;EAeQ,yBhBpCsB;EgBqCtB,qBhBrCsB;EgBsCtB,WPsCQ;AXigFhB;;AkBxjFA;;EAeQ,yBhBtB0B;EgBuB1B,qBhBvB0B;EgBwB1B,WPsCQ;AXwgFhB;;AkB/jFA;;EAeQ,yBhBpB0B;EgBqB1B,qBhBrB0B;EgBsB1B,WPsCQ;AX+gFhB;;AkBtkFA;;EAeQ,yBhBrB0B;EgBsB1B,qBhBtB0B;EgBuB1B,WPsCQ;AXshFhB;;AkB7kFA;;EAeQ,yBhBvB0B;EgBwB1B,qBhBxB0B;EgByB1B,WPsCQ;AX6hFhB;;AkBplFA;;EAeQ,yBhBxB0B;EgByB1B,qBhBzB0B;EgB0B1B,yBPoCa;AXsiFrB;;AkB3lFA;;EAeQ,yBhBlByB;EgBmBzB,qBhBnByB;EgBoBzB,WPsCQ;AX2iFhB;;AkBlmFA;;EAoBM,mBAAmB;EACnB,SAAS;AlBmlFf;;AkBxmFA;;EAuBM,yBhB9B4B;EgB+B5B,WP+BU;AXujFhB;;AkB9mFA;;;;EA2BQ,mBAAmB;AlB0lF3B;;AkBrnFA;;EA6BM,sBAAsB;AlB6lF5B;;AkB1nFA;EA+BI,chBpD0B;AFmpF9B;;AkB9nFA;EAiCM,mBAAmB;AlBimFzB;;AkBloFA;EAoCM,yBhB3C4B;EgB4C5B,WPkBU;AXglFhB;;AkBvoFA;;EAwCQ,mBAAmB;AlBomF3B;;AkB5oFA;;EA2CQ,kBPYQ;EOXR,mBAAmB;AlBsmF3B;;AkBlpFA;EA8CI,6BA5DqC;AlBoqFzC;;AkBtpFA;;EAiDM,qBApEgC;EAqEhC,chBvEwB;AFirF9B;;AkB5pFA;EAoDI,6BAhEqC;AlB4qFzC;;AkBhqFA;;EAuDM,qBAxEgC;EAyEhC,chB7EwB;AF2rF9B;;AkBtqFA;EA0DI,6BAvEqC;AlBurFzC;;AkB1qFA;;EA+DU,sBAAsB;AlBgnFhC;;AkB/qFA;;EAoEM,iBAAiB;AlBgnFvB;;AkBprFA;;EAyEU,wBAAwB;AlBgnFlC;;AkBzrFA;EA2EI,WAAW;AlBknFf;;AkB7rFA;EAgFU,yBhB7FoB;AF8sF9B;;AkBjsFA;EAqFY,yBhBlGkB;AFktF9B;;AkBrsFA;EAuFc,4BhBrGgB;AFutF9B;;AkBzsFA;;EA2FM,qBAAqB;AlBmnF3B;;AkB9sFA;EAgGU,yBhB7GoB;AF+tF9B;;AkBhnFA;EjB/DE,iCAAiC;EiBkEjC,cAAc;EACd,kBAAkB;EAClB,eAAe;AlBknFjB;;AmB7uFA;EACE,mBAAmB;EACnB,aAAa;EACb,eAAe;EACf,2BAA2B;AnBgvF7B;;AmBpvFA;EAMI,qBAAqB;AnBkvFzB;;AmBxvFA;ElByII,oBkBjIwC;AnBovF5C;;AmB5vFA;EAUI,sBAAsB;AnBsvF1B;;AmBhwFA;EAYI,mBAAmB;AnBwvFvB;;AmBpwFA;EAgBM,ejBcO;AF0uFb;;AmBxwFA;EAmBM,kBjBUU;AF+uFhB;;AmB5wFA;EAqBI,uBAAuB;AnB2vF3B;;AmBhxFA;EAuBM,qBAAqB;EACrB,oBAAoB;AnB6vF1B;;AmBrxFA;EA0BI,yBAAyB;AnB+vF7B;;AmBzxFA;EA6BQ,mBAAmB;AnBgwF3B;;AmB7xFA;EA+BQ,eAAe;AnBkwFvB;;AmBjyFA;ElByII,ekBvGmC;AnBmwFvC;;AmBryFA;ElByII,ckBrGqC;EAE/B,yBAAyB;EACzB,4BAA4B;AnBowFtC;;AmB3yFA;EA6CU,0BAA0B;EAC1B,6BAA6B;AnBkwFvC;;AmB7vFA;EACE,mBAAmB;EACnB,4BjB/C4B;EiBgD5B,kBjBQU;EiBPV,cjBvD4B;EiBwD5B,oBAAoB;EACpB,kBjB1Bc;EiB2Bd,WAAW;EACX,uBAAuB;EACvB,gBAAgB;EAChB,oBAAoB;EACpB,qBAAqB;EACrB,mBAAmB;AnBgwFrB;;AmB5wFA;ElBsFI,oBkBxEuC;ElBwEvC,uBkBvEyC;AnBkwF7C;;AmBjxFA;EAqBM,uBjBhEyB;EiBiEzB,cjB9EuB;AF80F7B;;AmBtxFA;EAqBM,yBjB7EuB;EiB8EvB,YjBjEyB;AFs0F/B;;AmB3xFA;EAqBM,4BjBlEwB;EiBmExB,yBRAe;AX0wFrB;;AmBhyFA;EAqBM,yBjBzEwB;EiB0ExB,WREU;AX6wFhB;;AmBryFA;EAqBM,yBjB3D4B;EiB4D5B,WREU;AXkxFhB;;AmB1yFA;EA4BU,yBRGsC;EQFtC,cRW2D;AXuwFrE;;AmB/yFA;EAqBM,yBjBzD4B;EiB0D5B,WREU;AX4xFhB;;AmBpzFA;EA4BU,yBRGsC;EQFtC,cRW2D;AXixFrE;;AmBzzFA;EAqBM,yBjB1D4B;EiB2D5B,WREU;AXsyFhB;;AmB9zFA;EA4BU,yBRGsC;EQFtC,cRW2D;AX2xFrE;;AmBn0FA;EAqBM,yBjB5D4B;EiB6D5B,WREU;AXgzFhB;;AmBx0FA;EA4BU,yBRGsC;EQFtC,cRW2D;AXqyFrE;;AmB70FA;EAqBM,yBjB7D4B;EiB8D5B,yBRAe;AX4zFrB;;AmBl1FA;EA4BU,yBRGsC;EQFtC,cRW2D;AX+yFrE;;AmBv1FA;EAqBM,yBjBvD2B;EiBwD3B,WREU;AXo0FhB;;AmB51FA;EA4BU,yBRGsC;EQFtC,cRW2D;AXyzFrE;;AmBj2FA;EAgCI,kBjBpDY;AFy3FhB;;AmBr2FA;EAkCI,ejBvDS;AF83Fb;;AmBz2FA;EAoCI,kBjB1DY;AFm4FhB;;AmB72FA;ElBsFI,qBkB/C0C;ElB+C1C,sBkB9C0C;AnB00F9C;;AmBl3FA;ElBsFI,qBkB5C0C;ElB4C1C,sBkB3C0C;AnB40F9C;;AmBv3FA;ElBsFI,qBkBzC0C;ElByC1C,sBkBxC0C;AnB80F9C;;AmB53FA;ElBsFI,gBkB7ImB;EAyGnB,UAAU;EACV,kBAAkB;EAClB,UAAU;AnB+0Fd;;AmBn4FA;EAuDM,8BAA8B;EAC9B,WAAW;EACX,cAAc;EACd,SAAS;EACT,kBAAkB;EAClB,QAAQ;EACR,0DAA0D;EAC1D,+BAA+B;AnBg1FrC;;AmB94FA;EAgEM,WAAW;EACX,UAAU;AnBk1FhB;;AmBn5FA;EAmEM,WAAW;EACX,UAAU;AnBo1FhB;;AmBx5FA;EAuEM,yBAAmD;AnBq1FzD;;AmB55FA;EAyEM,yBAAoD;AnBu1F1D;;AmBh6FA;EA2EI,uBjB9DqB;AFu5FzB;;AmBv1FA;EAEI,0BAA0B;AnBy1F9B;;AoB/8FA;;EAGE,sBAAsB;ApBi9FxB;;AoBp9FA;;;;EAMI,oBAAoB;ApBq9FxB;;AoB39FA;;EAQI,iBApBmB;ApB4+FvB;;AoBh+FA;;EAUI,iBArBmB;ApBg/FvB;;AoBr+FA;;EAYI,sBAAsB;ApB89F1B;;AoB59FA;EACE,clB5B4B;EkB+B5B,elBHW;EkBIX,gBlBKmB;EkBJnB,kBAnCuB;ApBggGzB;;AoBn+FA;EAQI,cApCwB;EAqCxB,oBApCyB;ApBmgG7B;;AoBx+FA;EAWI,oBAAoB;ApBi+FxB;;AoB5+FA;EAaI,oBA7B+B;ApBggGnC;;AoBh/FA;EAkBM,elBnBO;AFq/Fb;;AoBp/FA;EAkBM,iBlBlBS;AFw/Ff;;AoBx/FA;EAkBM,elBjBO;AF2/Fb;;AoB5/FA;EAkBM,iBlBhBS;AF8/Ff;;AoBhgGA;EAkBM,kBlBfU;AFigGhB;;AoBpgGA;EAkBM,elBdO;AFogGb;;AoBxgGA;EAkBM,kBlBbU;AFugGhB;;AoBx/FA;EACE,clB/C4B;EkBkD5B,kBlBrBc;EkBsBd,gBlBjBiB;EkBkBjB,iBA7CyB;ApBsiG3B;;AoB//FA;EAQI,clBvD0B;EkBwD1B,gBlBnBiB;AF8gGrB;;AoBpgGA;EAWI,oBA/C+B;ApB4iGnC;;AoBxgGA;EAgBM,elBrCO;AFiiGb;;AoB5gGA;EAgBM,iBlBpCS;AFoiGf;;AoBhhGA;EAgBM,elBnCO;AFuiGb;;AoBphGA;EAgBM,iBlBlCS;AF0iGf;;AoBxhGA;EAgBM,kBlBjCU;AF6iGhB;;AoB5hGA;EAgBM,elBhCO;AFgjGb;;AoBhiGA;EAgBM,kBlB/BU;AFmjGhB;;AqBnlGA;EACE,cAAc;EACd,eAAe;EACf,mBAAmB;EACnB,kBAAkB;EAClB,yBAAyB;ArBslG3B;;AqBplGA;EAEE,gBnB0BiB;EmBzBjB,eAAe;EACf,gBAAgB;EAChB,UAAU;ArBslGZ;;AqB3lGA;EAOI,cAAc;EACd,eAAe;ArBwlGnB;;AqBnlGA;EACE,mBAAmB;EACnB,4BnBf4B;EmBgB5B,uBnB0CuB;EmBzCvB,oBAAoB;EACpB,kBnBKc;EmBJd,WAAW;EACX,uBAAuB;EACvB,oBAAoB;EACpB,gBAAgB;EAChB,uBAAuB;EACvB,kBAAkB;EAClB,mBAAmB;ArBslGrB;;AsB5nGA,eAAA;ACuDA;EAxBE,uBrBhB6B;EqBiB7B,qBrBtB4B;EqBuB5B,kBrBoCU;EqBnCV,crB5B4B;AF8nG9B;;ACjkGI;EsB/BA,4BrB9B0B;AFkoG9B;;ACrkGI;EsB/BA,4BrB9B0B;AFsoG9B;;ACzkGI;EsB/BA,4BrB9B0B;AF0oG9B;;AC7kGI;EsB/BA,4BrB9B0B;AF8oG9B;;AuB/mGE;EAEE,qBrB9B0B;AF+oG9B;;AuBhnGE;EAIE,qBrBtB8B;EqBuB9B,kDrBvB8B;AFuoGlC;;AuB/mGE;;;;;EAEE,4BrBnC0B;EqBoC1B,wBrBpC0B;EqBqC1B,gBAAgB;EAChB,crB3C0B;AFgqG9B;;ACrmGI;;;;;EsBdE,+BrB7CwB;AFwqG9B;;AC7mGI;;;;;EsBdE,+BrB7CwB;AFgrG9B;;ACrnGI;;;;;EsBdE,+BrB7CwB;AFwrG9B;;AC7nGI;;;;;EsBdE,+BrB7CwB;AFgsG9B;;AwBlsGA;EAEE,2DtBN2B;EsBO3B,eAAe;EACf,WAAW;AxBosGb;;AwBnsGE;EACE,gBAAgB;AxBssGpB;;AwBlsGI;EACE,mBtBFyB;AFusG/B;;AwBtsGK;EAMG,mDtBPuB;AF2sG/B;;AwB1sGI;EACE,qBtBfuB;AF4tG7B;;AwB9sGK;EAMG,gDtBpBqB;AFguG7B;;AwBltGI;EACE,wBtBJwB;AFytG9B;;AwBttGK;EAMG,mDtBTsB;AF6tG9B;;AwB1tGI;EACE,qBtBXwB;AFwuG9B;;AwB9tGK;EAMG,gDtBhBsB;AF4uG9B;;AwBluGI;EACE,qBtBG4B;AFkuGlC;;AwBtuGK;EAMG,iDtBF0B;AFsuGlC;;AwB1uGI;EACE,qBtBK4B;AFwuGlC;;AwB9uGK;EAMG,kDtBA0B;AF4uGlC;;AwBlvGI;EACE,qBtBI4B;AFivGlC;;AwBtvGK;EAMG,kDtBD0B;AFqvGlC;;AwB1vGI;EACE,qBtBE4B;AF2vGlC;;AwB9vGK;EAMG,kDtBH0B;AF+vGlC;;AwBlwGI;EACE,qBtBC4B;AFowGlC;;AwBtwGK;EAMG,kDtBJ0B;AFwwGlC;;AwB1wGI;EACE,qBtBO2B;AFswGjC;;AwB9wGK;EAMG,kDtBEyB;AF0wGjC;;AwB1wGE;ErBoBA,kBDwBgB;ECvBhB,kBDPc;AFiwGhB;;AwB7wGE;ErBqBA,kBDXc;AFuwGhB;;AwB/wGE;ErBqBA,iBDda;AF4wGf;;AwBhxGE;EACE,cAAc;EACd,WAAW;AxBmxGf;;AwBlxGE;EACE,eAAe;EACf,WAAW;AxBqxGf;;AwBnxGA;EAGI,uBtB8BqB;EsB7BrB,gDAA4D;EAC5D,iDAA6D;AxBoxGjE;;AwBzxGA;EAOI,6BAA6B;EAC7B,yBAAyB;EACzB,gBAAgB;EAChB,eAAe;EACf,gBAAgB;AxBsxGpB;;AwBpxGA;EAEE,cAAc;EACd,eAAe;EACf,eAAe;EACf,2BrB/CkE;EqBgDlE,gBAAgB;AxBsxGlB;;AwB5xGA;EAQI,gBA1DsB;EA2DtB,eA1DqB;AxBk1GzB;;AwBjyGA;EAWI,eAAe;AxB0xGnB;;AwBryGA;EAcI,YAAY;AxB2xGhB;;AyB51GA;EACE,eAAe;EACf,qBAAqB;EACrB,iBAAiB;EACjB,kBAAkB;AzB+1GpB;;AyB91GE;EACE,eAAe;AzBi2GnB;;AyBh2GE;EACE,cvBF0B;AFq2G9B;;AyBl2GE;;;;;EAGE,cvBJ0B;EuBK1B,mBAAmB;AzBu2GvB;;AyBl2GA;ExB8HI,kBwB3HqC;AzBm2GzC;;A0Bt3GA;EACE,qBAAqB;EACrB,eAAe;EACf,kBAAkB;EAClB,mBAAmB;A1By3GrB;;A0B73GA;EAMI,avBHkB;AH83GtB;;A0Bj4GA;EAUM,qBxBU4B;EDkI9B,cyB3I+B;EAC7B,UAAU;A1B23GhB;;A0Bv4GA;EAeM,uBxBsDmB;EDyErB,iByB9HsC;A1B43G1C;;A0B54GA;EAmBI,eAAe;EACf,cAAc;EACd,cAAc;EACd,eAAe;EACf,aAAa;A1B63GjB;;A0Bp5GA;EAyBM,aAAa;A1B+3GnB;;A0Bx5GA;;EA4BM,wBxBjBwB;AFk5G9B;;A0B75GA;EzB8II,oByBhHwC;A1Bm4G5C;;A0Bj6GA;EAgCM,YAAY;EACZ,UAAU;A1Bq4GhB;;A0Bt6GA;EAmCQ,kBAAkB;A1Bu4G1B;;A0B16GA;EAuCM,qBxBnCwB;AF06G9B;;A0B96GA;EA6CQ,mBxBhCuB;AFq6G/B;;A0Bl7GA;EA+CQ,mBxBlCuB;AFy6G/B;;A0Bt7GA;EAkDU,qBfyDuB;AX+0GjC;;A0B17GA;EAuDU,mDxB1CqB;AFi7G/B;;A0B97GA;EA6CQ,qBxB7CqB;AFk8G7B;;A0Bl8GA;EA+CQ,qBxB/CqB;AFs8G7B;;A0Bt8GA;EAkDU,mBfyDuB;AX+1GjC;;A0B18GA;EAuDU,gDxBvDmB;AF88G7B;;A0B98GA;EA6CQ,wBxBlCsB;AFu8G9B;;A0Bl9GA;EA+CQ,wBxBpCsB;AF28G9B;;A0Bt9GA;EAkDU,qBfyDuB;AX+2GjC;;A0B19GA;EAuDU,mDxB5CoB;AFm9G9B;;A0B99GA;EA6CQ,qBxBzCsB;AF89G9B;;A0Bl+GA;EA+CQ,qBxB3CsB;AFk+G9B;;A0Bt+GA;EAkDU,qBfyDuB;AX+3GjC;;A0B1+GA;EAuDU,gDxBnDoB;AF0+G9B;;A0B9+GA;EA6CQ,qBxB3B0B;AFg+GlC;;A0Bl/GA;EA+CQ,qBxB7B0B;AFo+GlC;;A0Bt/GA;EAkDU,qBfyDuB;AX+4GjC;;A0B1/GA;EAuDU,iDxBrCwB;AF4+GlC;;A0B9/GA;EA6CQ,qBxBzB0B;AF8+GlC;;A0BlgHA;EA+CQ,qBxB3B0B;AFk/GlC;;A0BtgHA;EAkDU,qBfyDuB;AX+5GjC;;A0B1gHA;EAuDU,kDxBnCwB;AF0/GlC;;A0B9gHA;EA6CQ,qBxB1B0B;AF+/GlC;;A0BlhHA;EA+CQ,qBxB5B0B;AFmgHlC;;A0BthHA;EAkDU,qBfyDuB;AX+6GjC;;A0B1hHA;EAuDU,kDxBpCwB;AF2gHlC;;A0B9hHA;EA6CQ,qBxB5B0B;AFihHlC;;A0BliHA;EA+CQ,qBxB9B0B;AFqhHlC;;A0BtiHA;EAkDU,qBfyDuB;AX+7GjC;;A0B1iHA;EAuDU,kDxBtCwB;AF6hHlC;;A0B9iHA;EA6CQ,qBxB7B0B;AFkiHlC;;A0BljHA;EA+CQ,qBxB/B0B;AFsiHlC;;A0BtjHA;EAkDU,qBfyDuB;AX+8GjC;;A0B1jHA;EAuDU,kDxBvCwB;AF8iHlC;;A0B9jHA;EA6CQ,qBxBvByB;AF4iHjC;;A0BlkHA;EA+CQ,qBxBzByB;AFgjHjC;;A0BtkHA;EAkDU,qBfyDuB;AX+9GjC;;A0B1kHA;EAuDU,kDxBjCuB;AFwjHjC;;A0B9kHA;EvB0CE,kBDwBgB;ECvBhB,kBDPc;AF+iHhB;;A0BnlHA;EvB6CE,kBDXc;AFqjHhB;;A0BvlHA;EvB+CE,iBDda;AF0jHf;;A0B3lHA;EAkEM,qBxB5DwB;AFylH9B;;A0B/lHA;EAoEI,WAAW;A1B+hHf;;A0BnmHA;EAsEM,WAAW;A1BiiHjB;;A0BvmHA;EA0EM,aAAa;EACb,kBAAkB;EzB2EpB,cyB1E+B;EAC7B,YAAY;EACZ,eAAe;A1BiiHrB;;A0B/mHA;EAgFM,kBxB5CU;AF+kHhB;;A0BnnHA;EAkFM,kBxBhDU;AFqlHhB;;A0BvnHA;EAoFM,iBxBnDS;AF0lHf;;A2B9mHA;EAEE,oBAAoB;EACpB,aAAa;EACb,2BAA2B;EAC3B,kBAAkB;A3BgnHpB;;A2BrnHA;EAYQ,uBzBZuB;EyBavB,yBAAyB;EACzB,czB3BqB;AFwoH7B;;A2B3nHA;EAkBU,yBhB4EuB;EgB3EvB,yBAAyB;EACzB,czBjCmB;AF8oH7B;;A2BjoHA;EAwBU,yBAAyB;EACzB,+CzBzBqB;EyB0BrB,czBvCmB;AFopH7B;;A2BvoHA;EA8BU,yBhBgEuB;EgB/DvB,yBAAyB;EACzB,czB7CmB;AF0pH7B;;A2B7oHA;EAYQ,yBzBzBqB;EyB0BrB,yBAAyB;EACzB,YzBduB;AFmpH/B;;A2BnpHA;EAkBU,yBhB4EuB;EgB3EvB,yBAAyB;EACzB,YzBpBqB;AFypH/B;;A2BzpHA;EAwBU,yBAAyB;EACzB,4CzBtCmB;EyBuCnB,YzB1BqB;AF+pH/B;;A2B/pHA;EA8BU,uBhBgEuB;EgB/DvB,yBAAyB;EACzB,YzBhCqB;AFqqH/B;;A2BrqHA;EAYQ,4BzBdsB;EyBetB,yBAAyB;EACzB,yBhBmDa;AX0mHrB;;A2B3qHA;EAkBU,yBhB4EuB;EgB3EvB,yBAAyB;EACzB,yBhB6CW;AXgnHrB;;A2BjrHA;EAwBU,yBAAyB;EACzB,+CzB3BoB;EyB4BpB,yBhBuCW;AXsnHrB;;A2BvrHA;EA8BU,yBhBgEuB;EgB/DvB,yBAAyB;EACzB,yBhBiCW;AX4nHrB;;A2B7rHA;EAYQ,yBzBrBsB;EyBsBtB,yBAAyB;EACzB,WhBqDQ;AXgoHhB;;A2BnsHA;EAkBU,yBhB4EuB;EgB3EvB,yBAAyB;EACzB,WhB+CM;AXsoHhB;;A2BzsHA;EAwBU,yBAAyB;EACzB,4CzBlCoB;EyBmCpB,WhByCM;AX4oHhB;;A2B/sHA;EA8BU,yBhBgEuB;EgB/DvB,yBAAyB;EACzB,WhBmCM;AXkpHhB;;A2BrtHA;EAYQ,yBzBP0B;EyBQ1B,yBAAyB;EACzB,WhBqDQ;AXwpHhB;;A2B3tHA;EAkBU,yBhB4EuB;EgB3EvB,yBAAyB;EACzB,WhB+CM;AX8pHhB;;A2BjuHA;EAwBU,yBAAyB;EACzB,6CzBpBwB;EyBqBxB,WhByCM;AXoqHhB;;A2BvuHA;EA8BU,yBhBgEuB;EgB/DvB,yBAAyB;EACzB,WhBmCM;AX0qHhB;;A2B7uHA;EAYQ,yBzBL0B;EyBM1B,yBAAyB;EACzB,WhBqDQ;AXgrHhB;;A2BnvHA;EAkBU,yBhB4EuB;EgB3EvB,yBAAyB;EACzB,WhB+CM;AXsrHhB;;A2BzvHA;EAwBU,yBAAyB;EACzB,8CzBlBwB;EyBmBxB,WhByCM;AX4rHhB;;A2B/vHA;EA8BU,yBhBgEuB;EgB/DvB,yBAAyB;EACzB,WhBmCM;AXksHhB;;A2BrwHA;EAYQ,yBzBN0B;EyBO1B,yBAAyB;EACzB,WhBqDQ;AXwsHhB;;A2B3wHA;EAkBU,yBhB4EuB;EgB3EvB,yBAAyB;EACzB,WhB+CM;AX8sHhB;;A2BjxHA;EAwBU,yBAAyB;EACzB,8CzBnBwB;EyBoBxB,WhByCM;AXotHhB;;A2BvxHA;EA8BU,yBhBgEuB;EgB/DvB,yBAAyB;EACzB,WhBmCM;AX0tHhB;;A2B7xHA;EAYQ,yBzBR0B;EyBS1B,yBAAyB;EACzB,WhBqDQ;AXguHhB;;A2BnyHA;EAkBU,yBhB4EuB;EgB3EvB,yBAAyB;EACzB,WhB+CM;AXsuHhB;;A2BzyHA;EAwBU,yBAAyB;EACzB,8CzBrBwB;EyBsBxB,WhByCM;AX4uHhB;;A2B/yHA;EA8BU,yBhBgEuB;EgB/DvB,yBAAyB;EACzB,WhBmCM;AXkvHhB;;A2BrzHA;EAYQ,yBzBT0B;EyBU1B,yBAAyB;EACzB,yBhBmDa;AX0vHrB;;A2B3zHA;EAkBU,yBhB4EuB;EgB3EvB,yBAAyB;EACzB,yBhB6CW;AXgwHrB;;A2Bj0HA;EAwBU,yBAAyB;EACzB,8CzBtBwB;EyBuBxB,yBhBuCW;AXswHrB;;A2Bv0HA;EA8BU,yBhBgEuB;EgB/DvB,yBAAyB;EACzB,yBhBiCW;AX4wHrB;;A2B70HA;EAYQ,yBzBHyB;EyBIzB,yBAAyB;EACzB,WhBqDQ;AXgxHhB;;A2Bn1HA;EAkBU,yBhB4EuB;EgB3EvB,yBAAyB;EACzB,WhB+CM;AXsxHhB;;A2Bz1HA;EAwBU,yBAAyB;EACzB,8CzBhBuB;EyBiBvB,WhByCM;AX4xHhB;;A2B/1HA;EA8BU,yBhBgEuB;EgB/DvB,yBAAyB;EACzB,WhBmCM;AXkyHhB;;A2Br2HA;EAmCI,kBzBZY;AFk1HhB;;A2Bz2HA;EAqCI,kBzBhBY;AFw1HhB;;A2B72HA;EAwCQ,eAAe;A3By0HvB;;A2Bj3HA;EA0CI,iBzBtBW;AFi2Hf;;A2Br3HA;EA6CQ,eAAe;A3B40HvB;;A2Bz3HA;EAiDM,6BAA6B;EAC7B,0BAA0B;A3B40HhC;;A2B93HA;EAoDM,4BAA4B;EAC5B,yBAAyB;A3B80H/B;;A2Bn4HA;EAwDQ,kBzBFI;AFi1HZ;;A2Bv4HA;EA0DQ,aAAa;A3Bi1HrB;;A2B34HA;EA6DM,sBAAsB;A3Bk1H5B;;A2B/4HA;EA+DM,sBAAsB;EACtB,YAAY;EACZ,gBAAgB;A3Bo1HtB;;A2Br5HA;EAmEM,uBAAuB;A3Bs1H7B;;A2Bz5HA;EAqEM,aAAa;EACb,YAAY;A3Bw1HlB;;A2B95HA;EAwEQ,eAAe;A3B01HvB;;A2Bl6HA;EA2EQ,eAAe;A3B21HvB;;A2Bt6HA;EA8EQ,eAAe;A3B41HvB;;A2B16HA;EAiFQ,eAAe;A3B61HvB;;A2B96HA;EAoFQ,0BAA4C;A3B81HpD;;A2Bl7HA;EAsFQ,0BzBhCI;EyBiCJ,uBAAuB;A3Bg2H/B;;A2Bv7HA;EAyFI,uBAAuB;A3Bk2H3B;;A2B37HA;EA4FM,WAAW;A3Bm2HjB;;A2B/7HA;EA8FM,YAAY;EACZ,eAAe;A3Bq2HrB;;A2Bp8HA;EAiGI,yBAAyB;A3Bu2H7B;;A2Bx8HA;EAmGM,0BAA4C;A3By2HlD;;A2B58HA;EAqGM,0BzB/CM;EyBgDN,2BAA2B;EAC3B,SAAS;A3B22Hf;;A2Bz2HA;EACE,oBAAoB;EACpB,aAAa;EACb,eAAe;EACf,2BAA2B;EAC3B,gBAAgB;EAChB,kBAAkB;A3B42HpB;;A2Bl3HA;EASM,yBhBpB2B;EgBqB3B,czB5HwB;AFy+H9B;;A2Bv3HA;EAYM,qBhBvB2B;AXs4HjC;;A2B33HA;EAeM,yBhB1B2B;EgB2B3B,czBlIwB;AFk/H9B;;A2Bh4HA;EAkBM,qBhB7B2B;AX+4HjC;;A2Bh3HA;EACE,YAAY;EACZ,OAAO;EACP,UAAU;EACV,aAAa;EACb,kBAAkB;EAClB,MAAM;EACN,WAAW;A3Bm3Hb;;A2Bj3HA;;EAGE,qBzB9I4B;EyB+I5B,kBzBpFU;EyBqFV,cAAc;EACd,iBAAiB;EACjB,kBAAkB;EAClB,mBAAmB;A3Bm3HrB;;A2Bj3HA;EACE,4BzBnJ4B;EyBoJ5B,czB1J4B;AF8gI9B;;A2Bl3HA;EACE,qBzB1J4B;EyB2J5B,mBA5J4B;EA6J5B,2BA5JoC;EA6JpC,cAAc;EACd,eA7JwB;EA8JxB,gBAAgB;EAChB,mBAAmB;EACnB,uBAAuB;A3Bq3HzB;;A2Bn3HA;EACE,mBAAmB;EACnB,aAAa;EACb,WAAW;EACX,uBAAuB;E1BjCrB,mB0BkCmC;EACrC,UAAU;A3Bs3HZ;;A2B53HA;EAQI,eAAe;A3Bw3HnB;;A4BtiIA;EACE,c1BF4B;E0BG5B,cAAc;EACd,e1B2BW;E0B1BX,gB1BiCe;AFwgIjB;;A4B7iIA;EAMI,oBAAoB;A5B2iIxB;;A4BjjIA;EASI,kB1BsBY;AFshIhB;;A4BrjIA;EAWI,kB1BkBY;AF4hIhB;;A4BzjIA;EAaI,iB1BeW;AFiiIf;;A4B9iIA;EACE,cAAc;EACd,kB1Bcc;E0Bbd,mBAAmB;A5BijIrB;;A4BpjIA;EAOM,Y1BdyB;AF+jI/B;;A4BxjIA;EAOM,c1B3BuB;AFglI7B;;A4B5jIA;EAOM,iB1BhBwB;AFykI9B;;A4BhkIA;EAOM,c1BvBwB;AFolI9B;;A4BpkIA;EAOM,c1BT4B;AF0kIlC;;A4BxkIA;EAOM,c1BP4B;AF4kIlC;;A4B5kIA;EAOM,c1BR4B;AFilIlC;;A4BhlIA;EAOM,c1BV4B;AFulIlC;;A4BplIA;EAOM,c1BX4B;AF4lIlC;;A4BxlIA;EAOM,c1BL2B;AF0lIjC;;A4BjlIA;EAEI,sBAAsB;A5BmlI1B;;A4BrlIA;EAKI,aAAa;EACb,2BAA2B;A5BolI/B;;A4B1lIA;E3B+GI,kB2BtGwC;A5BqlI5C;;A4B9lIA;;;EAcU,gBAAgB;A5BslI1B;;A4BpmIA;;;EAoBY,6BAA6B;EAC7B,0BAA0B;A5BslItC;;A4B3mIA;;;EA8BY,4BAA4B;EAC5B,yBAAyB;A5BmlIrC;;A4BlnIA;;;;;EAyCY,UAAU;A5BilItB;;A4B1nIA;;;;;;;;;EA8CY,UAAU;A5BwlItB;;A4BtoIA;;;;;;;;;EAgDc,UAAU;A5BkmIxB;;A4BlpIA;EAkDQ,YAAY;EACZ,cAAc;A5BomItB;;A4BvpIA;EAqDM,uBAAuB;A5BsmI7B;;A4B3pIA;EAuDM,yBAAyB;A5BwmI/B;;A4B/pIA;EA0DQ,YAAY;EACZ,cAAc;A5BymItB;;A4BpqIA;EA6DI,aAAa;EACb,2BAA2B;A5B2mI/B;;A4BzqIA;EAgEM,cAAc;A5B6mIpB;;A4B7qIA;EAkEQ,gBAAgB;E3B6CpB,qB2B5C2C;A5B+mI/C;;A4BlrIA;EAqEQ,YAAY;EACZ,cAAc;A5BinItB;;A4BvrIA;EAwEM,uBAAuB;A5BmnI7B;;A4B3rIA;EA0EM,yBAAyB;A5BqnI/B;;A4B/rIA;EA4EM,eAAe;A5BunIrB;;A4BnsIA;EAgFU,sBAAsB;A5BunIhC;;A4BvsIA;EAkFQ,uBAAuB;A5BynI/B;;A4B3sIA;EAoFQ,gBAAgB;A5B2nIxB;;AC3pIE;E2BpDF;IAuFM,aAAa;E5B6nIjB;AACF;;A4B5nIA;EAEI,kBAAkB;A5B8nItB;;ACzqIE;E2ByCF;IAII,qBAAqB;E5BioIvB;AACF;;AC3qIE;E2BqCF;IAMI,aAAa;IACb,YAAY;IACZ,cAAc;I3Bcd,oB2BbsC;IACtC,iBAAiB;E5BqoInB;E4B/oIF;IAYM,kB1BhGU;I0BiGV,oBAAoB;E5BsoIxB;E4BnpIF;IAeM,oBAAoB;E5BuoIxB;E4BtpIF;IAiBM,kB1BvGU;I0BwGV,oBAAoB;E5BwoIxB;E4B1pIF;IAoBM,iB1B3GS;I0B4GT,oBAAoB;E5ByoIxB;AACF;;A4BxoIA;EAEI,gBAAgB;A5B0oIpB;;ACxsIE;E2B4DF;IAII,aAAa;IACb,aAAa;IACb,YAAY;IACZ,cAAc;E5B6oIhB;E4BppIF;IASM,gBAAgB;E5B8oIpB;E4BvpIF;IAWM,cAAc;E5B+oIlB;E4B1pIF;IAaQ,YAAY;E5BgpIlB;E4B7pIF;I3BDI,qB2BgB2C;E5BipI7C;AACF;;A4BhpIA;EACE,sBAAsB;EACtB,WAAW;EACX,e1BhIW;E0BiIX,kBAAkB;EAClB,mBAAmB;A5BmpIrB;;A4BxpIA;;;EAaU,c1BxKoB;AFyzI9B;;A4B9pIA;;;EAeQ,kB1B3IQ;AFgyIhB;;A4BpqIA;;;EAiBQ,kB1B/IQ;AFwyIhB;;A4B1qIA;;;EAmBQ,iB1BlJO;AF+yIf;;A4BhrIA;EAqBM,c1B7KwB;E0B8KxB,azBnLgB;EyBoLhB,oBAAoB;EACpB,kBAAkB;EAClB,MAAM;EACN,YzBvLgB;EyBwLhB,UAAU;A5B+pIhB;;A4B1rIA;;EA+BM,mBzB5LgB;AH41ItB;;A4B/rIA;EAiCM,OAAO;A5BkqIb;;A4BnsIA;;EAqCM,oBzBlMgB;AHq2ItB;;A4BxsIA;EAuCM,QAAQ;A5BqqId;;A4B5sIA;EA2CM,6BAA6B;E3BrD/B,c2BsD+B;EAC7B,YAAY;EACZ,UAAU;A5BqqIhB;;A4BntIA;EAgDM,kB1B5KU;AFm1IhB;;A4BvtIA;EAkDM,kB1BhLU;AFy1IhB;;A4B3tIA;EAoDM,iB1BnLS;AF81If;;A6Bj4IA,qBAAA;ACSA;EAGE,e5ByBW;E4BxBX,mBAAmB;A9B03IrB;;A8B93IA;EAMI,mBAAmB;EACnB,c5BM8B;E4BL9B,aAAa;EACb,uBAAuB;EACvB,iBAduC;A9B04I3C;;A8Bt4IA;EAYM,c5BfwB;AF64I9B;;A8B14IA;EAcI,mBAAmB;EACnB,aAAa;A9Bg4IjB;;A8B/4IA;E7BuII,e6BtHoC;A9Bk4IxC;;A8Bn5IA;EAoBQ,c5BvBsB;E4BwBtB,eAAe;EACf,oBAAoB;A9Bm4I5B;;A8Bz5IA;EAwBM,c5BxBwB;E4ByBxB,iBAAiB;A9Bq4IvB;;A8B95IA;;EA4BI,uBAAuB;EACvB,aAAa;EACb,eAAe;EACf,2BAA2B;A9Bu4I/B;;A8Bt6IA;E7BuII,mB6BrGuC;A9Bw4I3C;;A8B16IA;E7BuII,kB6BnGuC;A9B04I3C;;A8B96IA;;EAyCM,uBAAuB;A9B04I7B;;A8Bn7IA;;EA6CM,yBAAyB;A9B24I/B;;A8Bx7IA;EAgDI,kB5BnBY;AF+5IhB;;A8B57IA;EAkDI,kB5BvBY;AFq6IhB;;A8Bh8IA;EAoDI,iB5B1BW;AF06If;;A8Bp8IA;EAwDM,iBAAiB;A9Bg5IvB;;A8Bx8IA;EA2DM,iBAAiB;A9Bi5IvB;;A8B58IA;EA8DM,iBAAiB;A9Bk5IvB;;A8Bh9IA;EAiEM,iBAAiB;A9Bm5IvB;;A+Bx8IA;EACE,uB7BP6B;E6BQ7B,sBApBmB;EAqBnB,0F7BtB2B;E6BuB3B,c7BlB4B;E6BmB5B,eAAe;EACf,gBAvBoB;EAwBpB,kBAAkB;A/B28IpB;;A+Bz8IA;EACE,6BAzBwC;EA0BxC,oBAAoB;EACpB,kD7B/B2B;E6BgC3B,aAAa;A/B48If;;A+B18IA;EACE,mBAAmB;EACnB,c7BhC4B;E6BiC5B,aAAa;EACb,YAAY;EACZ,gB7BGe;E6BFf,qBAlCgC;A/B++IlC;;A+Bn9IA;EAQI,uBAAuB;A/B+8I3B;;A+B78IA;EACE,mBAAmB;EACnB,eAAe;EACf,aAAa;EACb,uBAAuB;EACvB,qBA3CgC;A/B2/IlC;;A+B98IA;EACE,cAAc;EACd,kBAAkB;A/Bi9IpB;;A+B/8IA;EACE,6BA9CyC;EA+CzC,eA9C2B;A/BggJ7B;;A+Bh9IA;EACE,6BA/CwC;EAgDxC,6B7BpD6B;E6BqD7B,oBAAoB;EACpB,aAAa;A/Bm9If;;A+Bj9IA;EACE,mBAAmB;EACnB,aAAa;EACb,aAAa;EACb,YAAY;EACZ,cAAc;EACd,uBAAuB;EACvB,gBAzD2B;A/B6gJ7B;;A+B39IA;E9B6EI,+BCrI2B;AFuhJ/B;;A+Bl9IA;EAEI,qB7BlCkB;AFs/ItB;;AgCnhJA;EACE,oBAAoB;EACpB,kBAAkB;EAClB,mBAAmB;AhCshJrB;;AgCzhJA;EAOM,cAAc;AhCshJpB;;AgC7hJA;EAUM,UAAU;EACV,QAAQ;AhCuhJd;;AgCliJA;EAcM,YAAY;EACZ,mBA9BuB;EA+BvB,oBAAoB;EACpB,SAAS;AhCwhJf;;AgCthJA;EACE,aAAa;E/BiHX,O+BhHqB;EACvB,gBAzC6B;EA0C7B,gBAtC2B;EAuC3B,kBAAkB;EAClB,SAAS;EACT,WApCqB;AhC6jJvB;;AgCvhJA;EACE,uB9BjC6B;E8BkC7B,kB9BoBU;E8BnBV,0F9BhD2B;E8BiD3B,sBA9CsC;EA+CtC,mBA9CmC;AhCwkJrC;;AgB5jJgB;EgBqCd,c9BhD4B;E8BiD5B,cAAc;EACd,mBAAmB;EACnB,gBAAgB;EAChB,sBAAsB;EACtB,kBAAkB;AhC2hJpB;;AgCzhJA;;E/BkFI,mB+BhFmC;EACrC,mBAAmB;EACnB,mBAAmB;EACnB,WAAW;AhC4hJb;;AgCjiJA;;EAOI,4B9BxD0B;E8ByD1B,c9BpEyB;AFmmJ7B;;AgCviJA;;EAUI,yB9BlD8B;E8BmD9B,WrBSY;AXyhJhB;;AgChiJA;EACE,yB9BjE6B;E8BkE7B,YAAY;EACZ,cAAc;EACd,WAAW;EACX,gBAAgB;AhCmiJlB;;AiCjnJA;EAEE,mBAAmB;EACnB,8BAA8B;AjCmnJhC;;AiCtnJA;EAKI,kB/B8DQ;AFujJZ;;AiC1nJA;EAOI,qBAAqB;EACrB,mBAAmB;AjCunJvB;;AiC/nJA;EAWI,aAAa;AjCwnJjB;;AiCnoJA;;EAcM,aAAa;AjC0nJnB;;AiCxoJA;EAgBM,aAAa;AjC4nJnB;;AiC5oJA;EAmBQ,gBAAgB;EhC2HpB,qBgChJqC;AjCmpJzC;;AiCjpJA;EAsBQ,YAAY;AjC+nJpB;;AClkJE;EgCnFF;IAyBI,aAAa;EjCioJf;EiC1pJF;IA4BQ,YAAY;EjCioJlB;AACF;;AiChoJA;EACE,mBAAmB;EACnB,aAAa;EACb,gBAAgB;EAChB,YAAY;EACZ,cAAc;EACd,uBAAuB;AjCmoJzB;;AiCzoJA;;EASI,gBAAgB;AjCqoJpB;;AC7lJE;EgCjDF;IAaM,sBA7CmC;EjCmrJvC;AACF;;AiCroJA;;EAEE,gBAAgB;EAChB,YAAY;EACZ,cAAc;AjCwoJhB;;AiC5oJA;;EAQM,YAAY;AjCyoJlB;;AC3mJE;EgCtCF;;IhCiGI,qBgChJqC;EjCssJvC;AACF;;AiC1oJA;EACE,mBAAmB;EACnB,2BAA2B;AjC6oJ7B;;AC3nJE;EgCpBF;IAMM,kBAAkB;EjC8oJtB;AACF;;AC7nJE;EgCxBF;IAQI,aAAa;EjCkpJf;AACF;;AiCjpJA;EACE,mBAAmB;EACnB,yBAAyB;AjCopJ3B;;ACxoJE;EgCdF;IAKI,aAAa;EjCspJf;AACF;;AkC/tJA;EACE,uBAAuB;EACvB,aAAa;EACb,mBAAmB;AlCkuJrB;;AkCruJA;EAKI,sBAAsB;AlCouJ1B;;AkCzuJA;EAOI,8ChCD0B;EgCE1B,aAAa;EACb,oBAAoB;AlCsuJxB;;AkC/uJA;;EAYM,qBAAqB;AlCwuJ3B;;AkCpvJA;EAcM,mBAAmB;AlC0uJzB;;AkCxvJA;EAgBQ,kBAAkB;AlC4uJ1B;;AkC5vJA;EAkBI,8ChCZ0B;EgCa1B,gBAtBgB;EAuBhB,iBAvBgB;AlCqwJpB;;AkClwJA;EAwBM,kBA1BsB;EA2BtB,mBA3BsB;AlCywJ5B;;AkC5uJA;;EAEE,gBAAgB;EAChB,YAAY;EACZ,cAAc;AlC+uJhB;;AkC7uJA;EjC2GI,kBiC/IgB;AlCqxJpB;;AkC9uJA;EjCwGI,iBiC/IgB;AlCyxJpB;;AkC/uJA;EACE,gBAAgB;EAChB,YAAY;EACZ,cAAc;EACd,mBAAmB;AlCkvJrB;;AChtJE;EiCtCF;IAQI,gBAAgB;ElCmvJlB;AACF;;AmCrxJA;EACE,ejCkBW;AFswJb;;AmCzxJA;EAII,kBjCgBY;AFywJhB;;AmC7xJA;EAMI,kBjCYY;AF+wJhB;;AmCjyJA;EAQI,iBjCSW;AFoxJf;;AmC3xJA;EACE,iBArB0B;AnCmzJ5B;;AmC/xJA;EAGI,kBjCqCc;EiCpCd,cjCzB0B;EiC0B1B,cAAc;EACd,qBAzBiC;AnCyzJrC;;AmCtyJA;EAQM,4BjCvBwB;EiCwBxB,cjC/BwB;AFi0J9B;;AmC3yJA;EAYM,yBjClB4B;EiCmB5B,WxByCU;AX0vJhB;;AmChzJA;ElCoHI,8BCtI0B;EiCmCxB,cAnC0B;ElCsI5B,oBkCrIkC;AnCu0JtC;;AmClyJA;EACE,cjCzC4B;EiC0C5B,iBApC2B;EAqC3B,qBApC+B;EAqC/B,yBAAyB;AnCqyJ3B;;AmCzyJA;EAMI,eAtCoB;AnC60JxB;;AmC7yJA;EAQI,kBAxCoB;AnCi1JxB;;AoC50JA;EAEE,4BlCV4B;EkCW5B,kBlC6CU;EkC5CV,elCYW;AFk0Jb;;AoCl1JA;EAMI,mBAAmB;ApCg1JvB;;AoCt1JA;EAQI,mBAAmB;EACnB,0BAA0B;ApCk1J9B;;AoC31JA;EAYI,kBlCKY;AF80JhB;;AoC/1JA;EAcI,kBlCCY;AFo1JhB;;AoCn2JA;EAgBI,iBlCFW;AFy1Jf;;AoCv2JA;EAsCM,uBAH+C;ApCw0JrD;;AoC32JA;EAwCQ,uBlC9CuB;EkC+CvB,clC5DqB;AFm4J7B;;AoCh3JA;EA2CQ,mBlCjDuB;AF03J/B;;AoCp3JA;EAsCM,yBAH+C;ApCq1JrD;;AoCx3JA;EAwCQ,yBlC3DqB;EkC4DrB,YlC/CuB;AFm4J/B;;AoC73JA;EA2CQ,qBlC9DqB;AFo5J7B;;AoCj4JA;EAsCM,yBAH+C;ApCk2JrD;;AoCr4JA;EAwCQ,4BlChDsB;EkCiDtB,yBzBkBa;AX+0JrB;;AoC14JA;EA2CQ,wBlCnDsB;AFs5J9B;;AoC94JA;EAsCM,yBAH+C;ApC+2JrD;;AoCl5JA;EAwCQ,yBlCvDsB;EkCwDtB,WzBoBQ;AX01JhB;;AoCv5JA;EA2CQ,qBlC1DsB;AF06J9B;;AoC35JA;EAsCM,yBzB8B0C;AX21JhD;;AoC/5JA;EAwCQ,yBlCzC0B;EkC0C1B,WzBoBQ;AXu2JhB;;AoCp6JA;EA2CQ,qBlC5C0B;EkC6C1B,czBiC6D;AX41JrE;;AoCz6JA;EAsCM,yBzB8B0C;AXy2JhD;;AoC76JA;EAwCQ,yBlCvC0B;EkCwC1B,WzBoBQ;AXq3JhB;;AoCl7JA;EA2CQ,qBlC1C0B;EkC2C1B,czBiC6D;AX02JrE;;AoCv7JA;EAsCM,yBzB8B0C;AXu3JhD;;AoC37JA;EAwCQ,yBlCxC0B;EkCyC1B,WzBoBQ;AXm4JhB;;AoCh8JA;EA2CQ,qBlC3C0B;EkC4C1B,czBiC6D;AXw3JrE;;AoCr8JA;EAsCM,yBzB8B0C;AXq4JhD;;AoCz8JA;EAwCQ,yBlC1C0B;EkC2C1B,WzBoBQ;AXi5JhB;;AoC98JA;EA2CQ,qBlC7C0B;EkC8C1B,czBiC6D;AXs4JrE;;AoCn9JA;EAsCM,yBzB8B0C;AXm5JhD;;AoCv9JA;EAwCQ,yBlC3C0B;EkC4C1B,yBzBkBa;AXi6JrB;;AoC59JA;EA2CQ,qBlC9C0B;EkC+C1B,czBiC6D;AXo5JrE;;AoCj+JA;EAsCM,yBzB8B0C;AXi6JhD;;AoCr+JA;EAwCQ,yBlCrCyB;EkCsCzB,WzBoBQ;AX66JhB;;AoC1+JA;EA2CQ,qBlCxCyB;EkCyCzB,czBiC6D;AXk6JrE;;AoCj8JA;EACE,mBAAmB;EACnB,yBlC9D4B;EkC+D5B,0BAAgE;EAChE,WzBWc;EyBVd,aAAa;EACb,gBlC7Be;EkC8Bf,8BAA8B;EAC9B,iBAAiB;EACjB,mBAtEiC;EAuEjC,kBAAkB;ApCo8JpB;;AoC98JA;EAYI,YAAY;EACZ,cAAc;EnCgEd,mBmC/DsC;ApCs8J1C;;AoCp9JA;EAgBI,eAjEgC;EAkEhC,yBAAyB;EACzB,0BAA0B;ApCw8J9B;;AoCt8JA;EACE,qBlC9E4B;EkC+E5B,kBlCpBU;EkCqBV,mBAAmB;EACnB,uBAjFmC;EAkFnC,clCrF4B;EkCsF5B,qBAjFiC;ApC0hKnC;;AoC/8JA;;EASI,uBlCjF2B;AF4hK/B;;AoCp9JA;EAWI,6BAlFgD;ApC+hKpD;;AqC/gKA;EAEE,mBAAmB;EACnB,aAAa;EACb,sBAAsB;EACtB,uBAAuB;EACvB,gBAAgB;EAChB,eAAe;EACf,WAxCU;ArCyjKZ;;AqCzhKA;EAWI,aAAa;ArCkhKjB;;AqChhKA;EAEE,wCnC7C2B;AF+jK7B;;AqChhKA;;EAEE,cA9CgC;EA+ChC,+BAA0D;EAC1D,cAAc;EACd,kBAAkB;EAClB,WAAW;ArCmhKb;;ACjgKE;EoCxBF;;IASI,cAAc;IACd,8BAA0D;IAC1D,YAxDuB;ErC8kKzB;AACF;;AqCrhKA;EAEE,gBAAgB;EAChB,YAxD2B;EAyD3B,eAAe;EpCsFb,WoC9IoB;EA0DtB,SAzDoB;EA0DpB,WA5D2B;ArCmlK7B;;AqCrhKA;EACE,aAAa;EACb,sBAAsB;EACtB,8BAAgD;EAChD,gBAAgB;EAChB,uBAAuB;ArCwhKzB;;AqCthKA;;EAEE,mBAAmB;EACnB,4BnCpE4B;EmCqE5B,aAAa;EACb,cAAc;EACd,2BAA2B;EAC3B,aApE4B;EAqE5B,kBAAkB;ArCyhKpB;;AqCvhKA;EACE,gCnC/E4B;EmCgF5B,2BnCpBgB;EmCqBhB,4BnCrBgB;AF+iKlB;;AqCxhKA;EACE,cnCxF4B;EmCyF5B,YAAY;EACZ,cAAc;EACd,iBnC9Da;EmC+Db,cA7E8B;ArCwmKhC;;AqCzhKA;EACE,8BnC/BgB;EmCgChB,+BnChCgB;EmCiChB,6BnC7F4B;AFynK9B;;AqC/hKA;EpC4CI,mBoCtCuC;ArC6hK3C;;AqC3hKA;EpC9CE,iCAAiC;EoCgDjC,uBnC/F6B;EmCgG7B,YAAY;EACZ,cAAc;EACd,cAAc;EACd,aAtF4B;ArConK9B;;AsCxlKA;EACE,uBpC1C6B;EoC2C7B,mBAvDqB;EAwDrB,kBAAkB;EAClB,WAtDW;AtCipKb;;AsC/lKA;EASM,uBpClDyB;EoCmDzB,cpChEuB;AF0pK7B;;AsCpmKA;;EAcU,cpCpEmB;AF+pK7B;;AsCzmKA;;;;EAoBY,yB3BiCqB;E2BhCrB,cpC3EiB;AFuqK7B;;AsCjnKA;EAwBY,qBpC9EiB;AF2qK7B;;AsCrnKA;EA0BQ,cpChFqB;AF+qK7B;;ACxmKE;EqCjBF;;;;IAgCY,cpCtFiB;EFurK3B;EsCjoKF;;;;;;;;;;IAsCc,yB3BemB;I2BdnB,cpC7Fe;EFosK3B;EsC9oKF;;IA0Cc,qBpChGe;EFwsK3B;EsClpKF;;;IA8CU,yB3BOuB;I2BNvB,cpCrGmB;EF8sK3B;EsCxpKF;IAmDc,uBpC5FiB;IoC6FjB,cpC1Ge;EFktK3B;AACF;;AsC7pKA;EASM,yBpC/DuB;EoCgEvB,YpCnDyB;AF2sK/B;;AsClqKA;;EAcU,YpCvDqB;AFgtK/B;;AsCvqKA;;;;EAoBY,uB3BiCqB;E2BhCrB,YpC9DmB;AFwtK/B;;AsC/qKA;EAwBY,mBpCjEmB;AF4tK/B;;AsCnrKA;EA0BQ,YpCnEuB;AFguK/B;;ACtqKE;EqCjBF;;;;IAgCY,YpCzEmB;EFwuK7B;EsC/rKF;;;;;;;;;;IAsCc,uB3BemB;I2BdnB,YpChFiB;EFqvK7B;EsC5sKF;;IA0Cc,mBpCnFiB;EFyvK7B;EsChtKF;;;IA8CU,uB3BOuB;I2BNvB,YpCxFqB;EF+vK7B;EsCttKF;IAmDc,yBpCzGe;IoC0Gf,YpC7FiB;EFmwK7B;AACF;;AsC3tKA;EASM,4BpCpDwB;EoCqDxB,yB3Bce;AXwsKrB;;AsChuKA;;EAcU,yB3BUW;AX6sKrB;;AsCruKA;;;;EAoBY,yB3BiCqB;E2BhCrB,yB3BGS;AXqtKrB;;AsC7uKA;EAwBY,gC3BAS;AXytKrB;;AsCjvKA;EA0BQ,yB3BFa;AX6tKrB;;ACpuKE;EqCjBF;;;;IAgCY,yB3BRS;EXquKnB;EsC7vKF;;;;;;;;;;IAsCc,yB3BemB;I2BdnB,yB3BfO;EXkvKnB;EsC1wKF;;IA0Cc,gC3BlBO;EXsvKnB;EsC9wKF;;;IA8CU,yB3BOuB;I2BNvB,yB3BvBW;EX4vKnB;EsCpxKF;IAmDc,4BpC9FgB;IoC+FhB,yB3B5BO;EXgwKnB;AACF;;AsCzxKA;EASM,yBpC3DwB;EoC4DxB,W3BgBU;AXowKhB;;AsC9xKA;;EAcU,W3BYM;AXywKhB;;AsCnyKA;;;;EAoBY,yB3BiCqB;E2BhCrB,W3BKI;AXixKhB;;AsC3yKA;EAwBY,kB3BEI;AXqxKhB;;AsC/yKA;EA0BQ,W3BAQ;AXyxKhB;;AClyKE;EqCjBF;;;;IAgCY,W3BNI;EXiyKd;EsC3zKF;;;;;;;;;;IAsCc,yB3BemB;I2BdnB,W3BbE;EX8yKd;EsCx0KF;;IA0Cc,kB3BhBE;EXkzKd;EsC50KF;;;IA8CU,yB3BOuB;I2BNvB,W3BrBM;EXwzKd;EsCl1KF;IAmDc,yBpCrGgB;IoCsGhB,W3B1BE;EX4zKd;AACF;;AsCv1KA;EASM,yBpC7C4B;EoC8C5B,W3BgBU;AXk0KhB;;AsC51KA;;EAcU,W3BYM;AXu0KhB;;AsCj2KA;;;;EAoBY,yB3BiCqB;E2BhCrB,W3BKI;AX+0KhB;;AsCz2KA;EAwBY,kB3BEI;AXm1KhB;;AsC72KA;EA0BQ,W3BAQ;AXu1KhB;;ACh2KE;EqCjBF;;;;IAgCY,W3BNI;EX+1Kd;EsCz3KF;;;;;;;;;;IAsCc,yB3BemB;I2BdnB,W3BbE;EX42Kd;EsCt4KF;;IA0Cc,kB3BhBE;EXg3Kd;EsC14KF;;;IA8CU,yB3BOuB;I2BNvB,W3BrBM;EXs3Kd;EsCh5KF;IAmDc,yBpCvFoB;IoCwFpB,W3B1BE;EX03Kd;AACF;;AsCr5KA;EASM,yBpC3C4B;EoC4C5B,W3BgBU;AXg4KhB;;AsC15KA;;EAcU,W3BYM;AXq4KhB;;AsC/5KA;;;;EAoBY,yB3BiCqB;E2BhCrB,W3BKI;AX64KhB;;AsCv6KA;EAwBY,kB3BEI;AXi5KhB;;AsC36KA;EA0BQ,W3BAQ;AXq5KhB;;AC95KE;EqCjBF;;;;IAgCY,W3BNI;EX65Kd;EsCv7KF;;;;;;;;;;IAsCc,yB3BemB;I2BdnB,W3BbE;EX06Kd;EsCp8KF;;IA0Cc,kB3BhBE;EX86Kd;EsCx8KF;;;IA8CU,yB3BOuB;I2BNvB,W3BrBM;EXo7Kd;EsC98KF;IAmDc,yBpCrFoB;IoCsFpB,W3B1BE;EXw7Kd;AACF;;AsCn9KA;EASM,yBpC5C4B;EoC6C5B,W3BgBU;AX87KhB;;AsCx9KA;;EAcU,W3BYM;AXm8KhB;;AsC79KA;;;;EAoBY,yB3BiCqB;E2BhCrB,W3BKI;AX28KhB;;AsCr+KA;EAwBY,kB3BEI;AX+8KhB;;AsCz+KA;EA0BQ,W3BAQ;AXm9KhB;;AC59KE;EqCjBF;;;;IAgCY,W3BNI;EX29Kd;EsCr/KF;;;;;;;;;;IAsCc,yB3BemB;I2BdnB,W3BbE;EXw+Kd;EsClgLF;;IA0Cc,kB3BhBE;EX4+Kd;EsCtgLF;;;IA8CU,yB3BOuB;I2BNvB,W3BrBM;EXk/Kd;EsC5gLF;IAmDc,yBpCtFoB;IoCuFpB,W3B1BE;EXs/Kd;AACF;;AsCjhLA;EASM,yBpC9C4B;EoC+C5B,W3BgBU;AX4/KhB;;AsCthLA;;EAcU,W3BYM;AXigLhB;;AsC3hLA;;;;EAoBY,yB3BiCqB;E2BhCrB,W3BKI;AXygLhB;;AsCniLA;EAwBY,kB3BEI;AX6gLhB;;AsCviLA;EA0BQ,W3BAQ;AXihLhB;;AC1hLE;EqCjBF;;;;IAgCY,W3BNI;EXyhLd;EsCnjLF;;;;;;;;;;IAsCc,yB3BemB;I2BdnB,W3BbE;EXsiLd;EsChkLF;;IA0Cc,kB3BhBE;EX0iLd;EsCpkLF;;;IA8CU,yB3BOuB;I2BNvB,W3BrBM;EXgjLd;EsC1kLF;IAmDc,yBpCxFoB;IoCyFpB,W3B1BE;EXojLd;AACF;;AsC/kLA;EASM,yBpC/C4B;EoCgD5B,yB3Bce;AX4jLrB;;AsCplLA;;EAcU,yB3BUW;AXikLrB;;AsCzlLA;;;;EAoBY,yB3BiCqB;E2BhCrB,yB3BGS;AXykLrB;;AsCjmLA;EAwBY,gC3BAS;AX6kLrB;;AsCrmLA;EA0BQ,yB3BFa;AXilLrB;;ACxlLE;EqCjBF;;;;IAgCY,yB3BRS;EXylLnB;EsCjnLF;;;;;;;;;;IAsCc,yB3BemB;I2BdnB,yB3BfO;EXsmLnB;EsC9nLF;;IA0Cc,gC3BlBO;EX0mLnB;EsCloLF;;;IA8CU,yB3BOuB;I2BNvB,yB3BvBW;EXgnLnB;EsCxoLF;IAmDc,yBpCzFoB;IoC0FpB,yB3B5BO;EXonLnB;AACF;;AsC7oLA;EASM,yBpCzC2B;EoC0C3B,W3BgBU;AXwnLhB;;AsClpLA;;EAcU,W3BYM;AX6nLhB;;AsCvpLA;;;;EAoBY,yB3BiCqB;E2BhCrB,W3BKI;AXqoLhB;;AsC/pLA;EAwBY,kB3BEI;AXyoLhB;;AsCnqLA;EA0BQ,W3BAQ;AX6oLhB;;ACtpLE;EqCjBF;;;;IAgCY,W3BNI;EXqpLd;EsC/qLF;;;;;;;;;;IAsCc,yB3BemB;I2BdnB,W3BbE;EXkqLd;EsC5rLF;;IA0Cc,kB3BhBE;EXsqLd;EsChsLF;;;IA8CU,yB3BOuB;I2BNvB,W3BrBM;EX4qLd;EsCtsLF;IAmDc,yBpCnFmB;IoCoFnB,W3B1BE;EXgrLd;AACF;;AsC3sLA;EAsDI,oBAAoB;EACpB,aAAa;EACb,mBA7GmB;EA8GnB,WAAW;AtCypLf;;AsCltLA;EA2DI,gCpCtG0B;AFiwL9B;;AsCttLA;EALE,OAAO;EACP,eAAe;EACf,QAAQ;EACR,WA/CiB;AtC8wLnB;;AsC7tLA;EAgEI,SAAS;AtCiqLb;;AsCjuLA;EAkEM,iCpC7GwB;AFgxL9B;;AsCruLA;EAoEI,MAAM;AtCqqLV;;AsCnqLA;;EAGI,oBA9HmB;AtCmyLvB;;AsCxqLA;;EAKI,uBAhImB;AtCwyLvB;;AsCtqLA;;EAEE,oBAAoB;EACpB,aAAa;EACb,cAAc;EACd,mBAvIqB;AtCgzLvB;;AsCvqLA;EAIM,6BAA6B;AtCuqLnC;;AsCrqLA;ErCpFE,iCAAiC;EqCsFjC,gBAAgB;EAChB,gBAAgB;EAChB,kBAAkB;AtCwqLpB;;AsCtqLA;EACE,cpClJ4B;EDoB5B,eAAe;EACf,cAAc;EACd,eqC1BqB;ErC2BrB,kBAAkB;EAClB,cqC5BqB;ErC6InB,iBqCWkC;AtC6qLtC;;ACxyLE;EACE,8BAA8B;EAC9B,cAAc;EACd,WAAW;EACX,qBAAqB;EACrB,kBAAkB;EAClB,wBAAwB;EACxB,yBCiCQ;EDhCR,yDAAyD;EACzD,oCC0Ba;EDzBb,WAAW;AD2yLf;;AC1yLI;EACE,oBAAoB;AD6yL1B;;AC5yLI;EACE,oBAAoB;AD+yL1B;;AC9yLI;EACE,oBAAoB;ADizL1B;;AChzLE;EACE,qCAAiC;ADmzLrC;;AC/yLM;EACE,wCAAwC;ADkzLhD;;ACjzLM;EACE,UAAU;ADozLlB;;ACnzLM;EACE,0CAA0C;ADszLlD;;AsCptLA;EACE,aAAa;AtCutLf;;AsCrtLA;;EAEE,cpC3J4B;EoC4J5B,cAAc;EACd,gBAAgB;EAChB,uBAAuB;EACvB,kBAAkB;AtCwtLpB;;AsC9tLA;;EASM,qBAAqB;EACrB,sBAAsB;AtC0tL5B;;AsCxtLA;;EAEE,eAAe;AtC2tLjB;;AsC7tLA;;;;;EAOI,yBpCrK0B;EoCsK1B,cpC9J8B;AF43LlC;;AsC5tLA;EACE,YAAY;EACZ,cAAc;AtC+tLhB;;AsCjuLA;EAII,mBA5KgC;AtC64LpC;;AsCruLA;EAMI,UAAU;AtCmuLd;;AsCzuLA;EAQI,YAAY;EACZ,cAAc;AtCquLlB;;AsC9uLA;EAWI,oCAAoC;EACpC,mBA/LmB;EAgMnB,kCAAkC;AtCuuLtC;;AsCpvLA;EAgBM,6BApLyC;EAqLzC,4BpCjL4B;AFy5LlC;;AsCzvLA;EAmBM,6BApL0C;EAqL1C,4BpCpL4B;EoCqL5B,0BApLuC;EAqLvC,wBApLqC;EAqLrC,cpCvL4B;EoCwL5B,kCAAwE;AtC0uL9E;;AsCxuLA;EACE,YAAY;EACZ,cAAc;AtC2uLhB;;AsCzuLA;ErCpEI,oBqCqEoC;AtC4uLxC;;AsC7uLA;EAII,qBpClM8B;EoCmM9B,oBAAoB;ErCjEpB,cqCkE6B;AtC6uLjC;;AsC3uLA;EACE,mBAAmB;EACnB,sBAAsB;EACtB,mBAAmB;AtC8uLrB;;AsCjvLA;EAKI,oBAAoB;EACpB,qBAAqB;AtCgvLzB;;AsC9uLA;EACE,4BpCxN4B;EoCyN5B,YAAY;EACZ,aAAa;EACb,WA9LyB;EA+LzB,gBAAgB;AtCivLlB;;AC74LE;EqCrBF;IAqLI,cAAc;EtCkvLhB;EsCjvLA;;IAGI,mBAAmB;IACnB,aAAa;EtCkvLjB;EsCjvLA;IAEI,aAAa;EtCkvLjB;EsC10LF;IA0FI,uBpCxO2B;IoCyO3B,4CpCtPyB;IoCuPzB,iBAAiB;EtCmvLnB;EsCtvLA;IAKI,cAAc;EtCovLlB;EsClvLA;IA1MA,OAAO;IACP,eAAe;IACf,QAAQ;IACR,WA/CiB;EtC8+LjB;EsCxvLA;IAKI,SAAS;EtCsvLb;EsC3vLA;IAOM,4CpClQqB;EFy/L3B;EsC9vLA;IASI,MAAM;EtCwvLV;EsCjwLA;IrC/LA,iCAAiC;IqC6M3B,iCAA2C;IAC3C,cAAc;EtCuvLpB;EsCtvLA;;IAGI,oBA7QiB;EtCogMrB;EsC1vLA;;IAKI,uBA/QiB;EtCwgMrB;AACF;;ACn8LE;EqC4MA;;;;IAIE,oBAAoB;IACpB,aAAa;EtC2vLf;EsC79LF;IAoOI,mBAzRmB;EtCqhMrB;EsC7vLA;IAGI,kBAzR0B;EtCshM9B;EsChwLA;;IAMM,mBAAmB;EtC8vLzB;EsCpwLA;;IASM,kBpC/NI;EF89LV;EsCxwLA;;;;IAgBQ,wCAAwC;EtC8vLhD;EsC9wLA;IAuBU,wCAAwC;EtC0vLlD;EsCjxLA;IA4BU,4BpC1SkB;IoC2SlB,cpCtTiB;EF8iM3B;EsCrxLA;IA+BU,4BpC7SkB;IoC8SlB,cpCrSsB;EF8hMhC;EsC55LF;IAqKI,aAAa;EtC0vLf;EsCv5LF;;IAgKI,mBAAmB;IACnB,aAAa;EtC2vLf;EsCt4LF;IA8IM,oBAAoB;EtC2vLxB;EsC7vLA;IAKM,oDAAoD;EtC2vL1D;EsChwLA;IAOM,gCpC/TsB;IoCgUtB,0BAAkE;IAClE,gBAAgB;IAChB,YAAY;IACZ,4CpC3UqB;IoC4UrB,SAAS;EtC4vLf;EsCxwLA;IAkBM,cAAc;EtCyvLpB;EsCxvLM;IAEE,UAAU;IACV,oBAAoB;IACpB,wBAAwB;EtCyvLhC;EsCr7LF;IA8LI,YAAY;IACZ,cAAc;EtC0vLhB;EsCzvLA;IACE,2BAA2B;IrC9M3B,kBqC+MoC;EtC2vLtC;EsC1vLA;IACE,yBAAyB;IrCjNzB,iBqCkNoC;EtC4vLtC;EsCl4LF;IAwII,uBpCrV2B;IoCsV3B,8BpC/Rc;IoCgSd,+BpChSc;IoCiSd,6BpC7V0B;IoC8V1B,2CpCtWyB;IoCuWzB,aAAa;IACb,mBAAmB;IrClNnB,OqCmNuB;IACvB,eAAe;IACf,kBAAkB;IAClB,SAAS;IACT,WAhVkB;EtC6kMpB;EsCh5LF;IAqJM,sBAAsB;IACtB,mBAAmB;EtC8vLvB;EsC7wLA;IrCnNE,mBqCoOuC;EtC+vLzC;EsChxLA;IAoBM,4BpC1WsB;IoC2WtB,cpCtXqB;EFqnM3B;EsCpxLA;IAuBM,4BpC7WsB;IoC8WtB,cpCrW0B;EFqmMhC;EsC/vLE;IAEE,kBpCxTY;IoCyTZ,gBAAgB;IAChB,4EpC9XuB;IoC+XvB,cAAc;IACd,UAAU;IACV,oBAAoB;IACpB,wBAA8C;IAC9C,2BAA2B;IAC3B,yBpC9TM;IoC+TN,uCAAuC;EtCgwL3C;EsCpyLA;IAsCI,UAAU;IACV,QAAQ;EtCiwLZ;EsCv6LF;IAwKI,cAAc;EtCkwLhB;EsCjwLA;;IrC7PE,qBqCgQyC;EtCkwL3C;EsCrwLA;;IrC7PE,sBqCkQyC;EtCowL3C;EsClwLA;IAjWA,OAAO;IACP,eAAe;IACf,QAAQ;IACR,WA/CiB;EtCqpMjB;EsCxwLA;IAKI,SAAS;EtCswLb;EsC3wLA;IAOM,4CpCzZqB;EFgqM3B;EsC9wLA;IASI,MAAM;EtCwwLV;EsCvwLA;;IAGI,oBA9ZiB;EtCsqMrB;EsC3wLA;;IAKI,uBAhaiB;EtC0qMrB;EsC/wLA;;IAOI,oBAA4D;EtC4wLhE;EsCnxLA;;IASI,uBAA+D;EtC8wLnE;EsC5wLA;;IAGI,cpC1auB;EFurM3B;EsChxLA;;IAKI,6BAja2C;EtCgrM/C;EsC9wLA;IAKM,yBpCtasB;EFkrM5B;AACF;;AsCzwLA;EAEI,iCAA2C;AtC2wL/C;;AuCtqMA;EAEE,erCIW;EqCHX,gBAhC0B;AvCwsM5B;;AuC3qMA;EAMI,kBrCCY;AFwqMhB;;AuC/qMA;EAQI,kBrCHY;AF8qMhB;;AuCnrMA;EAUI,iBrCNW;AFmrMf;;AuCvrMA;;EAcM,iBAAiB;EACjB,kBAAkB;EAClB,uBrCwBmB;AFspMzB;;AuC9rMA;EAkBM,uBrCsBmB;AF0pMzB;;AuC9qMA;;EAEE,mBAAmB;EACnB,aAAa;EACb,uBAAuB;EACvB,kBAAkB;AvCirMpB;;AuC/qMA;;;;EAME,cA3D6B;EA4D7B,uBAAuB;EACvB,eA5D8B;EA6D9B,mBA5DkC;EA6DlC,oBA5DmC;EA6DnC,kBAAkB;AvCgrMpB;;AuC9qMA;;;EAGE,qBrChE4B;EqCiE5B,crCrE4B;EqCsE5B,gBpCvEoB;AHwvMtB;;AuCtrMA;;;EAOI,qBrCrE0B;EqCsE1B,crCzE0B;AF8vM9B;;AuC7rMA;;;EAUI,qBrC3D8B;AFovMlC;;AuCnsMA;;;EAYI,iDrCjFyB;AF8wM7B;;AuCzsMA;;;EAcI,yBrC3E0B;EqC4E1B,qBrC5E0B;EqC6E1B,gBAAgB;EAChB,crChF0B;EqCiF1B,YAAY;AvCisMhB;;AuC/rMA;;EAEE,oBAAoB;EACpB,qBAAqB;EACrB,mBAAmB;AvCksMrB;;AuChsMA;EAEI,yBrC7E8B;EqC8E9B,qBrC9E8B;EqC+E9B,W5BnBY;AXqtMhB;;AuChsMA;EACE,crC/F4B;EqCgG5B,oBAAoB;AvCmsMtB;;AuCjsMA;EACE,eAAe;AvCosMjB;;AC/tME;EsClDF;IAiFI,eAAe;EvCqsMjB;EuC1tMF;;IAwBI,YAAY;IACZ,cAAc;EvCssMhB;EuCrsMA;IAEI,YAAY;IACZ,cAAc;EvCssMlB;AACF;;AC1uME;EsCsBF;IAiBI,YAAY;IACZ,cAAc;IACd,2BAA2B;IAC3B,QAAQ;EvCwsMV;EuCvsMA;IACE,QAAQ;EvCysMV;EuCxsMA;IACE,QAAQ;EvC0sMV;EuC9yMF;IAsGI,8BAA8B;EvC2sMhC;EuC5sMA;IAIM,QAAQ;EvC2sMd;EuC/sMA;IAMM,uBAAuB;IACvB,QAAQ;EvC4sMd;EuCntMA;IASM,QAAQ;EvC6sMd;EuCttMA;IAYM,QAAQ;EvC6sMd;EuCztMA;IAcM,QAAQ;EvC8sMd;EuC5tMA;IAgBM,yBAAyB;IACzB,QAAQ;EvC+sMd;AACF;;AwCv0MA;EACE,kBtCuCgB;EsCtChB,0FtC9B2B;EsC+B3B,etCIW;AFs0Mb;;AwC70MA;EAKI,qBtCakB;AF+zMtB;;AwCj1MA;EAYQ,uBtC3BuB;EsC4BvB,ctCzCqB;AFk3M7B;;AwCt1MA;EAeQ,0BtC9BuB;AFy2M/B;;AwC11MA;EAiBQ,YtChCuB;AF62M/B;;AwC91MA;EAYQ,yBtCxCqB;EsCyCrB,YtC5BuB;AFk3M/B;;AwCn2MA;EAeQ,4BtC3CqB;AFm4M7B;;AwCv2MA;EAiBQ,ctC7CqB;AFu4M7B;;AwC32MA;EAYQ,4BtC7BsB;EsC8BtB,yB7BqCa;AX8zMrB;;AwCh3MA;EAeQ,+BtChCsB;AFq4M9B;;AwCp3MA;EAiBQ,iBtClCsB;AFy4M9B;;AwCx3MA;EAYQ,yBtCpCsB;EsCqCtB,W7BuCQ;AXy0MhB;;AwC73MA;EAeQ,4BtCvCsB;AFy5M9B;;AwCj4MA;EAiBQ,ctCzCsB;AF65M9B;;AwCr4MA;EAYQ,yBtCtB0B;EsCuB1B,W7BuCQ;AXs1MhB;;AwC14MA;EAeQ,4BtCzB0B;AFw5MlC;;AwC94MA;EAiBQ,ctC3B0B;AF45MlC;;AwCl5MA;EAYQ,yBtCpB0B;EsCqB1B,W7BuCQ;AXm2MhB;;AwCv5MA;EAeQ,4BtCvB0B;AFm6MlC;;AwC35MA;EAiBQ,ctCzB0B;AFu6MlC;;AwC/5MA;EAYQ,yBtCrB0B;EsCsB1B,W7BuCQ;AXg3MhB;;AwCp6MA;EAeQ,4BtCxB0B;AFi7MlC;;AwCx6MA;EAiBQ,ctC1B0B;AFq7MlC;;AwC56MA;EAYQ,yBtCvB0B;EsCwB1B,W7BuCQ;AX63MhB;;AwCj7MA;EAeQ,4BtC1B0B;AFg8MlC;;AwCr7MA;EAiBQ,ctC5B0B;AFo8MlC;;AwCz7MA;EAYQ,yBtCxB0B;EsCyB1B,yB7BqCa;AX44MrB;;AwC97MA;EAeQ,4BtC3B0B;AF88MlC;;AwCl8MA;EAiBQ,ctC7B0B;AFk9MlC;;AwCt8MA;EAYQ,yBtClByB;EsCmBzB,W7BuCQ;AXu5MhB;;AwC38MA;EAeQ,4BtCrByB;AFq9MjC;;AwC/8MA;EAiBQ,ctCvByB;AFy9MjC;;AwCh8MA;;EAGI,gCtCzC2B;AF2+M/B;;AwCh8MA;EACE,yBtC5C6B;EsC6C7B,0BAA8C;EAC9C,ctCnD4B;EsCoD5B,iBAhDyB;EAiDzB,gBtCfe;EsCgBf,iBArD8B;EAsD9B,mBArDgC;AxCw/MlC;;AwCj8MA;EACE,qBAAqB;EACrB,aAAa;EACb,kBArD4B;EAsD5B,uBAAuB;AxCo8MzB;;AwCx8MA;EAMI,gCtC3D0B;EsC4D1B,mBAAmB;EACnB,cAAc;AxCs8MlB;;AwC98MA;EAWM,4BtCnEwB;EsCoExB,ctCrEwB;AF4gN9B;;AwCr8MA;EAEI,ctCxE0B;AF+gN9B;;AwCz8MA;EAIM,ctC3D4B;AFogNlC;;AwCv8MA;EACE,mBAAmB;EACnB,ctC/E4B;EsCgF5B,aAAa;EACb,2BAA2B;EAC3B,qBAAqB;AxC08MvB;;AwC/8MA;EvC6DI,oBuCtDsC;AxC48M1C;;AwCn9MA;EASI,YAAY;EACZ,cAAc;EACd,WAAW;AxC88Mf;;AwCz9MA;EAaI,eAAe;AxCg9MnB;;AwC79MA;EAeI,0BtC5E8B;EsC6E9B,ctC7F0B;AF+iN9B;;AwCl+MA;EAkBM,ctC/E4B;AFmiNlC;;AwCt+MA;EAoBI,8BtCjCc;EsCkCd,+BtClCc;AFw/MlB;;AwCp9MA;;EAEE,eAAe;AxCu9MjB;;AwCz9MA;;EAII,4BtCjG0B;AF2jN9B;;AwCx9MA;EvC9FE,qBAAqB;EACrB,euC8FgB;EvC7FhB,WuC6FqB;EvC5FrB,gBuC4FqB;EvC3FrB,kBAAkB;EAClB,mBAAmB;EACnB,UuCyFqB;EACrB,ctC1G4B;EDwI1B,oBuC7BoC;AxCi+MxC;;AwCp+MA;EAKI,kBAAkB;EAClB,oBAAoB;AxCm+MxB;;AyC7jNA;ExCkCE,iCAAiC;EwC9BjC,oBAAoB;EACpB,aAAa;EACb,evCGW;EuCFX,8BAA8B;EAC9B,gBAAgB;EAChB,gBAAgB;EAChB,mBAAmB;AzC8jNrB;;AyCxkNA;EAYI,mBAAmB;EACnB,4BvC/B0B;EuCgC1B,0BAzC4B;EA0C5B,wBAzC0B;EA0C1B,cvCrC0B;EuCsC1B,aAAa;EACb,uBAAuB;EACvB,mBAA6C;EAC7C,kBAxCyB;EAyCzB,mBAAmB;AzCgkNvB;;AyCrlNA;EAuBM,4BvC7CwB;EuC8CxB,cvC9CwB;AFgnN9B;;AyC1lNA;EA0BI,cAAc;AzCokNlB;;AyC9lNA;EA6BQ,4BvCnC0B;EuCoC1B,cvCpC0B;AFymNlC;;AyCnmNA;EAgCI,mBAAmB;EACnB,4BvCnD0B;EuCoD1B,0BA7D4B;EA8D5B,wBA7D0B;EA8D1B,aAAa;EACb,YAAY;EACZ,cAAc;EACd,2BAA2B;AzCukN/B;;AyC9mNA;EAyCM,qBAAqB;AzCykN3B;;AyClnNA;EA2CM,UAAU;EACV,uBAAuB;EACvB,oBAAoB;EACpB,qBAAqB;AzC2kN3B;;AyCznNA;EAgDM,yBAAyB;EACzB,oBAAoB;AzC6kN1B;;AyC9nNA;ExCoHI,mBwChEuC;AzC8kN3C;;AyCloNA;ExCoHI,kBwC9DuC;AzCglN3C;;AyCtoNA;EA0DM,uBAAuB;AzCglN7B;;AyC1oNA;EA6DM,yBAAyB;AzCilN/B;;AyC9oNA;EAiEM,6BAA6B;EAE3B,0BAAkE;AzCglN1E;;AyCnpNA;EAuEQ,4BvCtFsB;EuCuFtB,4BvC1FsB;AF0qN9B;;AyCxpNA;EA4EU,uBvCzFqB;EuC0FrB,qBvC/FoB;EuCgGpB,2CAA2E;AzCglNrF;;AyC9pNA;EAiFM,YAAY;EACZ,cAAc;AzCilNpB;;AyCnqNA;EAqFM,qBvCvGwB;EuCwGxB,mBA/F+B;EAgG/B,iBA/F6B;EAgG7B,gBAAgB;EAChB,kBAAkB;AzCklNxB;;AyC3qNA;EA2FQ,4BvC1GsB;EuC2GtB,qBvC/GsB;EuCgHtB,UAAU;AzColNlB;;AyCjrNA;ExCoHI,iBwCpBuE;AzCqlN3E;;AyCrrNA;EAmGU,2BvC1DE;EuC2DF,8BvC3DE;AFipNZ;;AyC1rNA;EA0GU,4BvCjEE;EuCkEF,+BvClEE;AFspNZ;;AyC/rNA;EAiHU,yBvCvHwB;EuCwHxB,qBvCxHwB;EuCyHxB,W9B7DM;E8B8DN,UAAU;AzCklNpB;;AyCtsNA;EAsHM,mBAAmB;AzColNzB;;AyC1sNA;EA2HY,mCvChFa;EuCiFb,gCvCjFa;EuCkFb,oBAAoB;AzCmlNhC;;AyChtNA;EAoIY,oCvCzFa;EuC0Fb,iCvC1Fa;EuC2Fb,qBAAqB;AzCglNjC;;AyCttNA;EA6II,kBvCnIY;AFgtNhB;;AyC1tNA;EA+II,kBvCvIY;AFstNhB;;AyC9tNA;EAiJI,iBvC1IW;AF2tNf;;A0C9vNA,eAAA;ACEA;EACE,cAAc;EACd,aAAa;EACb,YAAY;EACZ,cAAc;EACd,gBAPkB;A3CuwNpB;;A2C/vNE;EACE,UAAU;A3CkwNd;;A2CjwNE;EACE,UAAU;EACV,WAAW;A3CowNf;;A2CnwNE;EACE,UAAU;EACV,UAAU;A3CswNd;;A2CrwNE;EACE,UAAU;EACV,eAAe;A3CwwNnB;;A2CvwNE;EACE,UAAU;EACV,UAAU;A3C0wNd;;A2CzwNE;EACE,UAAU;EACV,eAAe;A3C4wNnB;;A2C3wNE;EACE,UAAU;EACV,UAAU;A3C8wNd;;A2C7wNE;EACE,UAAU;EACV,UAAU;A3CgxNd;;A2C/wNE;EACE,UAAU;EACV,UAAU;A3CkxNd;;A2CjxNE;EACE,UAAU;EACV,UAAU;A3CoxNd;;A2CnxNE;EACE,UAAU;EACV,UAAU;A3CsxNd;;A2CrxNE;E1CwGE,gB0CvGmC;A3CwxNvC;;A2CvxNE;E1CsGE,qB0CrGwC;A3C0xN5C;;A2CzxNE;E1CoGE,gB0CnGmC;A3C4xNvC;;A2C3xNE;E1CkGE,qB0CjGwC;A3C8xN5C;;A2C7xNE;E1CgGE,gB0C/FmC;A3CgyNvC;;A2C/xNE;E1C8FE,gB0C7FmC;A3CkyNvC;;A2CjyNE;E1C4FE,gB0C3FmC;A3CoyNvC;;A2CnyNE;E1C0FE,gB0CzFmC;A3CsyNvC;;A2CryNE;E1CwFE,gB0CvFmC;A3CwyNvC;;A2CtyNI;EACE,UAAU;EACV,SAA0B;A3CyyNhC;;A2CxyNI;E1CkFA,e0CjFqD;A3C2yNzD;;A2C/yNI;EACE,UAAU;EACV,eAA0B;A3CkzNhC;;A2CjzNI;E1CkFA,qB0CjFqD;A3CozNzD;;A2CxzNI;EACE,UAAU;EACV,gBAA0B;A3C2zNhC;;A2C1zNI;E1CkFA,sB0CjFqD;A3C6zNzD;;A2Cj0NI;EACE,UAAU;EACV,UAA0B;A3Co0NhC;;A2Cn0NI;E1CkFA,gB0CjFqD;A3Cs0NzD;;A2C10NI;EACE,UAAU;EACV,gBAA0B;A3C60NhC;;A2C50NI;E1CkFA,sB0CjFqD;A3C+0NzD;;A2Cn1NI;EACE,UAAU;EACV,gBAA0B;A3Cs1NhC;;A2Cr1NI;E1CkFA,sB0CjFqD;A3Cw1NzD;;A2C51NI;EACE,UAAU;EACV,UAA0B;A3C+1NhC;;A2C91NI;E1CkFA,gB0CjFqD;A3Ci2NzD;;A2Cr2NI;EACE,UAAU;EACV,gBAA0B;A3Cw2NhC;;A2Cv2NI;E1CkFA,sB0CjFqD;A3C02NzD;;A2C92NI;EACE,UAAU;EACV,gBAA0B;A3Ci3NhC;;A2Ch3NI;E1CkFA,sB0CjFqD;A3Cm3NzD;;A2Cv3NI;EACE,UAAU;EACV,UAA0B;A3C03NhC;;A2Cz3NI;E1CkFA,gB0CjFqD;A3C43NzD;;A2Ch4NI;EACE,UAAU;EACV,gBAA0B;A3Cm4NhC;;A2Cl4NI;E1CkFA,sB0CjFqD;A3Cq4NzD;;A2Cz4NI;EACE,UAAU;EACV,gBAA0B;A3C44NhC;;A2C34NI;E1CkFA,sB0CjFqD;A3C84NzD;;A2Cl5NI;EACE,UAAU;EACV,WAA0B;A3Cq5NhC;;A2Cp5NI;E1CkFA,iB0CjFqD;A3Cu5NzD;;ACr4NE;E0C/EF;IAgEM,UAAU;E3Cy5Nd;E2Cz9NF;IAkEM,UAAU;IACV,WAAW;E3C05Nf;E2C79NF;IAqEM,UAAU;IACV,UAAU;E3C25Nd;E2Cj+NF;IAwEM,UAAU;IACV,eAAe;E3C45NnB;E2Cr+NF;IA2EM,UAAU;IACV,UAAU;E3C65Nd;E2Cz+NF;IA8EM,UAAU;IACV,eAAe;E3C85NnB;E2C7+NF;IAiFM,UAAU;IACV,UAAU;E3C+5Nd;E2Cj/NF;IAoFM,UAAU;IACV,UAAU;E3Cg6Nd;E2Cr/NF;IAuFM,UAAU;IACV,UAAU;E3Ci6Nd;E2Cz/NF;IA0FM,UAAU;IACV,UAAU;E3Ck6Nd;E2C7/NF;IA6FM,UAAU;IACV,UAAU;E3Cm6Nd;E2CjgOF;I1C8II,gB0C9CqC;E3Co6NvC;E2CpgOF;I1C8II,qB0C5C0C;E3Cq6N5C;E2CvgOF;I1C8II,gB0C1CqC;E3Cs6NvC;E2C1gOF;I1C8II,qB0CxC0C;E3Cu6N5C;E2C7gOF;I1C8II,gB0CtCqC;E3Cw6NvC;E2ChhOF;I1C8II,gB0CpCqC;E3Cy6NvC;E2CnhOF;I1C8II,gB0ClCqC;E3C06NvC;E2CthOF;I1C8II,gB0ChCqC;E3C26NvC;E2CzhOF;I1C8II,gB0C9BqC;E3C46NvC;E2C5hOF;IAmHQ,UAAU;IACV,SAA0B;E3C46NhC;E2ChiOF;I1C8II,e0CxBuD;E3C66NzD;E2CniOF;IAmHQ,UAAU;IACV,eAA0B;E3Cm7NhC;E2CviOF;I1C8II,qB0CxBuD;E3Co7NzD;E2C1iOF;IAmHQ,UAAU;IACV,gBAA0B;E3C07NhC;E2C9iOF;I1C8II,sB0CxBuD;E3C27NzD;E2CjjOF;IAmHQ,UAAU;IACV,UAA0B;E3Ci8NhC;E2CrjOF;I1C8II,gB0CxBuD;E3Ck8NzD;E2CxjOF;IAmHQ,UAAU;IACV,gBAA0B;E3Cw8NhC;E2C5jOF;I1C8II,sB0CxBuD;E3Cy8NzD;E2C/jOF;IAmHQ,UAAU;IACV,gBAA0B;E3C+8NhC;E2CnkOF;I1C8II,sB0CxBuD;E3Cg9NzD;E2CtkOF;IAmHQ,UAAU;IACV,UAA0B;E3Cs9NhC;E2C1kOF;I1C8II,gB0CxBuD;E3Cu9NzD;E2C7kOF;IAmHQ,UAAU;IACV,gBAA0B;E3C69NhC;E2CjlOF;I1C8II,sB0CxBuD;E3C89NzD;E2CplOF;IAmHQ,UAAU;IACV,gBAA0B;E3Co+NhC;E2CxlOF;I1C8II,sB0CxBuD;E3Cq+NzD;E2C3lOF;IAmHQ,UAAU;IACV,UAA0B;E3C2+NhC;E2C/lOF;I1C8II,gB0CxBuD;E3C4+NzD;E2ClmOF;IAmHQ,UAAU;IACV,gBAA0B;E3Ck/NhC;E2CtmOF;I1C8II,sB0CxBuD;E3Cm/NzD;E2CzmOF;IAmHQ,UAAU;IACV,gBAA0B;E3Cy/NhC;E2C7mOF;I1C8II,sB0CxBuD;E3C0/NzD;E2ChnOF;IAmHQ,UAAU;IACV,WAA0B;E3CggOhC;E2CpnOF;I1C8II,iB0CxBuD;E3CigOzD;AACF;;ACriOE;E0CnFF;IA0HM,UAAU;E3CmgOd;E2C7nOF;IA6HM,UAAU;IACV,WAAW;E3CmgOf;E2CjoOF;IAiIM,UAAU;IACV,UAAU;E3CmgOd;E2CroOF;IAqIM,UAAU;IACV,eAAe;E3CmgOnB;E2CzoOF;IAyIM,UAAU;IACV,UAAU;E3CmgOd;E2C7oOF;IA6IM,UAAU;IACV,eAAe;E3CmgOnB;E2CjpOF;IAiJM,UAAU;IACV,UAAU;E3CmgOd;E2CrpOF;IAqJM,UAAU;IACV,UAAU;E3CmgOd;E2CzpOF;IAyJM,UAAU;IACV,UAAU;E3CmgOd;E2C7pOF;IA6JM,UAAU;IACV,UAAU;E3CmgOd;E2CjqOF;IAiKM,UAAU;IACV,UAAU;E3CmgOd;E2CrqOF;I1C8II,gB0CuBqC;E3CmgOvC;E2CxqOF;I1C8II,qB0C0B0C;E3CmgO5C;E2C3qOF;I1C8II,gB0C6BqC;E3CmgOvC;E2C9qOF;I1C8II,qB0CgC0C;E3CmgO5C;E2CjrOF;I1C8II,gB0CmCqC;E3CmgOvC;E2CprOF;I1C8II,gB0CsCqC;E3CmgOvC;E2CvrOF;I1C8II,gB0CyCqC;E3CmgOvC;E2C1rOF;I1C8II,gB0C4CqC;E3CmgOvC;E2C7rOF;I1C8II,gB0C+CqC;E3CmgOvC;E2ChsOF;IAiMQ,UAAU;IACV,SAA0B;E3CkgOhC;E2CpsOF;I1C8II,e0CuDuD;E3CkgOzD;E2CvsOF;IAiMQ,UAAU;IACV,eAA0B;E3CygOhC;E2C3sOF;I1C8II,qB0CuDuD;E3CygOzD;E2C9sOF;IAiMQ,UAAU;IACV,gBAA0B;E3CghOhC;E2CltOF;I1C8II,sB0CuDuD;E3CghOzD;E2CrtOF;IAiMQ,UAAU;IACV,UAA0B;E3CuhOhC;E2CztOF;I1C8II,gB0CuDuD;E3CuhOzD;E2C5tOF;IAiMQ,UAAU;IACV,gBAA0B;E3C8hOhC;E2ChuOF;I1C8II,sB0CuDuD;E3C8hOzD;E2CnuOF;IAiMQ,UAAU;IACV,gBAA0B;E3CqiOhC;E2CvuOF;I1C8II,sB0CuDuD;E3CqiOzD;E2C1uOF;IAiMQ,UAAU;IACV,UAA0B;E3C4iOhC;E2C9uOF;I1C8II,gB0CuDuD;E3C4iOzD;E2CjvOF;IAiMQ,UAAU;IACV,gBAA0B;E3CmjOhC;E2CrvOF;I1C8II,sB0CuDuD;E3CmjOzD;E2CxvOF;IAiMQ,UAAU;IACV,gBAA0B;E3C0jOhC;E2C5vOF;I1C8II,sB0CuDuD;E3C0jOzD;E2C/vOF;IAiMQ,UAAU;IACV,UAA0B;E3CikOhC;E2CnwOF;I1C8II,gB0CuDuD;E3CikOzD;E2CtwOF;IAiMQ,UAAU;IACV,gBAA0B;E3CwkOhC;E2C1wOF;I1C8II,sB0CuDuD;E3CwkOzD;E2C7wOF;IAiMQ,UAAU;IACV,gBAA0B;E3C+kOhC;E2CjxOF;I1C8II,sB0CuDuD;E3C+kOzD;E2CpxOF;IAiMQ,UAAU;IACV,WAA0B;E3CslOhC;E2CxxOF;I1C8II,iB0CuDuD;E3CslOzD;AACF;;ACjsOE;E0C3FF;IAwMM,UAAU;E3CylOd;E2CjyOF;IA0MM,UAAU;IACV,WAAW;E3C0lOf;E2CryOF;IA6MM,UAAU;IACV,UAAU;E3C2lOd;E2CzyOF;IAgNM,UAAU;IACV,eAAe;E3C4lOnB;E2C7yOF;IAmNM,UAAU;IACV,UAAU;E3C6lOd;E2CjzOF;IAsNM,UAAU;IACV,eAAe;E3C8lOnB;E2CrzOF;IAyNM,UAAU;IACV,UAAU;E3C+lOd;E2CzzOF;IA4NM,UAAU;IACV,UAAU;E3CgmOd;E2C7zOF;IA+NM,UAAU;IACV,UAAU;E3CimOd;E2Cj0OF;IAkOM,UAAU;IACV,UAAU;E3CkmOd;E2Cr0OF;IAqOM,UAAU;IACV,UAAU;E3CmmOd;E2Cz0OF;I1C8II,gB0C0FqC;E3ComOvC;E2C50OF;I1C8II,qB0C4F0C;E3CqmO5C;E2C/0OF;I1C8II,gB0C8FqC;E3CsmOvC;E2Cl1OF;I1C8II,qB0CgG0C;E3CumO5C;E2Cr1OF;I1C8II,gB0CkGqC;E3CwmOvC;E2Cx1OF;I1C8II,gB0CoGqC;E3CymOvC;E2C31OF;I1C8II,gB0CsGqC;E3C0mOvC;E2C91OF;I1C8II,gB0CwGqC;E3C2mOvC;E2Cj2OF;I1C8II,gB0C0GqC;E3C4mOvC;E2Cp2OF;IA2PQ,UAAU;IACV,SAA0B;E3C4mOhC;E2Cx2OF;I1C8II,e0CgHuD;E3C6mOzD;E2C32OF;IA2PQ,UAAU;IACV,eAA0B;E3CmnOhC;E2C/2OF;I1C8II,qB0CgHuD;E3ConOzD;E2Cl3OF;IA2PQ,UAAU;IACV,gBAA0B;E3C0nOhC;E2Ct3OF;I1C8II,sB0CgHuD;E3C2nOzD;E2Cz3OF;IA2PQ,UAAU;IACV,UAA0B;E3CioOhC;E2C73OF;I1C8II,gB0CgHuD;E3CkoOzD;E2Ch4OF;IA2PQ,UAAU;IACV,gBAA0B;E3CwoOhC;E2Cp4OF;I1C8II,sB0CgHuD;E3CyoOzD;E2Cv4OF;IA2PQ,UAAU;IACV,gBAA0B;E3C+oOhC;E2C34OF;I1C8II,sB0CgHuD;E3CgpOzD;E2C94OF;IA2PQ,UAAU;IACV,UAA0B;E3CspOhC;E2Cl5OF;I1C8II,gB0CgHuD;E3CupOzD;E2Cr5OF;IA2PQ,UAAU;IACV,gBAA0B;E3C6pOhC;E2Cz5OF;I1C8II,sB0CgHuD;E3C8pOzD;E2C55OF;IA2PQ,UAAU;IACV,gBAA0B;E3CoqOhC;E2Ch6OF;I1C8II,sB0CgHuD;E3CqqOzD;E2Cn6OF;IA2PQ,UAAU;IACV,UAA0B;E3C2qOhC;E2Cv6OF;I1C8II,gB0CgHuD;E3C4qOzD;E2C16OF;IA2PQ,UAAU;IACV,gBAA0B;E3CkrOhC;E2C96OF;I1C8II,sB0CgHuD;E3CmrOzD;E2Cj7OF;IA2PQ,UAAU;IACV,gBAA0B;E3CyrOhC;E2Cr7OF;I1C8II,sB0CgHuD;E3C0rOzD;E2Cx7OF;IA2PQ,UAAU;IACV,WAA0B;E3CgsOhC;E2C57OF;I1C8II,iB0CgHuD;E3CisOzD;AACF;;ACj2OE;E0C/FF;IAiQM,UAAU;E3CosOd;E2Cr8OF;IAmQM,UAAU;IACV,WAAW;E3CqsOf;E2Cz8OF;IAsQM,UAAU;IACV,UAAU;E3CssOd;E2C78OF;IAyQM,UAAU;IACV,eAAe;E3CusOnB;E2Cj9OF;IA4QM,UAAU;IACV,UAAU;E3CwsOd;E2Cr9OF;IA+QM,UAAU;IACV,eAAe;E3CysOnB;E2Cz9OF;IAkRM,UAAU;IACV,UAAU;E3C0sOd;E2C79OF;IAqRM,UAAU;IACV,UAAU;E3C2sOd;E2Cj+OF;IAwRM,UAAU;IACV,UAAU;E3C4sOd;E2Cr+OF;IA2RM,UAAU;IACV,UAAU;E3C6sOd;E2Cz+OF;IA8RM,UAAU;IACV,UAAU;E3C8sOd;E2C7+OF;I1C8II,gB0CmJqC;E3C+sOvC;E2Ch/OF;I1C8II,qB0CqJ0C;E3CgtO5C;E2Cn/OF;I1C8II,gB0CuJqC;E3CitOvC;E2Ct/OF;I1C8II,qB0CyJ0C;E3CktO5C;E2Cz/OF;I1C8II,gB0C2JqC;E3CmtOvC;E2C5/OF;I1C8II,gB0C6JqC;E3CotOvC;E2C//OF;I1C8II,gB0C+JqC;E3CqtOvC;E2ClgPF;I1C8II,gB0CiKqC;E3CstOvC;E2CrgPF;I1C8II,gB0CmKqC;E3CutOvC;E2CxgPF;IAoTQ,UAAU;IACV,SAA0B;E3CutOhC;E2C5gPF;I1C8II,e0CyKuD;E3CwtOzD;E2C/gPF;IAoTQ,UAAU;IACV,eAA0B;E3C8tOhC;E2CnhPF;I1C8II,qB0CyKuD;E3C+tOzD;E2CthPF;IAoTQ,UAAU;IACV,gBAA0B;E3CquOhC;E2C1hPF;I1C8II,sB0CyKuD;E3CsuOzD;E2C7hPF;IAoTQ,UAAU;IACV,UAA0B;E3C4uOhC;E2CjiPF;I1C8II,gB0CyKuD;E3C6uOzD;E2CpiPF;IAoTQ,UAAU;IACV,gBAA0B;E3CmvOhC;E2CxiPF;I1C8II,sB0CyKuD;E3CovOzD;E2C3iPF;IAoTQ,UAAU;IACV,gBAA0B;E3C0vOhC;E2C/iPF;I1C8II,sB0CyKuD;E3C2vOzD;E2CljPF;IAoTQ,UAAU;IACV,UAA0B;E3CiwOhC;E2CtjPF;I1C8II,gB0CyKuD;E3CkwOzD;E2CzjPF;IAoTQ,UAAU;IACV,gBAA0B;E3CwwOhC;E2C7jPF;I1C8II,sB0CyKuD;E3CywOzD;E2ChkPF;IAoTQ,UAAU;IACV,gBAA0B;E3C+wOhC;E2CpkPF;I1C8II,sB0CyKuD;E3CgxOzD;E2CvkPF;IAoTQ,UAAU;IACV,UAA0B;E3CsxOhC;E2C3kPF;I1C8II,gB0CyKuD;E3CuxOzD;E2C9kPF;IAoTQ,UAAU;IACV,gBAA0B;E3C6xOhC;E2CllPF;I1C8II,sB0CyKuD;E3C8xOzD;E2CrlPF;IAoTQ,UAAU;IACV,gBAA0B;E3CoyOhC;E2CzlPF;I1C8II,sB0CyKuD;E3CqyOzD;E2C5lPF;IAoTQ,UAAU;IACV,WAA0B;E3C2yOhC;E2ChmPF;I1C8II,iB0CyKuD;E3C4yOzD;AACF;;ACt/OI;E0C9GJ;IA0TM,UAAU;E3C+yOd;E2CzmPF;IA4TM,UAAU;IACV,WAAW;E3CgzOf;E2C7mPF;IA+TM,UAAU;IACV,UAAU;E3CizOd;E2CjnPF;IAkUM,UAAU;IACV,eAAe;E3CkzOnB;E2CrnPF;IAqUM,UAAU;IACV,UAAU;E3CmzOd;E2CznPF;IAwUM,UAAU;IACV,eAAe;E3CozOnB;E2C7nPF;IA2UM,UAAU;IACV,UAAU;E3CqzOd;E2CjoPF;IA8UM,UAAU;IACV,UAAU;E3CszOd;E2CroPF;IAiVM,UAAU;IACV,UAAU;E3CuzOd;E2CzoPF;IAoVM,UAAU;IACV,UAAU;E3CwzOd;E2C7oPF;IAuVM,UAAU;IACV,UAAU;E3CyzOd;E2CjpPF;I1C8II,gB0C4MqC;E3C0zOvC;E2CppPF;I1C8II,qB0C8M0C;E3C2zO5C;E2CvpPF;I1C8II,gB0CgNqC;E3C4zOvC;E2C1pPF;I1C8II,qB0CkN0C;E3C6zO5C;E2C7pPF;I1C8II,gB0CoNqC;E3C8zOvC;E2ChqPF;I1C8II,gB0CsNqC;E3C+zOvC;E2CnqPF;I1C8II,gB0CwNqC;E3Cg0OvC;E2CtqPF;I1C8II,gB0C0NqC;E3Ci0OvC;E2CzqPF;I1C8II,gB0C4NqC;E3Ck0OvC;E2C5qPF;IA6WQ,UAAU;IACV,SAA0B;E3Ck0OhC;E2ChrPF;I1C8II,e0CkOuD;E3Cm0OzD;E2CnrPF;IA6WQ,UAAU;IACV,eAA0B;E3Cy0OhC;E2CvrPF;I1C8II,qB0CkOuD;E3C00OzD;E2C1rPF;IA6WQ,UAAU;IACV,gBAA0B;E3Cg1OhC;E2C9rPF;I1C8II,sB0CkOuD;E3Ci1OzD;E2CjsPF;IA6WQ,UAAU;IACV,UAA0B;E3Cu1OhC;E2CrsPF;I1C8II,gB0CkOuD;E3Cw1OzD;E2CxsPF;IA6WQ,UAAU;IACV,gBAA0B;E3C81OhC;E2C5sPF;I1C8II,sB0CkOuD;E3C+1OzD;E2C/sPF;IA6WQ,UAAU;IACV,gBAA0B;E3Cq2OhC;E2CntPF;I1C8II,sB0CkOuD;E3Cs2OzD;E2CttPF;IA6WQ,UAAU;IACV,UAA0B;E3C42OhC;E2C1tPF;I1C8II,gB0CkOuD;E3C62OzD;E2C7tPF;IA6WQ,UAAU;IACV,gBAA0B;E3Cm3OhC;E2CjuPF;I1C8II,sB0CkOuD;E3Co3OzD;E2CpuPF;IA6WQ,UAAU;IACV,gBAA0B;E3C03OhC;E2CxuPF;I1C8II,sB0CkOuD;E3C23OzD;E2C3uPF;IA6WQ,UAAU;IACV,UAA0B;E3Ci4OhC;E2C/uPF;I1C8II,gB0CkOuD;E3Ck4OzD;E2ClvPF;IA6WQ,UAAU;IACV,gBAA0B;E3Cw4OhC;E2CtvPF;I1C8II,sB0CkOuD;E3Cy4OzD;E2CzvPF;IA6WQ,UAAU;IACV,gBAA0B;E3C+4OhC;E2C7vPF;I1C8II,sB0CkOuD;E3Cg5OzD;E2ChwPF;IA6WQ,UAAU;IACV,WAA0B;E3Cs5OhC;E2CpwPF;I1C8II,iB0CkOuD;E3Cu5OzD;AACF;;AC3oPI;E0C7HJ;IAmXM,UAAU;E3C05Od;E2C7wPF;IAqXM,UAAU;IACV,WAAW;E3C25Of;E2CjxPF;IAwXM,UAAU;IACV,UAAU;E3C45Od;E2CrxPF;IA2XM,UAAU;IACV,eAAe;E3C65OnB;E2CzxPF;IA8XM,UAAU;IACV,UAAU;E3C85Od;E2C7xPF;IAiYM,UAAU;IACV,eAAe;E3C+5OnB;E2CjyPF;IAoYM,UAAU;IACV,UAAU;E3Cg6Od;E2CryPF;IAuYM,UAAU;IACV,UAAU;E3Ci6Od;E2CzyPF;IA0YM,UAAU;IACV,UAAU;E3Ck6Od;E2C7yPF;IA6YM,UAAU;IACV,UAAU;E3Cm6Od;E2CjzPF;IAgZM,UAAU;IACV,UAAU;E3Co6Od;E2CrzPF;I1C8II,gB0CqQqC;E3Cq6OvC;E2CxzPF;I1C8II,qB0CuQ0C;E3Cs6O5C;E2C3zPF;I1C8II,gB0CyQqC;E3Cu6OvC;E2C9zPF;I1C8II,qB0C2Q0C;E3Cw6O5C;E2Cj0PF;I1C8II,gB0C6QqC;E3Cy6OvC;E2Cp0PF;I1C8II,gB0C+QqC;E3C06OvC;E2Cv0PF;I1C8II,gB0CiRqC;E3C26OvC;E2C10PF;I1C8II,gB0CmRqC;E3C46OvC;E2C70PF;I1C8II,gB0CqRqC;E3C66OvC;E2Ch1PF;IAsaQ,UAAU;IACV,SAA0B;E3C66OhC;E2Cp1PF;I1C8II,e0C2RuD;E3C86OzD;E2Cv1PF;IAsaQ,UAAU;IACV,eAA0B;E3Co7OhC;E2C31PF;I1C8II,qB0C2RuD;E3Cq7OzD;E2C91PF;IAsaQ,UAAU;IACV,gBAA0B;E3C27OhC;E2Cl2PF;I1C8II,sB0C2RuD;E3C47OzD;E2Cr2PF;IAsaQ,UAAU;IACV,UAA0B;E3Ck8OhC;E2Cz2PF;I1C8II,gB0C2RuD;E3Cm8OzD;E2C52PF;IAsaQ,UAAU;IACV,gBAA0B;E3Cy8OhC;E2Ch3PF;I1C8II,sB0C2RuD;E3C08OzD;E2Cn3PF;IAsaQ,UAAU;IACV,gBAA0B;E3Cg9OhC;E2Cv3PF;I1C8II,sB0C2RuD;E3Ci9OzD;E2C13PF;IAsaQ,UAAU;IACV,UAA0B;E3Cu9OhC;E2C93PF;I1C8II,gB0C2RuD;E3Cw9OzD;E2Cj4PF;IAsaQ,UAAU;IACV,gBAA0B;E3C89OhC;E2Cr4PF;I1C8II,sB0C2RuD;E3C+9OzD;E2Cx4PF;IAsaQ,UAAU;IACV,gBAA0B;E3Cq+OhC;E2C54PF;I1C8II,sB0C2RuD;E3Cs+OzD;E2C/4PF;IAsaQ,UAAU;IACV,UAA0B;E3C4+OhC;E2Cn5PF;I1C8II,gB0C2RuD;E3C6+OzD;E2Ct5PF;IAsaQ,UAAU;IACV,gBAA0B;E3Cm/OhC;E2C15PF;I1C8II,sB0C2RuD;E3Co/OzD;E2C75PF;IAsaQ,UAAU;IACV,gBAA0B;E3C0/OhC;E2Cj6PF;I1C8II,sB0C2RuD;E3C2/OzD;E2Cp6PF;IAsaQ,UAAU;IACV,WAA0B;E3CigPhC;E2Cx6PF;I1C8II,iB0C2RuD;E3CkgPzD;AACF;;A2CjgPA;E1C7RI,qB0ChJgB;E1CgJhB,sB0ChJgB;EAgblB,oBAhbkB;A3Co7PpB;;A2CvgPA;EAKI,uBAlbgB;A3Cw7PpB;;A2C3gPA;EAOI,qCAA4C;A3CwgPhD;;A2C/gPA;EAUI,uBAAuB;A3CygP3B;;A2CnhPA;E1C7RI,c0CySiC;E1CzSjC,e0C0SiC;EACjC,aAAa;A3C2gPjB;;A2CzhPA;EAgBM,SAAS;EACT,qBAAqB;A3C6gP3B;;A2C9hPA;EAmBM,qBAAqB;A3C+gP3B;;A2CliPA;EAqBM,gBAAgB;A3CihPtB;;A2CtiPA;EAuBI,aAAa;A3CmhPjB;;A2C1iPA;EAyBI,eAAe;A3CqhPnB;;A2C9iPA;EA2BI,mBAAmB;A3CuhPvB;;AC14PE;E0CwVF;IA+BM,aAAa;E3CwhPjB;AACF;;ACp4PE;E0C4UF;IAmCM,aAAa;E3C0hPjB;AACF;;A2CxhPE;EACE,oBAAY;E1CpUZ,wC0CqU2D;E1CrU3D,yC0CsU2D;A3C2hP/D;;A2C9hPE;EAKI,8BAA8B;EAC9B,+BAA+B;A3C6hPrC;;A2CniPE;EASM,iBAAY;A3C8hPpB;;ACz6PE;E0CkYA;IAYQ,iBAAY;E3CgiPpB;AACF;;AC36PE;E0C8XA;IAeQ,iBAAY;E3CmiPpB;AACF;;AC76PE;E0C0XA;IAkBQ,iBAAY;E3CsiPpB;AACF;;AC/6PE;E0CsXA;IAqBQ,iBAAY;E3CyiPpB;AACF;;ACj7PE;E0CkXA;IAwBQ,iBAAY;E3C4iPpB;AACF;;ACl7PI;E0C6WF;IA2BQ,iBAAY;E3C+iPpB;AACF;;AC96PI;E0CmWF;IA8BQ,iBAAY;E3CkjPpB;AACF;;AC/6PI;E0C8VF;IAiCQ,iBAAY;E3CqjPpB;AACF;;AC36PI;E0CoVF;IAoCQ,iBAAY;E3CwjPpB;AACF;;A2C7lPE;EASM,oBAAY;A3CwlPpB;;ACn+PE;E0CkYA;IAYQ,oBAAY;E3C0lPpB;AACF;;ACr+PE;E0C8XA;IAeQ,oBAAY;E3C6lPpB;AACF;;ACv+PE;E0C0XA;IAkBQ,oBAAY;E3CgmPpB;AACF;;ACz+PE;E0CsXA;IAqBQ,oBAAY;E3CmmPpB;AACF;;AC3+PE;E0CkXA;IAwBQ,oBAAY;E3CsmPpB;AACF;;AC5+PI;E0C6WF;IA2BQ,oBAAY;E3CymPpB;AACF;;ACx+PI;E0CmWF;IA8BQ,oBAAY;E3C4mPpB;AACF;;ACz+PI;E0C8VF;IAiCQ,oBAAY;E3C+mPpB;AACF;;ACr+PI;E0CoVF;IAoCQ,oBAAY;E3CknPpB;AACF;;A2CvpPE;EASM,mBAAY;A3CkpPpB;;AC7hQE;E0CkYA;IAYQ,mBAAY;E3CopPpB;AACF;;AC/hQE;E0C8XA;IAeQ,mBAAY;E3CupPpB;AACF;;ACjiQE;E0C0XA;IAkBQ,mBAAY;E3C0pPpB;AACF;;ACniQE;E0CsXA;IAqBQ,mBAAY;E3C6pPpB;AACF;;ACriQE;E0CkXA;IAwBQ,mBAAY;E3CgqPpB;AACF;;ACtiQI;E0C6WF;IA2BQ,mBAAY;E3CmqPpB;AACF;;ACliQI;E0CmWF;IA8BQ,mBAAY;E3CsqPpB;AACF;;ACniQI;E0C8VF;IAiCQ,mBAAY;E3CyqPpB;AACF;;AC/hQI;E0CoVF;IAoCQ,mBAAY;E3C4qPpB;AACF;;A2CjtPE;EASM,oBAAY;A3C4sPpB;;ACvlQE;E0CkYA;IAYQ,oBAAY;E3C8sPpB;AACF;;ACzlQE;E0C8XA;IAeQ,oBAAY;E3CitPpB;AACF;;AC3lQE;E0C0XA;IAkBQ,oBAAY;E3CotPpB;AACF;;AC7lQE;E0CsXA;IAqBQ,oBAAY;E3CutPpB;AACF;;AC/lQE;E0CkXA;IAwBQ,oBAAY;E3C0tPpB;AACF;;AChmQI;E0C6WF;IA2BQ,oBAAY;E3C6tPpB;AACF;;AC5lQI;E0CmWF;IA8BQ,oBAAY;E3CguPpB;AACF;;AC7lQI;E0C8VF;IAiCQ,oBAAY;E3CmuPpB;AACF;;ACzlQI;E0CoVF;IAoCQ,oBAAY;E3CsuPpB;AACF;;A2C3wPE;EASM,iBAAY;A3CswPpB;;ACjpQE;E0CkYA;IAYQ,iBAAY;E3CwwPpB;AACF;;ACnpQE;E0C8XA;IAeQ,iBAAY;E3C2wPpB;AACF;;ACrpQE;E0C0XA;IAkBQ,iBAAY;E3C8wPpB;AACF;;ACvpQE;E0CsXA;IAqBQ,iBAAY;E3CixPpB;AACF;;ACzpQE;E0CkXA;IAwBQ,iBAAY;E3CoxPpB;AACF;;AC1pQI;E0C6WF;IA2BQ,iBAAY;E3CuxPpB;AACF;;ACtpQI;E0CmWF;IA8BQ,iBAAY;E3C0xPpB;AACF;;ACvpQI;E0C8VF;IAiCQ,iBAAY;E3C6xPpB;AACF;;ACnpQI;E0CoVF;IAoCQ,iBAAY;E3CgyPpB;AACF;;A2Cr0PE;EASM,oBAAY;A3Cg0PpB;;AC3sQE;E0CkYA;IAYQ,oBAAY;E3Ck0PpB;AACF;;AC7sQE;E0C8XA;IAeQ,oBAAY;E3Cq0PpB;AACF;;AC/sQE;E0C0XA;IAkBQ,oBAAY;E3Cw0PpB;AACF;;ACjtQE;E0CsXA;IAqBQ,oBAAY;E3C20PpB;AACF;;ACntQE;E0CkXA;IAwBQ,oBAAY;E3C80PpB;AACF;;ACptQI;E0C6WF;IA2BQ,oBAAY;E3Ci1PpB;AACF;;AChtQI;E0CmWF;IA8BQ,oBAAY;E3Co1PpB;AACF;;ACjtQI;E0C8VF;IAiCQ,oBAAY;E3Cu1PpB;AACF;;AC7sQI;E0CoVF;IAoCQ,oBAAY;E3C01PpB;AACF;;A2C/3PE;EASM,mBAAY;A3C03PpB;;ACrwQE;E0CkYA;IAYQ,mBAAY;E3C43PpB;AACF;;ACvwQE;E0C8XA;IAeQ,mBAAY;E3C+3PpB;AACF;;ACzwQE;E0C0XA;IAkBQ,mBAAY;E3Ck4PpB;AACF;;AC3wQE;E0CsXA;IAqBQ,mBAAY;E3Cq4PpB;AACF;;AC7wQE;E0CkXA;IAwBQ,mBAAY;E3Cw4PpB;AACF;;AC9wQI;E0C6WF;IA2BQ,mBAAY;E3C24PpB;AACF;;AC1wQI;E0CmWF;IA8BQ,mBAAY;E3C84PpB;AACF;;AC3wQI;E0C8VF;IAiCQ,mBAAY;E3Ci5PpB;AACF;;ACvwQI;E0CoVF;IAoCQ,mBAAY;E3Co5PpB;AACF;;A2Cz7PE;EASM,oBAAY;A3Co7PpB;;AC/zQE;E0CkYA;IAYQ,oBAAY;E3Cs7PpB;AACF;;ACj0QE;E0C8XA;IAeQ,oBAAY;E3Cy7PpB;AACF;;ACn0QE;E0C0XA;IAkBQ,oBAAY;E3C47PpB;AACF;;ACr0QE;E0CsXA;IAqBQ,oBAAY;E3C+7PpB;AACF;;ACv0QE;E0CkXA;IAwBQ,oBAAY;E3Ck8PpB;AACF;;ACx0QI;E0C6WF;IA2BQ,oBAAY;E3Cq8PpB;AACF;;ACp0QI;E0CmWF;IA8BQ,oBAAY;E3Cw8PpB;AACF;;ACr0QI;E0C8VF;IAiCQ,oBAAY;E3C28PpB;AACF;;ACj0QI;E0CoVF;IAoCQ,oBAAY;E3C88PpB;AACF;;A2Cn/PE;EASM,iBAAY;A3C8+PpB;;ACz3QE;E0CkYA;IAYQ,iBAAY;E3Cg/PpB;AACF;;AC33QE;E0C8XA;IAeQ,iBAAY;E3Cm/PpB;AACF;;AC73QE;E0C0XA;IAkBQ,iBAAY;E3Cs/PpB;AACF;;AC/3QE;E0CsXA;IAqBQ,iBAAY;E3Cy/PpB;AACF;;ACj4QE;E0CkXA;IAwBQ,iBAAY;E3C4/PpB;AACF;;ACl4QI;E0C6WF;IA2BQ,iBAAY;E3C+/PpB;AACF;;AC93QI;E0CmWF;IA8BQ,iBAAY;E3CkgQpB;AACF;;AC/3QI;E0C8VF;IAiCQ,iBAAY;E3CqgQpB;AACF;;AC33QI;E0CoVF;IAoCQ,iBAAY;E3CwgQpB;AACF;;A4C9/QA;EACE,oBAAoB;EACpB,cAAc;EACd,aAAa;EACb,YAAY;EACZ,cAAc;EACd,+BAAuB;EAAvB,4BAAuB;EAAvB,uBAAuB;A5CigRzB;;A4CvgRA;EASI,qBAA+B;EAC/B,sBAAgC;EAChC,oBAA8B;A5CkgRlC;;A4C7gRA;EAaM,uBAAiC;A5CogRvC;;A4CjhRA;EAeM,sBAjBgB;A5CuhRtB;;A4CrhRA;EAiBI,oBAAoB;A5CwgRxB;;A4CzhRA;EAmBI,gBArBkB;A5C+hRtB;;A4C7hRA;EAqBI,sBAAsB;A5C4gR1B;;A4CjiRA;EAuBM,gCAAgC;A5C8gRtC;;ACl9QE;E2CnFF;IA2BM,aAAa;E5C+gRjB;E4C1iRF;IA8BQ,UAAU;IACV,eAAuB;E5C+gR7B;E4C9iRF;IA8BQ,UAAU;IACV,gBAAuB;E5CmhR7B;E4CljRF;IA8BQ,UAAU;IACV,UAAuB;E5CuhR7B;E4CtjRF;IA8BQ,UAAU;IACV,gBAAuB;E5C2hR7B;E4C1jRF;IA8BQ,UAAU;IACV,gBAAuB;E5C+hR7B;E4C9jRF;IA8BQ,UAAU;IACV,UAAuB;E5CmiR7B;E4ClkRF;IA8BQ,UAAU;IACV,gBAAuB;E5CuiR7B;E4CtkRF;IA8BQ,UAAU;IACV,gBAAuB;E5C2iR7B;E4C1kRF;IA8BQ,UAAU;IACV,UAAuB;E5C+iR7B;E4C9kRF;IA8BQ,UAAU;IACV,gBAAuB;E5CmjR7B;E4CllRF;IA8BQ,UAAU;IACV,gBAAuB;E5CujR7B;E4CtlRF;IA8BQ,UAAU;IACV,WAAuB;E5C2jR7B;AACF;;A6C7lRA,kBAAA;ACEE;EACE,uBAAwB;A9C+lR5B;;A8C9lRE;EAGI,yBAA0C;A9C+lRhD;;A8C9lRE;EACE,kCAAmC;A9CimRvC;;A8CxmRE;EACE,yBAAwB;A9C2mR5B;;A8C1mRE;EAGI,uBAA0C;A9C2mRhD;;A8C1mRE;EACE,oCAAmC;A9C6mRvC;;A8CpnRE;EACE,4BAAwB;A9CunR5B;;A8CtnRE;EAGI,yBAA0C;A9CunRhD;;A8CtnRE;EACE,uCAAmC;A9CynRvC;;A8ChoRE;EACE,yBAAwB;A9CmoR5B;;A8CloRE;EAGI,yBAA0C;A9CmoRhD;;A8CloRE;EACE,oCAAmC;A9CqoRvC;;A8C5oRE;EACE,yBAAwB;A9C+oR5B;;A8C9oRE;EAGI,yBAA0C;A9C+oRhD;;A8C9oRE;EACE,oCAAmC;A9CipRvC;;A8C5oRI;EACE,yBAA8B;A9C+oRpC;;A8C9oRI;EAGI,yBAAgD;A9C+oRxD;;A8C9oRI;EACE,oCAAyC;A9CipR/C;;A8C/oRI;EACE,yBAA6B;A9CkpRnC;;A8CjpRI;EAGI,yBAAgD;A9CkpRxD;;A8CjpRI;EACE,oCAAwC;A9CopR9C;;A8ChrRE;EACE,yBAAwB;A9CmrR5B;;A8ClrRE;EAGI,yBAA0C;A9CmrRhD;;A8ClrRE;EACE,oCAAmC;A9CqrRvC;;A8ChrRI;EACE,yBAA8B;A9CmrRpC;;A8ClrRI;EAGI,yBAAgD;A9CmrRxD;;A8ClrRI;EACE,oCAAyC;A9CqrR/C;;A8CnrRI;EACE,yBAA6B;A9CsrRnC;;A8CrrRI;EAGI,yBAAgD;A9CsrRxD;;A8CrrRI;EACE,oCAAwC;A9CwrR9C;;A8CptRE;EACE,yBAAwB;A9CutR5B;;A8CttRE;EAGI,yBAA0C;A9CutRhD;;A8CttRE;EACE,oCAAmC;A9CytRvC;;A8CptRI;EACE,yBAA8B;A9CutRpC;;A8CttRI;EAGI,yBAAgD;A9CutRxD;;A8CttRI;EACE,oCAAyC;A9CytR/C;;A8CvtRI;EACE,yBAA6B;A9C0tRnC;;A8CztRI;EAGI,yBAAgD;A9C0tRxD;;A8CztRI;EACE,oCAAwC;A9C4tR9C;;A8CxvRE;EACE,yBAAwB;A9C2vR5B;;A8C1vRE;EAGI,yBAA0C;A9C2vRhD;;A8C1vRE;EACE,oCAAmC;A9C6vRvC;;A8CxvRI;EACE,yBAA8B;A9C2vRpC;;A8C1vRI;EAGI,yBAAgD;A9C2vRxD;;A8C1vRI;EACE,oCAAyC;A9C6vR/C;;A8C3vRI;EACE,yBAA6B;A9C8vRnC;;A8C7vRI;EAGI,yBAAgD;A9C8vRxD;;A8C7vRI;EACE,oCAAwC;A9CgwR9C;;A8C5xRE;EACE,yBAAwB;A9C+xR5B;;A8C9xRE;EAGI,yBAA0C;A9C+xRhD;;A8C9xRE;EACE,oCAAmC;A9CiyRvC;;A8C5xRI;EACE,yBAA8B;A9C+xRpC;;A8C9xRI;EAGI,yBAAgD;A9C+xRxD;;A8C9xRI;EACE,oCAAyC;A9CiyR/C;;A8C/xRI;EACE,yBAA6B;A9CkyRnC;;A8CjyRI;EAGI,yBAAgD;A9CkyRxD;;A8CjyRI;EACE,oCAAwC;A9CoyR9C;;A8Ch0RE;EACE,yBAAwB;A9Cm0R5B;;A8Cl0RE;EAGI,yBAA0C;A9Cm0RhD;;A8Cl0RE;EACE,oCAAmC;A9Cq0RvC;;A8Ch0RI;EACE,yBAA8B;A9Cm0RpC;;A8Cl0RI;EAGI,yBAAgD;A9Cm0RxD;;A8Cl0RI;EACE,oCAAyC;A9Cq0R/C;;A8Cn0RI;EACE,yBAA6B;A9Cs0RnC;;A8Cr0RI;EAGI,yBAAgD;A9Cs0RxD;;A8Cr0RI;EACE,oCAAwC;A9Cw0R9C;;A8Cr0RE;EACE,yBAAwB;A9Cw0R5B;;A8Cv0RE;EACE,oCAAmC;A9C00RvC;;A8C70RE;EACE,yBAAwB;A9Cg1R5B;;A8C/0RE;EACE,oCAAmC;A9Ck1RvC;;A8Cr1RE;EACE,yBAAwB;A9Cw1R5B;;A8Cv1RE;EACE,oCAAmC;A9C01RvC;;A8C71RE;EACE,yBAAwB;A9Cg2R5B;;A8C/1RE;EACE,oCAAmC;A9Ck2RvC;;A8Cr2RE;EACE,yBAAwB;A9Cw2R5B;;A8Cv2RE;EACE,oCAAmC;A9C02RvC;;A8C72RE;EACE,yBAAwB;A9Cg3R5B;;A8C/2RE;EACE,oCAAmC;A9Ck3RvC;;A8Cr3RE;EACE,yBAAwB;A9Cw3R5B;;A8Cv3RE;EACE,oCAAmC;A9C03RvC;;A8C73RE;EACE,4BAAwB;A9Cg4R5B;;A8C/3RE;EACE,uCAAmC;A9Ck4RvC;;A8Cr4RE;EACE,yBAAwB;A9Cw4R5B;;A8Cv4RE;EACE,oCAAmC;A9C04RvC;;A+C56RE;EACE,8BAAiC;A/C+6RrC;;A+Ch7RE;EACE,sCAAiC;A/Cm7RrC;;A+Cp7RE;EACE,iCAAiC;A/Cu7RrC;;A+Cx7RE;EACE,yCAAiC;A/C27RrC;;A+Cv7RE;EACE,4BAA4B;A/C07RhC;;A+C37RE;EACE,0BAA4B;A/C87RhC;;A+C/7RE;EACE,kCAA4B;A/Ck8RhC;;A+C97RE;EACE,sCAAkC;A/Ci8RtC;;A+Cl8RE;EACE,oCAAkC;A/Cq8RtC;;A+Ct8RE;EACE,kCAAkC;A/Cy8RtC;;A+C18RE;EACE,yCAAkC;A/C68RtC;;A+C98RE;EACE,wCAAkC;A/Ci9RtC;;A+Cl9RE;EACE,wCAAkC;A/Cq9RtC;;A+Ct9RE;EACE,iCAAkC;A/Cy9RtC;;A+C19RE;EACE,+BAAkC;A/C69RtC;;A+C99RE;EACE,gCAAkC;A/Ci+RtC;;A+Cl+RE;EACE,iCAAkC;A/Cq+RtC;;A+Cj+RE;EACE,oCAAgC;A/Co+RpC;;A+Cr+RE;EACE,kCAAgC;A/Cw+RpC;;A+Cz+RE;EACE,gCAAgC;A/C4+RpC;;A+C7+RE;EACE,uCAAgC;A/Cg/RpC;;A+Cj/RE;EACE,sCAAgC;A/Co/RpC;;A+Cr/RE;EACE,sCAAgC;A/Cw/RpC;;A+Cz/RE;EACE,iCAAgC;A/C4/RpC;;A+C7/RE;EACE,+BAAgC;A/CggSpC;;A+CjgSE;EACE,6BAAgC;A/CogSpC;;A+CrgSE;EACE,kCAAgC;A/CwgSpC;;A+CpgSE;EACE,+BAA8B;A/CugSlC;;A+CxgSE;EACE,kCAA8B;A/C2gSlC;;A+C5gSE;EACE,gCAA8B;A/C+gSlC;;A+ChhSE;EACE,8BAA8B;A/CmhSlC;;A+CphSE;EACE,gCAA8B;A/CuhSlC;;A+CxhSE;EACE,6BAA8B;A/C2hSlC;;A+C5hSE;EACE,2BAA8B;A/C+hSlC;;A+ChiSE;EACE,kCAA8B;A/CmiSlC;;A+CpiSE;EACE,gCAA8B;A/CuiSlC;;A+CniSE;EACE,2BAA6B;A/CsiSjC;;A+CviSE;EACE,iCAA6B;A/C0iSjC;;A+C3iSE;EACE,+BAA6B;A/C8iSjC;;A+C/iSE;EACE,6BAA6B;A/CkjSjC;;A+CnjSE;EACE,+BAA6B;A/CsjSjC;;A+CvjSE;EACE,8BAA6B;A/C0jSjC;;A+CrjSI;EACE,uBAAqC;A/CwjS3C;;A+CzjSI;EACE,uBAAqC;A/C4jS3C;;A+C7jSI;EACE,uBAAqC;A/CgkS3C;;A+CjkSI;EACE,uBAAqC;A/CokS3C;;A+CrkSI;EACE,uBAAqC;A/CwkS3C;;A+CzkSI;EACE,uBAAqC;A/C4kS3C;;A+C7kSI;EACE,yBAAqC;A/CglS3C;;A+CjlSI;EACE,yBAAqC;A/ColS3C;;A+CrlSI;EACE,yBAAqC;A/CwlS3C;;A+CzlSI;EACE,yBAAqC;A/C4lS3C;;A+C7lSI;EACE,yBAAqC;A/CgmS3C;;A+CjmSI;EACE,yBAAqC;A/ComS3C;;ACnoSE;EACE,WAAW;EACX,YAAY;EACZ,cAAc;ADsoSlB;;AgDzoSA;EACE,sBAAsB;AhD4oSxB;;AgD1oSA;EACE,uBAAuB;AhD6oSzB;;AiDppSA;EACE,2BAA2B;AjDupS7B;;AiDrpSA;EACE,2BAA2B;AjDwpS7B;;AiDtpSA;EACE,0BAA0B;AjDypS5B;;AkDhqSA;EACE,2BAA2B;AlDmqS7B;;AmDjqSA;EACE,6BAA6B;AnDoqS/B;;AoDxqSA;EACE,oBAAoB;ApD2qStB;;AoDzqSA;EACE,qBAAqB;ApD4qSvB;;AoDjqSI;EACE,oBAA+B;ApDoqSrC;;AoDjqSM;EACE,wBAA8C;ApDoqStD;;AoDrqSM;EACE,0BAA8C;ApDwqStD;;AoDzqSM;EACE,2BAA8C;ApD4qStD;;AoD7qSM;EACE,yBAA8C;ApDgrStD;;AoD7qSM;EACE,yBAAyC;EACzC,0BAA2C;ApDgrSnD;;AoD7qSM;EACE,wBAAuC;EACvC,2BAA6C;ApDgrSrD;;AoD/rSI;EACE,0BAA+B;ApDksSrC;;AoD/rSM;EACE,8BAA8C;ApDksStD;;AoDnsSM;EACE,gCAA8C;ApDssStD;;AoDvsSM;EACE,iCAA8C;ApD0sStD;;AoD3sSM;EACE,+BAA8C;ApD8sStD;;AoD3sSM;EACE,+BAAyC;EACzC,gCAA2C;ApD8sSnD;;AoD3sSM;EACE,8BAAuC;EACvC,iCAA6C;ApD8sSrD;;AoD7tSI;EACE,yBAA+B;ApDguSrC;;AoD7tSM;EACE,6BAA8C;ApDguStD;;AoDjuSM;EACE,+BAA8C;ApDouStD;;AoDruSM;EACE,gCAA8C;ApDwuStD;;AoDzuSM;EACE,8BAA8C;ApD4uStD;;AoDzuSM;EACE,8BAAyC;EACzC,+BAA2C;ApD4uSnD;;AoDzuSM;EACE,6BAAuC;EACvC,gCAA6C;ApD4uSrD;;AoD3vSI;EACE,0BAA+B;ApD8vSrC;;AoD3vSM;EACE,8BAA8C;ApD8vStD;;AoD/vSM;EACE,gCAA8C;ApDkwStD;;AoDnwSM;EACE,iCAA8C;ApDswStD;;AoDvwSM;EACE,+BAA8C;ApD0wStD;;AoDvwSM;EACE,+BAAyC;EACzC,gCAA2C;ApD0wSnD;;AoDvwSM;EACE,8BAAuC;EACvC,iCAA6C;ApD0wSrD;;AoDzxSI;EACE,uBAA+B;ApD4xSrC;;AoDzxSM;EACE,2BAA8C;ApD4xStD;;AoD7xSM;EACE,6BAA8C;ApDgyStD;;AoDjySM;EACE,8BAA8C;ApDoyStD;;AoDrySM;EACE,4BAA8C;ApDwyStD;;AoDrySM;EACE,4BAAyC;EACzC,6BAA2C;ApDwySnD;;AoDrySM;EACE,2BAAuC;EACvC,8BAA6C;ApDwySrD;;AoDvzSI;EACE,yBAA+B;ApD0zSrC;;AoDvzSM;EACE,6BAA8C;ApD0zStD;;AoD3zSM;EACE,+BAA8C;ApD8zStD;;AoD/zSM;EACE,gCAA8C;ApDk0StD;;AoDn0SM;EACE,8BAA8C;ApDs0StD;;AoDn0SM;EACE,8BAAyC;EACzC,+BAA2C;ApDs0SnD;;AoDn0SM;EACE,6BAAuC;EACvC,gCAA6C;ApDs0SrD;;AoDr1SI;EACE,uBAA+B;ApDw1SrC;;AoDr1SM;EACE,2BAA8C;ApDw1StD;;AoDz1SM;EACE,6BAA8C;ApD41StD;;AoD71SM;EACE,8BAA8C;ApDg2StD;;AoDj2SM;EACE,4BAA8C;ApDo2StD;;AoDj2SM;EACE,4BAAyC;EACzC,6BAA2C;ApDo2SnD;;AoDj2SM;EACE,2BAAuC;EACvC,8BAA6C;ApDo2SrD;;AoDn3SI;EACE,qBAA+B;ApDs3SrC;;AoDn3SM;EACE,yBAA8C;ApDs3StD;;AoDv3SM;EACE,2BAA8C;ApD03StD;;AoD33SM;EACE,4BAA8C;ApD83StD;;AoD/3SM;EACE,0BAA8C;ApDk4StD;;AoD/3SM;EACE,0BAAyC;EACzC,2BAA2C;ApDk4SnD;;AoD/3SM;EACE,yBAAuC;EACvC,4BAA6C;ApDk4SrD;;AoDj5SI;EACE,2BAA+B;ApDo5SrC;;AoDj5SM;EACE,+BAA8C;ApDo5StD;;AoDr5SM;EACE,iCAA8C;ApDw5StD;;AoDz5SM;EACE,kCAA8C;ApD45StD;;AoD75SM;EACE,gCAA8C;ApDg6StD;;AoD75SM;EACE,gCAAyC;EACzC,iCAA2C;ApDg6SnD;;AoD75SM;EACE,+BAAuC;EACvC,kCAA6C;ApDg6SrD;;AoD/6SI;EACE,0BAA+B;ApDk7SrC;;AoD/6SM;EACE,8BAA8C;ApDk7StD;;AoDn7SM;EACE,gCAA8C;ApDs7StD;;AoDv7SM;EACE,iCAA8C;ApD07StD;;AoD37SM;EACE,+BAA8C;ApD87StD;;AoD37SM;EACE,+BAAyC;EACzC,gCAA2C;ApD87SnD;;AoD37SM;EACE,8BAAuC;EACvC,iCAA6C;ApD87SrD;;AoD78SI;EACE,2BAA+B;ApDg9SrC;;AoD78SM;EACE,+BAA8C;ApDg9StD;;AoDj9SM;EACE,iCAA8C;ApDo9StD;;AoDr9SM;EACE,kCAA8C;ApDw9StD;;AoDz9SM;EACE,gCAA8C;ApD49StD;;AoDz9SM;EACE,gCAAyC;EACzC,iCAA2C;ApD49SnD;;AoDz9SM;EACE,+BAAuC;EACvC,kCAA6C;ApD49SrD;;AoD3+SI;EACE,wBAA+B;ApD8+SrC;;AoD3+SM;EACE,4BAA8C;ApD8+StD;;AoD/+SM;EACE,8BAA8C;ApDk/StD;;AoDn/SM;EACE,+BAA8C;ApDs/StD;;AoDv/SM;EACE,6BAA8C;ApD0/StD;;AoDv/SM;EACE,6BAAyC;EACzC,8BAA2C;ApD0/SnD;;AoDv/SM;EACE,4BAAuC;EACvC,+BAA6C;ApD0/SrD;;AoDzgTI;EACE,0BAA+B;ApD4gTrC;;AoDzgTM;EACE,8BAA8C;ApD4gTtD;;AoD7gTM;EACE,gCAA8C;ApDghTtD;;AoDjhTM;EACE,iCAA8C;ApDohTtD;;AoDrhTM;EACE,+BAA8C;ApDwhTtD;;AoDrhTM;EACE,+BAAyC;EACzC,gCAA2C;ApDwhTnD;;AoDrhTM;EACE,8BAAuC;EACvC,iCAA6C;ApDwhTrD;;AoDviTI;EACE,wBAA+B;ApD0iTrC;;AoDviTM;EACE,4BAA8C;ApD0iTtD;;AoD3iTM;EACE,8BAA8C;ApD8iTtD;;AoD/iTM;EACE,+BAA8C;ApDkjTtD;;AoDnjTM;EACE,6BAA8C;ApDsjTtD;;AoDnjTM;EACE,6BAAyC;EACzC,8BAA2C;ApDsjTnD;;AoDnjTM;EACE,4BAAuC;EACvC,+BAA6C;ApDsjTrD;;AqDjlTI;EACE,0BAA2B;ArDolTjC;;AqDrlTI;EACE,4BAA2B;ArDwlTjC;;AqDzlTI;EACE,0BAA2B;ArD4lTjC;;AqD7lTI;EACE,4BAA2B;ArDgmTjC;;AqDjmTI;EACE,6BAA2B;ArDomTjC;;AqDrmTI;EACE,0BAA2B;ArDwmTjC;;AqDzmTI;EACE,6BAA2B;ArD4mTjC;;AC/hTE;EoD9EE;IACE,0BAA2B;ErDinT/B;EqDlnTE;IACE,4BAA2B;ErDonT/B;EqDrnTE;IACE,0BAA2B;ErDunT/B;EqDxnTE;IACE,4BAA2B;ErD0nT/B;EqD3nTE;IACE,6BAA2B;ErD6nT/B;EqD9nTE;IACE,0BAA2B;ErDgoT/B;EqDjoTE;IACE,6BAA2B;ErDmoT/B;AACF;;ACnjTE;EoDlFE;IACE,0BAA2B;ErDyoT/B;EqD1oTE;IACE,4BAA2B;ErD4oT/B;EqD7oTE;IACE,0BAA2B;ErD+oT/B;EqDhpTE;IACE,4BAA2B;ErDkpT/B;EqDnpTE;IACE,6BAA2B;ErDqpT/B;EqDtpTE;IACE,0BAA2B;ErDwpT/B;EqDzpTE;IACE,6BAA2B;ErD2pT/B;AACF;;ACnkTE;EoD1FE;IACE,0BAA2B;ErDiqT/B;EqDlqTE;IACE,4BAA2B;ErDoqT/B;EqDrqTE;IACE,0BAA2B;ErDuqT/B;EqDxqTE;IACE,4BAA2B;ErD0qT/B;EqD3qTE;IACE,6BAA2B;ErD6qT/B;EqD9qTE;IACE,0BAA2B;ErDgrT/B;EqDjrTE;IACE,6BAA2B;ErDmrT/B;AACF;;ACvlTE;EoD9FE;IACE,0BAA2B;ErDyrT/B;EqD1rTE;IACE,4BAA2B;ErD4rT/B;EqD7rTE;IACE,0BAA2B;ErD+rT/B;EqDhsTE;IACE,4BAA2B;ErDksT/B;EqDnsTE;IACE,6BAA2B;ErDqsT/B;EqDtsTE;IACE,0BAA2B;ErDwsT/B;EqDzsTE;IACE,6BAA2B;ErD2sT/B;AACF;;AChmTI;EoD7GA;IACE,0BAA2B;ErDitT/B;EqDltTE;IACE,4BAA2B;ErDotT/B;EqDrtTE;IACE,0BAA2B;ErDutT/B;EqDxtTE;IACE,4BAA2B;ErD0tT/B;EqD3tTE;IACE,6BAA2B;ErD6tT/B;EqD9tTE;IACE,0BAA2B;ErDguT/B;EqDjuTE;IACE,6BAA2B;ErDmuT/B;AACF;;ACzmTI;EoD5HA;IACE,0BAA2B;ErDyuT/B;EqD1uTE;IACE,4BAA2B;ErD4uT/B;EqD7uTE;IACE,0BAA2B;ErD+uT/B;EqDhvTE;IACE,4BAA2B;ErDkvT/B;EqDnvTE;IACE,6BAA2B;ErDqvT/B;EqDtvTE;IACE,0BAA2B;ErDwvT/B;EqDzvTE;IACE,6BAA2B;ErD2vT/B;AACF;;AqDnuTE;EACE,6BAAqC;ArDsuTzC;;AqDvuTE;EACE,8BAAqC;ArD0uTzC;;AqD3uTE;EACE,2BAAqC;ArD8uTzC;;AqD/uTE;EACE,4BAAqC;ArDkvTzC;;AC/rTE;EoD/CE;IACE,6BAAqC;ErDkvTzC;AACF;;ACjsTE;EoDhDE;IACE,6BAAqC;ErDqvTzC;AACF;;ACnsTE;EoDjDE;IACE,6BAAqC;ErDwvTzC;AACF;;ACrsTE;EoDlDE;IACE,6BAAqC;ErD2vTzC;AACF;;ACvsTE;EoDnDE;IACE,6BAAqC;ErD8vTzC;AACF;;ACxsTI;EoDrDA;IACE,6BAAqC;ErDiwTzC;AACF;;ACpsTI;EoD5DA;IACE,6BAAqC;ErDowTzC;AACF;;ACrsTI;EoD9DA;IACE,6BAAqC;ErDuwTzC;AACF;;ACjsTI;EoDrEA;IACE,6BAAqC;ErD0wTzC;AACF;;ACrvTE;EoD/CE;IACE,8BAAqC;ErDwyTzC;AACF;;ACvvTE;EoDhDE;IACE,8BAAqC;ErD2yTzC;AACF;;ACzvTE;EoDjDE;IACE,8BAAqC;ErD8yTzC;AACF;;AC3vTE;EoDlDE;IACE,8BAAqC;ErDizTzC;AACF;;AC7vTE;EoDnDE;IACE,8BAAqC;ErDozTzC;AACF;;AC9vTI;EoDrDA;IACE,8BAAqC;ErDuzTzC;AACF;;AC1vTI;EoD5DA;IACE,8BAAqC;ErD0zTzC;AACF;;AC3vTI;EoD9DA;IACE,8BAAqC;ErD6zTzC;AACF;;ACvvTI;EoDrEA;IACE,8BAAqC;ErDg0TzC;AACF;;AC3yTE;EoD/CE;IACE,2BAAqC;ErD81TzC;AACF;;AC7yTE;EoDhDE;IACE,2BAAqC;ErDi2TzC;AACF;;AC/yTE;EoDjDE;IACE,2BAAqC;ErDo2TzC;AACF;;ACjzTE;EoDlDE;IACE,2BAAqC;ErDu2TzC;AACF;;ACnzTE;EoDnDE;IACE,2BAAqC;ErD02TzC;AACF;;ACpzTI;EoDrDA;IACE,2BAAqC;ErD62TzC;AACF;;AChzTI;EoD5DA;IACE,2BAAqC;ErDg3TzC;AACF;;ACjzTI;EoD9DA;IACE,2BAAqC;ErDm3TzC;AACF;;AC7yTI;EoDrEA;IACE,2BAAqC;ErDs3TzC;AACF;;ACj2TE;EoD/CE;IACE,4BAAqC;ErDo5TzC;AACF;;ACn2TE;EoDhDE;IACE,4BAAqC;ErDu5TzC;AACF;;ACr2TE;EoDjDE;IACE,4BAAqC;ErD05TzC;AACF;;ACv2TE;EoDlDE;IACE,4BAAqC;ErD65TzC;AACF;;ACz2TE;EoDnDE;IACE,4BAAqC;ErDg6TzC;AACF;;AC12TI;EoDrDA;IACE,4BAAqC;ErDm6TzC;AACF;;ACt2TI;EoD5DA;IACE,4BAAqC;ErDs6TzC;AACF;;ACv2TI;EoD9DA;IACE,4BAAqC;ErDy6TzC;AACF;;ACn2TI;EoDrEA;IACE,4BAAqC;ErD46TzC;AACF;;AqD36TA;EACE,qCAAqC;ArD86TvC;;AqD56TA;EACE,oCAAoC;ArD+6TtC;;AqD76TA;EACE,oCAAoC;ArDg7TtC;;AqD96TA;EACE,6BAA6B;ArDi7T/B;;AqD/6TA;EACE,2BAAqC;ArDk7TvC;;AqDj7TA;EACE,2BAAsC;ArDo7TxC;;AqDn7TA;EACE,2BAAsC;ArDs7TxC;;AqDr7TA;EACE,2BAAwC;ArDw7T1C;;AqDv7TA;EACE,2BAAoC;ArD07TtC;;AqDx7TA;EACE,+LAAuC;ArD27TzC;;AqDz7TA;EACE,+LAAyC;ArD47T3C;;AqD17TA;EACE,+LAA0C;ArD67T5C;;AqD37TA;EACE,iCAAyC;ArD87T3C;;AqD57TA;EACE,iCAAoC;ArD+7TtC;;AsD3hUE;EACE,yBAA+B;AtD8hUnC;;ACn9TE;EqDzEE;IACE,yBAA+B;EtDgiUnC;AACF;;ACr9TE;EqD1EE;IACE,yBAA+B;EtDmiUnC;AACF;;ACv9TE;EqD3EE;IACE,yBAA+B;EtDsiUnC;AACF;;ACz9TE;EqD5EE;IACE,yBAA+B;EtDyiUnC;AACF;;AC39TE;EqD7EE;IACE,yBAA+B;EtD4iUnC;AACF;;AC59TI;EqD/EA;IACE,yBAA+B;EtD+iUnC;AACF;;ACx9TI;EqDtFA;IACE,yBAA+B;EtDkjUnC;AACF;;ACz9TI;EqDxFA;IACE,yBAA+B;EtDqjUnC;AACF;;ACr9TI;EqD/FA;IACE,yBAA+B;EtDwjUnC;AACF;;AsDrlUE;EACE,wBAA+B;AtDwlUnC;;AC7gUE;EqDzEE;IACE,wBAA+B;EtD0lUnC;AACF;;AC/gUE;EqD1EE;IACE,wBAA+B;EtD6lUnC;AACF;;ACjhUE;EqD3EE;IACE,wBAA+B;EtDgmUnC;AACF;;ACnhUE;EqD5EE;IACE,wBAA+B;EtDmmUnC;AACF;;ACrhUE;EqD7EE;IACE,wBAA+B;EtDsmUnC;AACF;;ACthUI;EqD/EA;IACE,wBAA+B;EtDymUnC;AACF;;AClhUI;EqDtFA;IACE,wBAA+B;EtD4mUnC;AACF;;ACnhUI;EqDxFA;IACE,wBAA+B;EtD+mUnC;AACF;;AC/gUI;EqD/FA;IACE,wBAA+B;EtDknUnC;AACF;;AsD/oUE;EACE,0BAA+B;AtDkpUnC;;ACvkUE;EqDzEE;IACE,0BAA+B;EtDopUnC;AACF;;ACzkUE;EqD1EE;IACE,0BAA+B;EtDupUnC;AACF;;AC3kUE;EqD3EE;IACE,0BAA+B;EtD0pUnC;AACF;;AC7kUE;EqD5EE;IACE,0BAA+B;EtD6pUnC;AACF;;AC/kUE;EqD7EE;IACE,0BAA+B;EtDgqUnC;AACF;;AChlUI;EqD/EA;IACE,0BAA+B;EtDmqUnC;AACF;;AC5kUI;EqDtFA;IACE,0BAA+B;EtDsqUnC;AACF;;AC7kUI;EqDxFA;IACE,0BAA+B;EtDyqUnC;AACF;;ACzkUI;EqD/FA;IACE,0BAA+B;EtD4qUnC;AACF;;AsDzsUE;EACE,gCAA+B;AtD4sUnC;;ACjoUE;EqDzEE;IACE,gCAA+B;EtD8sUnC;AACF;;ACnoUE;EqD1EE;IACE,gCAA+B;EtDitUnC;AACF;;ACroUE;EqD3EE;IACE,gCAA+B;EtDotUnC;AACF;;ACvoUE;EqD5EE;IACE,gCAA+B;EtDutUnC;AACF;;ACzoUE;EqD7EE;IACE,gCAA+B;EtD0tUnC;AACF;;AC1oUI;EqD/EA;IACE,gCAA+B;EtD6tUnC;AACF;;ACtoUI;EqDtFA;IACE,gCAA+B;EtDguUnC;AACF;;ACvoUI;EqDxFA;IACE,gCAA+B;EtDmuUnC;AACF;;ACnoUI;EqD/FA;IACE,gCAA+B;EtDsuUnC;AACF;;AsDnwUE;EACE,+BAA+B;AtDswUnC;;AC3rUE;EqDzEE;IACE,+BAA+B;EtDwwUnC;AACF;;AC7rUE;EqD1EE;IACE,+BAA+B;EtD2wUnC;AACF;;AC/rUE;EqD3EE;IACE,+BAA+B;EtD8wUnC;AACF;;ACjsUE;EqD5EE;IACE,+BAA+B;EtDixUnC;AACF;;ACnsUE;EqD7EE;IACE,+BAA+B;EtDoxUnC;AACF;;ACpsUI;EqD/EA;IACE,+BAA+B;EtDuxUnC;AACF;;AChsUI;EqDtFA;IACE,+BAA+B;EtD0xUnC;AACF;;ACjsUI;EqDxFA;IACE,+BAA+B;EtD6xUnC;AACF;;AC7rUI;EqD/FA;IACE,+BAA+B;EtDgyUnC;AACF;;AsD/xUA;EACE,wBAAwB;AtDkyU1B;;AsDhyUA;EACE,uBAAuB;EACvB,iCAAiC;EACjC,yBAAyB;EACzB,2BAA2B;EAC3B,qBAAqB;EACrB,6BAA6B;EAC7B,8BAA8B;EAC9B,wBAAwB;AtDmyU1B;;AChwUE;EqDhCA;IACE,wBAAwB;EtDoyU1B;AACF;;AClwUE;EqDhCA;IACE,wBAAwB;EtDsyU1B;AACF;;ACpwUE;EqDhCA;IACE,wBAAwB;EtDwyU1B;AACF;;ACtwUE;EqDhCA;IACE,wBAAwB;EtD0yU1B;AACF;;ACxwUE;EqDhCA;IACE,wBAAwB;EtD4yU1B;AACF;;ACzwUI;EqDjCF;IACE,wBAAwB;EtD8yU1B;AACF;;ACrwUI;EqDvCF;IACE,wBAAwB;EtDgzU1B;AACF;;ACtwUI;EqDxCF;IACE,wBAAwB;EtDkzU1B;AACF;;AClwUI;EqD9CF;IACE,wBAAwB;EtDozU1B;AACF;;AsDnzUA;EACE,6BAA6B;AtDszU/B;;AC1zUE;EqDOA;IACE,6BAA6B;EtDuzU/B;AACF;;AC5zUE;EqDOA;IACE,6BAA6B;EtDyzU/B;AACF;;AC9zUE;EqDOA;IACE,6BAA6B;EtD2zU/B;AACF;;ACh0UE;EqDOA;IACE,6BAA6B;EtD6zU/B;AACF;;ACl0UE;EqDOA;IACE,6BAA6B;EtD+zU/B;AACF;;ACn0UI;EqDMF;IACE,6BAA6B;EtDi0U/B;AACF;;AC/zUI;EqDAF;IACE,6BAA6B;EtDm0U/B;AACF;;ACh0UI;EqDDF;IACE,6BAA6B;EtDq0U/B;AACF;;AC5zUI;EqDPF;IACE,6BAA6B;EtDu0U/B;AACF;;AuDj8UA,iBAAA;ACQA;EACE,oBAAoB;EACpB,aAAa;EACb,sBAAsB;EACtB,8BAA8B;AxD67UhC;;AwDj8UA;EAMI,gBAAgB;AxD+7UpB;;AwDr8UA;EASM,mBAAmB;AxDg8UzB;;AwDz8UA;EAeM,uBtDRyB;EsDSzB,ctDtBuB;AFo9U7B;;AwD98UA;;EAmBQ,cAAc;AxDg8UtB;;AwDn9UA;EAqBQ,ctD3BqB;AF69U7B;;AwDv9UA;EAuBQ,4BtD7BqB;AFi+U7B;;AwD39UA;;EA0BU,ctDhCmB;AFs+U7B;;AC34UE;EuDrFF;IA6BU,uBtDtBqB;EF89U7B;AACF;;AwDt+UA;;EAgCQ,4BtDtCqB;AFi/U7B;;AwD3+UA;;;EAqCU,yB7CgEuB;E6C/DvB,ctD5CmB;AFw/U7B;;AwDl/UA;EAyCU,ctD/CmB;EsDgDnB,YAAY;AxD68UtB;;AwDv/UA;EA4CY,UAAU;AxD+8UtB;;AwD3/UA;EA+CY,UAAU;AxDg9UtB;;AwD//UA;EAmDY,ctDzDiB;AFygV7B;;AwDngVA;EAqDc,uCtD3De;AF6gV7B;;AwDvgVA;EAyDc,yBtD/De;EsDgEf,qBtDhEe;EsDiEf,YtDpDiB;AFsgV/B;;AwD7gVA;EAiEU,4EAAyG;AxDg9UnH;;ACx8UE;EuDzEF;IAoEc,4EAAyG;ExDk9UrH;AACF;;AwDvhVA;EAeM,yBtDrBuB;EsDsBvB,YtDTyB;AFqhV/B;;AwD5hVA;;EAmBQ,cAAc;AxD8gVtB;;AwDjiVA;EAqBQ,YtDduB;AF8hV/B;;AwDriVA;EAuBQ,+BtDhBuB;AFkiV/B;;AwDziVA;;EA0BU,YtDnBqB;AFuiV/B;;ACz9UE;EuDrFF;IA6BU,yBtDnCmB;EFyjV3B;AACF;;AwDpjVA;;EAgCQ,+BtDzBuB;AFkjV/B;;AwDzjVA;;;EAqCU,uB7CgEuB;E6C/DvB,YtD/BqB;AFyjV/B;;AwDhkVA;EAyCU,YtDlCqB;EsDmCrB,YAAY;AxD2hVtB;;AwDrkVA;EA4CY,UAAU;AxD6hVtB;;AwDzkVA;EA+CY,UAAU;AxD8hVtB;;AwD7kVA;EAmDY,YtD5CmB;AF0kV/B;;AwDjlVA;EAqDc,uCtD3De;AF2lV7B;;AwDrlVA;EAyDc,uBtDlDiB;EsDmDjB,mBtDnDiB;EsDoDjB,ctDjEe;AFimV7B;;AwD3lVA;EAiEU,8EAAyG;AxD8hVnH;;ACthVE;EuDzEF;IAoEc,8EAAyG;ExDgiVrH;AACF;;AwDrmVA;EAeM,4BtDVwB;EsDWxB,yB7CwDe;AXkiVrB;;AwD1mVA;;EAmBQ,cAAc;AxD4lVtB;;AwD/mVA;EAqBQ,yB7CmDa;AX2iVrB;;AwDnnVA;EAuBQ,yB7CiDa;AX+iVrB;;AwDvnVA;;EA0BU,yB7C8CW;AXojVrB;;ACviVE;EuDrFF;IA6BU,4BtDxBoB;EF4nV5B;AACF;;AwDloVA;;EAgCQ,yB7CwCa;AX+jVrB;;AwDvoVA;;;EAqCU,yB7CgEuB;E6C/DvB,yB7CkCW;AXskVrB;;AwD9oVA;EAyCU,yB7C+BW;E6C9BX,YAAY;AxDymVtB;;AwDnpVA;EA4CY,UAAU;AxD2mVtB;;AwDvpVA;EA+CY,UAAU;AxD4mVtB;;AwD3pVA;EAmDY,yB7CqBS;AXulVrB;;AwD/pVA;EAqDc,uCtD3De;AFyqV7B;;AwDnqVA;EAyDc,oC7CeO;E6CdP,gC7CcO;E6CbP,iBtDtDgB;AFoqV9B;;AwDzqVA;EAiEU,iFAAyG;AxD4mVnH;;ACpmVE;EuDzEF;IAoEc,iFAAyG;ExD8mVrH;AACF;;AwDnrVA;EAeM,yBtDjBwB;EsDkBxB,W7C0DU;AX8mVhB;;AwDxrVA;;EAmBQ,cAAc;AxD0qVtB;;AwD7rVA;EAqBQ,W7CqDQ;AXunVhB;;AwDjsVA;EAuBQ,+B7CmDQ;AX2nVhB;;AwDrsVA;;EA0BU,W7CgDM;AXgoVhB;;ACrnVE;EuDrFF;IA6BU,yBtD/BoB;EFitV5B;AACF;;AwDhtVA;;EAgCQ,+B7C0CQ;AX2oVhB;;AwDrtVA;;;EAqCU,yB7CgEuB;E6C/DvB,W7CoCM;AXkpVhB;;AwD5tVA;EAyCU,W7CiCM;E6ChCN,YAAY;AxDurVtB;;AwDjuVA;EA4CY,UAAU;AxDyrVtB;;AwDruVA;EA+CY,UAAU;AxD0rVtB;;AwDzuVA;EAmDY,W7CuBI;AXmqVhB;;AwD7uVA;EAqDc,uCtD3De;AFuvV7B;;AwDjvVA;EAyDc,sB7CiBE;E6ChBF,kB7CgBE;E6CfF,ctD7DgB;AFyvV9B;;AwDvvVA;EAiEU,gFAAyG;AxD0rVnH;;AClrVE;EuDzEF;IAoEc,gFAAyG;ExD4rVrH;AACF;;AwDjwVA;EAeM,yBtDH4B;EsDI5B,W7C0DU;AX4rVhB;;AwDtwVA;;EAmBQ,cAAc;AxDwvVtB;;AwD3wVA;EAqBQ,W7CqDQ;AXqsVhB;;AwD/wVA;EAuBQ,+B7CmDQ;AXysVhB;;AwDnxVA;;EA0BU,W7CgDM;AX8sVhB;;ACnsVE;EuDrFF;IA6BU,yBtDjBwB;EFixVhC;AACF;;AwD9xVA;;EAgCQ,+B7C0CQ;AXytVhB;;AwDnyVA;;;EAqCU,yB7CgEuB;E6C/DvB,W7CoCM;AXguVhB;;AwD1yVA;EAyCU,W7CiCM;E6ChCN,YAAY;AxDqwVtB;;AwD/yVA;EA4CY,UAAU;AxDuwVtB;;AwDnzVA;EA+CY,UAAU;AxDwwVtB;;AwDvzVA;EAmDY,W7CuBI;AXivVhB;;AwD3zVA;EAqDc,uCtD3De;AFq0V7B;;AwD/zVA;EAyDc,sB7CiBE;E6ChBF,kB7CgBE;E6CfF,ctD/CoB;AFyzVlC;;AwDr0VA;EAiEU,gFAAyG;AxDwwVnH;;AChwVE;EuDzEF;IAoEc,gFAAyG;ExD0wVrH;AACF;;AwD/0VA;EAeM,yBtDD4B;EsDE5B,W7C0DU;AX0wVhB;;AwDp1VA;;EAmBQ,cAAc;AxDs0VtB;;AwDz1VA;EAqBQ,W7CqDQ;AXmxVhB;;AwD71VA;EAuBQ,+B7CmDQ;AXuxVhB;;AwDj2VA;;EA0BU,W7CgDM;AX4xVhB;;ACjxVE;EuDrFF;IA6BU,yBtDfwB;EF61VhC;AACF;;AwD52VA;;EAgCQ,+B7C0CQ;AXuyVhB;;AwDj3VA;;;EAqCU,yB7CgEuB;E6C/DvB,W7CoCM;AX8yVhB;;AwDx3VA;EAyCU,W7CiCM;E6ChCN,YAAY;AxDm1VtB;;AwD73VA;EA4CY,UAAU;AxDq1VtB;;AwDj4VA;EA+CY,UAAU;AxDs1VtB;;AwDr4VA;EAmDY,W7CuBI;AX+zVhB;;AwDz4VA;EAqDc,uCtD3De;AFm5V7B;;AwD74VA;EAyDc,sB7CiBE;E6ChBF,kB7CgBE;E6CfF,ctD7CoB;AFq4VlC;;AwDn5VA;EAiEU,gFAAyG;AxDs1VnH;;AC90VE;EuDzEF;IAoEc,gFAAyG;ExDw1VrH;AACF;;AwD75VA;EAeM,yBtDF4B;EsDG5B,W7C0DU;AXw1VhB;;AwDl6VA;;EAmBQ,cAAc;AxDo5VtB;;AwDv6VA;EAqBQ,W7CqDQ;AXi2VhB;;AwD36VA;EAuBQ,+B7CmDQ;AXq2VhB;;AwD/6VA;;EA0BU,W7CgDM;AX02VhB;;AC/1VE;EuDrFF;IA6BU,yBtDhBwB;EF46VhC;AACF;;AwD17VA;;EAgCQ,+B7C0CQ;AXq3VhB;;AwD/7VA;;;EAqCU,yB7CgEuB;E6C/DvB,W7CoCM;AX43VhB;;AwDt8VA;EAyCU,W7CiCM;E6ChCN,YAAY;AxDi6VtB;;AwD38VA;EA4CY,UAAU;AxDm6VtB;;AwD/8VA;EA+CY,UAAU;AxDo6VtB;;AwDn9VA;EAmDY,W7CuBI;AX64VhB;;AwDv9VA;EAqDc,uCtD3De;AFi+V7B;;AwD39VA;EAyDc,sB7CiBE;E6ChBF,kB7CgBE;E6CfF,ctD9CoB;AFo9VlC;;AwDj+VA;EAiEU,gFAAyG;AxDo6VnH;;AC55VE;EuDzEF;IAoEc,gFAAyG;ExDs6VrH;AACF;;AwD3+VA;EAeM,yBtDJ4B;EsDK5B,W7C0DU;AXs6VhB;;AwDh/VA;;EAmBQ,cAAc;AxDk+VtB;;AwDr/VA;EAqBQ,W7CqDQ;AX+6VhB;;AwDz/VA;EAuBQ,+B7CmDQ;AXm7VhB;;AwD7/VA;;EA0BU,W7CgDM;AXw7VhB;;AC76VE;EuDrFF;IA6BU,yBtDlBwB;EF4/VhC;AACF;;AwDxgWA;;EAgCQ,+B7C0CQ;AXm8VhB;;AwD7gWA;;;EAqCU,yB7CgEuB;E6C/DvB,W7CoCM;AX08VhB;;AwDphWA;EAyCU,W7CiCM;E6ChCN,YAAY;AxD++VtB;;AwDzhWA;EA4CY,UAAU;AxDi/VtB;;AwD7hWA;EA+CY,UAAU;AxDk/VtB;;AwDjiWA;EAmDY,W7CuBI;AX29VhB;;AwDriWA;EAqDc,uCtD3De;AF+iW7B;;AwDziWA;EAyDc,sB7CiBE;E6ChBF,kB7CgBE;E6CfF,ctDhDoB;AFoiWlC;;AwD/iWA;EAiEU,gFAAyG;AxDk/VnH;;AC1+VE;EuDzEF;IAoEc,gFAAyG;ExDo/VrH;AACF;;AwDzjWA;EAeM,yBtDL4B;EsDM5B,yB7CwDe;AXs/VrB;;AwD9jWA;;EAmBQ,cAAc;AxDgjWtB;;AwDnkWA;EAqBQ,yB7CmDa;AX+/VrB;;AwDvkWA;EAuBQ,yB7CiDa;AXmgWrB;;AwD3kWA;;EA0BU,yB7C8CW;AXwgWrB;;AC3/VE;EuDrFF;IA6BU,yBtDnBwB;EF2kWhC;AACF;;AwDtlWA;;EAgCQ,yB7CwCa;AXmhWrB;;AwD3lWA;;;EAqCU,yB7CgEuB;E6C/DvB,yB7CkCW;AX0hWrB;;AwDlmWA;EAyCU,yB7C+BW;E6C9BX,YAAY;AxD6jWtB;;AwDvmWA;EA4CY,UAAU;AxD+jWtB;;AwD3mWA;EA+CY,UAAU;AxDgkWtB;;AwD/mWA;EAmDY,yB7CqBS;AX2iWrB;;AwDnnWA;EAqDc,uCtD3De;AF6nW7B;;AwDvnWA;EAyDc,oC7CeO;E6CdP,gC7CcO;E6CbP,ctDjDoB;AFmnWlC;;AwD7nWA;EAiEU,gFAAyG;AxDgkWnH;;ACxjWE;EuDzEF;IAoEc,gFAAyG;ExDkkWrH;AACF;;AwDvoWA;EAeM,yBtDC2B;EsDA3B,W7C0DU;AXkkWhB;;AwD5oWA;;EAmBQ,cAAc;AxD8nWtB;;AwDjpWA;EAqBQ,W7CqDQ;AX2kWhB;;AwDrpWA;EAuBQ,+B7CmDQ;AX+kWhB;;AwDzpWA;;EA0BU,W7CgDM;AXolWhB;;ACzkWE;EuDrFF;IA6BU,yBtDbuB;EFmpW/B;AACF;;AwDpqWA;;EAgCQ,+B7C0CQ;AX+lWhB;;AwDzqWA;;;EAqCU,yB7CgEuB;E6C/DvB,W7CoCM;AXsmWhB;;AwDhrWA;EAyCU,W7CiCM;E6ChCN,YAAY;AxD2oWtB;;AwDrrWA;EA4CY,UAAU;AxD6oWtB;;AwDzrWA;EA+CY,UAAU;AxD8oWtB;;AwD7rWA;EAmDY,W7CuBI;AXunWhB;;AwDjsWA;EAqDc,uCtD3De;AF2sW7B;;AwDrsWA;EAyDc,sB7CiBE;E6ChBF,kB7CgBE;E6CfF,ctD3CmB;AF2rWjC;;AwD3sWA;EAiEU,gFAAyG;AxD8oWnH;;ACtoWE;EuDzEF;IAoEc,gFAAyG;ExDgpWrH;AACF;;AwDrtWA;EAwEM,eA/E0B;AxDguWhC;;AC5oWE;EuD7EF;IA4EQ,oBAlF8B;ExDouWpC;AACF;;AClpWE;EuD7EF;IAgFQ,qBArF8B;ExDyuWpC;AACF;;AwDruWA;EAqFM,mBAAmB;EACnB,aAAa;AxDopWnB;;AwD1uWA;EAwFQ,YAAY;EACZ,cAAc;AxDspWtB;;AwD/uWA;EA2FI,gBAAgB;AxDwpWpB;;AwDnvWA;EA6FI,iBAAiB;AxD0pWrB;;AwDtpWA;EAEE,gBAAgB;AxDwpWlB;;AwD1pWA;EAII,SAAS;EACT,gBAAgB;EAChB,eAAe;EACf,kBAAkB;EAClB,QAAQ;EACR,qCAAqC;AxD0pWzC;;AwDnqWA;EAYI,YAAY;AxD2pWhB;;AC/rWE;EuDwBF;IAeI,aAAa;ExD6pWf;AACF;;AwD5pWA;EACE,kBAAkB;AxD+pWpB;;ACzsWE;EuDyCF;IAKM,aAAa;ExDgqWjB;EwDrqWF;IAOQ,sBAAsB;ExDiqW5B;AACF;;AC9sWE;EuDqCF;IASI,aAAa;IACb,uBAAuB;ExDqqWzB;EwD/qWF;IvDsBI,oBuDVwC;ExDsqW1C;AACF;;AwDnqWA;;EAEE,YAAY;EACZ,cAAc;AxDsqWhB;;AwDpqWA;EACE,YAAY;EACZ,cAAc;EACd,oBAlJ6B;AxDyzW/B;;AyDrzWA;EACE,oBAL2B;AzD6zW7B;;AC5tWE;EwD7FF;IAMM,oBAT8B;EzDi0WlC;EyD9zWF;IAQM,qBAV8B;EzDm0WlC;AACF;;A0Dl0WA;EACE,yBxDS4B;EwDR5B,yBAJ+B;A1Dy0WjC\",\"file\":\"bulma.css\"}" }, { "path": "docs/static/css/index.css", "content": "body {\n font-family: 'Noto Sans', sans-serif;\n}\n\n\n.footer .icon-link {\n font-size: 25px;\n color: #000;\n}\n\n.link-block a {\n margin-top: 5px;\n margin-bottom: 5px;\n}\n\n.dnerf {\n font-variant: small-caps;\n}\n\n\n.teaser .hero-body {\n padding-top: 0;\n padding-bottom: 3rem;\n}\n\n.teaser {\n font-family: 'Google Sans', sans-serif;\n}\n\n\n.publication-title {\n}\n\n.publication-banner {\n max-height: parent;\n\n}\n\n.publication-banner video {\n position: relative;\n left: auto;\n top: auto;\n transform: none;\n object-fit: fit;\n}\n\n.publication-header .hero-body {\n}\n\n.publication-title {\n font-family: 'Google Sans', sans-serif;\n}\n\n.publication-authors {\n font-family: 'Google Sans', sans-serif;\n}\n\n.publication-venue {\n color: #555;\n width: fit-content;\n font-weight: bold;\n}\n\n.publication-awards {\n color: #ff3860;\n width: fit-content;\n font-weight: bolder;\n}\n\n.publication-authors {\n}\n\n.publication-authors a {\n color: hsl(204, 86%, 53%) !important;\n}\n\n.publication-authors a:hover {\n text-decoration: underline;\n}\n\n.author-block {\n display: inline-block;\n}\n\n.publication-banner img {\n}\n\n.publication-authors {\n /*color: #4286f4;*/\n}\n\n.publication-video {\n position: relative;\n width: 100%;\n height: 0;\n padding-bottom: 56.25%;\n\n overflow: hidden;\n border-radius: 10px !important;\n}\n\n.publication-video iframe {\n position: absolute;\n top: 0;\n left: 0;\n width: 100%;\n height: 100%;\n}\n\n.publication-body img {\n}\n\n.results-carousel {\n overflow: hidden;\n}\n\n.results-carousel .item {\n margin: 5px;\n overflow: hidden;\n border: 1px solid #bbb;\n border-radius: 10px;\n padding: 0;\n font-size: 0;\n}\n\n.results-carousel video {\n margin: 0;\n}\n\n\n.interpolation-panel {\n background: #f5f5f5;\n border-radius: 10px;\n}\n\n.interpolation-panel .interpolation-image {\n width: 100%;\n border-radius: 5px;\n}\n\n.interpolation-video-column {\n}\n\n.interpolation-panel .slider {\n margin: 0 !important;\n}\n\n.interpolation-panel .slider {\n margin: 0 !important;\n}\n\n#interpolation-image-wrapper {\n width: 100%;\n}\n#interpolation-image-wrapper img {\n border-radius: 5px;\n}\n" }, { "path": "docs/static/js/bulma-carousel.js", "content": "(function webpackUniversalModuleDefinition(root, factory) {\n\tif(typeof exports === 'object' && typeof module === 'object')\n\t\tmodule.exports = factory();\n\telse if(typeof define === 'function' && define.amd)\n\t\tdefine([], factory);\n\telse if(typeof exports === 'object')\n\t\texports[\"bulmaCarousel\"] = factory();\n\telse\n\t\troot[\"bulmaCarousel\"] = factory();\n})(typeof self !== 'undefined' ? self : this, function() {\nreturn /******/ (function(modules) { // webpackBootstrap\n/******/ \t// The module cache\n/******/ \tvar installedModules = {};\n/******/\n/******/ \t// The require function\n/******/ \tfunction __webpack_require__(moduleId) {\n/******/\n/******/ \t\t// Check if module is in cache\n/******/ \t\tif(installedModules[moduleId]) {\n/******/ \t\t\treturn installedModules[moduleId].exports;\n/******/ \t\t}\n/******/ \t\t// Create a new module (and put it into the cache)\n/******/ \t\tvar module = installedModules[moduleId] = {\n/******/ \t\t\ti: moduleId,\n/******/ \t\t\tl: false,\n/******/ \t\t\texports: {}\n/******/ \t\t};\n/******/\n/******/ \t\t// Execute the module function\n/******/ \t\tmodules[moduleId].call(module.exports, module, module.exports, __webpack_require__);\n/******/\n/******/ \t\t// Flag the module as loaded\n/******/ \t\tmodule.l = true;\n/******/\n/******/ \t\t// Return the exports of the module\n/******/ \t\treturn module.exports;\n/******/ \t}\n/******/\n/******/\n/******/ \t// expose the modules object (__webpack_modules__)\n/******/ \t__webpack_require__.m = modules;\n/******/\n/******/ \t// expose the module cache\n/******/ \t__webpack_require__.c = installedModules;\n/******/\n/******/ \t// define getter function for harmony exports\n/******/ \t__webpack_require__.d = function(exports, name, getter) {\n/******/ \t\tif(!__webpack_require__.o(exports, name)) {\n/******/ \t\t\tObject.defineProperty(exports, name, {\n/******/ \t\t\t\tconfigurable: false,\n/******/ \t\t\t\tenumerable: true,\n/******/ \t\t\t\tget: getter\n/******/ \t\t\t});\n/******/ \t\t}\n/******/ \t};\n/******/\n/******/ \t// getDefaultExport function for compatibility with non-harmony modules\n/******/ \t__webpack_require__.n = function(module) {\n/******/ \t\tvar getter = module && module.__esModule ?\n/******/ \t\t\tfunction getDefault() { return module['default']; } :\n/******/ \t\t\tfunction getModuleExports() { return module; };\n/******/ \t\t__webpack_require__.d(getter, 'a', getter);\n/******/ \t\treturn getter;\n/******/ \t};\n/******/\n/******/ \t// Object.prototype.hasOwnProperty.call\n/******/ \t__webpack_require__.o = function(object, property) { return Object.prototype.hasOwnProperty.call(object, property); };\n/******/\n/******/ \t// __webpack_public_path__\n/******/ \t__webpack_require__.p = \"\";\n/******/\n/******/ \t// Load entry module and return exports\n/******/ \treturn __webpack_require__(__webpack_require__.s = 5);\n/******/ })\n/************************************************************************/\n/******/ ([\n/* 0 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* unused harmony export addClasses */\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"d\", function() { return removeClasses; });\n/* unused harmony export show */\n/* unused harmony export hide */\n/* unused harmony export offset */\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"e\", function() { return width; });\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"b\", function() { return height; });\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"c\", function() { return outerHeight; });\n/* unused harmony export outerWidth */\n/* unused harmony export position */\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"a\", function() { return css; });\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_0__type__ = __webpack_require__(2);\n\n\nvar addClasses = function addClasses(element, classes) {\n\tclasses = Array.isArray(classes) ? classes : classes.split(' ');\n\tclasses.forEach(function (cls) {\n\t\telement.classList.add(cls);\n\t});\n};\n\nvar removeClasses = function removeClasses(element, classes) {\n\tclasses = Array.isArray(classes) ? classes : classes.split(' ');\n\tclasses.forEach(function (cls) {\n\t\telement.classList.remove(cls);\n\t});\n};\n\nvar show = function show(elements) {\n\telements = Array.isArray(elements) ? elements : [elements];\n\telements.forEach(function (element) {\n\t\telement.style.display = '';\n\t});\n};\n\nvar hide = function hide(elements) {\n\telements = Array.isArray(elements) ? elements : [elements];\n\telements.forEach(function (element) {\n\t\telement.style.display = 'none';\n\t});\n};\n\nvar offset = function offset(element) {\n\tvar rect = element.getBoundingClientRect();\n\treturn {\n\t\ttop: rect.top + document.body.scrollTop,\n\t\tleft: rect.left + document.body.scrollLeft\n\t};\n};\n\n// returns an element's width\nvar width = function width(element) {\n\treturn element.getBoundingClientRect().width || element.offsetWidth;\n};\n// returns an element's height\nvar height = function height(element) {\n\treturn element.getBoundingClientRect().height || element.offsetHeight;\n};\n\nvar outerHeight = function outerHeight(element) {\n\tvar withMargin = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : false;\n\n\tvar height = element.offsetHeight;\n\tif (withMargin) {\n\t\tvar style = window.getComputedStyle(element);\n\t\theight += parseInt(style.marginTop) + parseInt(style.marginBottom);\n\t}\n\treturn height;\n};\n\nvar outerWidth = function outerWidth(element) {\n\tvar withMargin = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : false;\n\n\tvar width = element.offsetWidth;\n\tif (withMargin) {\n\t\tvar style = window.getComputedStyle(element);\n\t\twidth += parseInt(style.marginLeft) + parseInt(style.marginRight);\n\t}\n\treturn width;\n};\n\nvar position = function position(element) {\n\treturn {\n\t\tleft: element.offsetLeft,\n\t\ttop: element.offsetTop\n\t};\n};\n\nvar css = function css(element, obj) {\n\tif (!obj) {\n\t\treturn window.getComputedStyle(element);\n\t}\n\tif (Object(__WEBPACK_IMPORTED_MODULE_0__type__[\"b\" /* isObject */])(obj)) {\n\t\tvar style = '';\n\t\tObject.keys(obj).forEach(function (key) {\n\t\t\tstyle += key + ': ' + obj[key] + ';';\n\t\t});\n\n\t\telement.style.cssText += style;\n\t}\n};\n\n/***/ }),\n/* 1 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony export (immutable) */ __webpack_exports__[\"a\"] = detectSupportsPassive;\nfunction detectSupportsPassive() {\n\tvar supportsPassive = false;\n\n\ttry {\n\t\tvar opts = Object.defineProperty({}, 'passive', {\n\t\t\tget: function get() {\n\t\t\t\tsupportsPassive = true;\n\t\t\t}\n\t\t});\n\n\t\twindow.addEventListener('testPassive', null, opts);\n\t\twindow.removeEventListener('testPassive', null, opts);\n\t} catch (e) {}\n\n\treturn supportsPassive;\n}\n\n/***/ }),\n/* 2 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"a\", function() { return isFunction; });\n/* unused harmony export isNumber */\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"c\", function() { return isString; });\n/* unused harmony export isDate */\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"b\", function() { return isObject; });\n/* unused harmony export isEmptyObject */\n/* unused harmony export isNode */\n/* unused harmony export isVideo */\n/* unused harmony export isHTML5 */\n/* unused harmony export isIFrame */\n/* unused harmony export isYoutube */\n/* unused harmony export isVimeo */\nvar _typeof = typeof Symbol === \"function\" && typeof Symbol.iterator === \"symbol\" ? function (obj) { return typeof obj; } : function (obj) { return obj && typeof Symbol === \"function\" && obj.constructor === Symbol && obj !== Symbol.prototype ? \"symbol\" : typeof obj; };\n\nvar isFunction = function isFunction(unknown) {\n\treturn typeof unknown === 'function';\n};\nvar isNumber = function isNumber(unknown) {\n\treturn typeof unknown === \"number\";\n};\nvar isString = function isString(unknown) {\n\treturn typeof unknown === 'string' || !!unknown && (typeof unknown === 'undefined' ? 'undefined' : _typeof(unknown)) === 'object' && Object.prototype.toString.call(unknown) === '[object String]';\n};\nvar isDate = function isDate(unknown) {\n\treturn (Object.prototype.toString.call(unknown) === '[object Date]' || unknown instanceof Date) && !isNaN(unknown.valueOf());\n};\nvar isObject = function isObject(unknown) {\n\treturn (typeof unknown === 'function' || (typeof unknown === 'undefined' ? 'undefined' : _typeof(unknown)) === 'object' && !!unknown) && !Array.isArray(unknown);\n};\nvar isEmptyObject = function isEmptyObject(unknown) {\n\tfor (var name in unknown) {\n\t\tif (unknown.hasOwnProperty(name)) {\n\t\t\treturn false;\n\t\t}\n\t}\n\treturn true;\n};\n\nvar isNode = function isNode(unknown) {\n\treturn !!(unknown && unknown.nodeType === HTMLElement | SVGElement);\n};\nvar isVideo = function isVideo(unknown) {\n\treturn isYoutube(unknown) || isVimeo(unknown) || isHTML5(unknown);\n};\nvar isHTML5 = function isHTML5(unknown) {\n\treturn isNode(unknown) && unknown.tagName === 'VIDEO';\n};\nvar isIFrame = function isIFrame(unknown) {\n\treturn isNode(unknown) && unknown.tagName === 'IFRAME';\n};\nvar isYoutube = function isYoutube(unknown) {\n\treturn isIFrame(unknown) && !!unknown.src.match(/\\/\\/.*?youtube(-nocookie)?\\.[a-z]+\\/(watch\\?v=[^&\\s]+|embed)|youtu\\.be\\/.*/);\n};\nvar isVimeo = function isVimeo(unknown) {\n\treturn isIFrame(unknown) && !!unknown.src.match(/vimeo\\.com\\/video\\/.*/);\n};\n\n/***/ }),\n/* 3 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _toConsumableArray(arr) { if (Array.isArray(arr)) { for (var i = 0, arr2 = Array(arr.length); i < arr.length; i++) { arr2[i] = arr[i]; } return arr2; } else { return Array.from(arr); } }\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\nvar EventEmitter = function () {\n function EventEmitter() {\n var events = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : [];\n\n _classCallCheck(this, EventEmitter);\n\n this.events = new Map(events);\n }\n\n _createClass(EventEmitter, [{\n key: \"on\",\n value: function on(name, cb) {\n var _this = this;\n\n this.events.set(name, [].concat(_toConsumableArray(this.events.has(name) ? this.events.get(name) : []), [cb]));\n\n return function () {\n return _this.events.set(name, _this.events.get(name).filter(function (fn) {\n return fn !== cb;\n }));\n };\n }\n }, {\n key: \"emit\",\n value: function emit(name) {\n for (var _len = arguments.length, args = Array(_len > 1 ? _len - 1 : 0), _key = 1; _key < _len; _key++) {\n args[_key - 1] = arguments[_key];\n }\n\n return this.events.has(name) && this.events.get(name).map(function (fn) {\n return fn.apply(undefined, args);\n });\n }\n }]);\n\n return EventEmitter;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (EventEmitter);\n\n/***/ }),\n/* 4 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\nvar Coordinate = function () {\n\tfunction Coordinate() {\n\t\tvar x = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : 0;\n\t\tvar y = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : 0;\n\n\t\t_classCallCheck(this, Coordinate);\n\n\t\tthis._x = x;\n\t\tthis._y = y;\n\t}\n\n\t_createClass(Coordinate, [{\n\t\tkey: 'add',\n\t\tvalue: function add(coord) {\n\t\t\treturn new Coordinate(this._x + coord._x, this._y + coord._y);\n\t\t}\n\t}, {\n\t\tkey: 'sub',\n\t\tvalue: function sub(coord) {\n\t\t\treturn new Coordinate(this._x - coord._x, this._y - coord._y);\n\t\t}\n\t}, {\n\t\tkey: 'distance',\n\t\tvalue: function distance(coord) {\n\t\t\tvar deltaX = this._x - coord._x;\n\t\t\tvar deltaY = this._y - coord._y;\n\n\t\t\treturn Math.sqrt(Math.pow(deltaX, 2) + Math.pow(deltaY, 2));\n\t\t}\n\t}, {\n\t\tkey: 'max',\n\t\tvalue: function max(coord) {\n\t\t\tvar x = Math.max(this._x, coord._x);\n\t\t\tvar y = Math.max(this._y, coord._y);\n\n\t\t\treturn new Coordinate(x, y);\n\t\t}\n\t}, {\n\t\tkey: 'equals',\n\t\tvalue: function equals(coord) {\n\t\t\tif (this == coord) {\n\t\t\t\treturn true;\n\t\t\t}\n\t\t\tif (!coord || coord == null) {\n\t\t\t\treturn false;\n\t\t\t}\n\t\t\treturn this._x == coord._x && this._y == coord._y;\n\t\t}\n\t}, {\n\t\tkey: 'inside',\n\t\tvalue: function inside(northwest, southeast) {\n\t\t\tif (this._x >= northwest._x && this._x <= southeast._x && this._y >= northwest._y && this._y <= southeast._y) {\n\n\t\t\t\treturn true;\n\t\t\t}\n\t\t\treturn false;\n\t\t}\n\t}, {\n\t\tkey: 'constrain',\n\t\tvalue: function constrain(min, max) {\n\t\t\tif (min._x > max._x || min._y > max._y) {\n\t\t\t\treturn this;\n\t\t\t}\n\n\t\t\tvar x = this._x,\n\t\t\t y = this._y;\n\n\t\t\tif (min._x !== null) {\n\t\t\t\tx = Math.max(x, min._x);\n\t\t\t}\n\t\t\tif (max._x !== null) {\n\t\t\t\tx = Math.min(x, max._x);\n\t\t\t}\n\t\t\tif (min._y !== null) {\n\t\t\t\ty = Math.max(y, min._y);\n\t\t\t}\n\t\t\tif (max._y !== null) {\n\t\t\t\ty = Math.min(y, max._y);\n\t\t\t}\n\n\t\t\treturn new Coordinate(x, y);\n\t\t}\n\t}, {\n\t\tkey: 'reposition',\n\t\tvalue: function reposition(element) {\n\t\t\telement.style['top'] = this._y + 'px';\n\t\t\telement.style['left'] = this._x + 'px';\n\t\t}\n\t}, {\n\t\tkey: 'toString',\n\t\tvalue: function toString() {\n\t\t\treturn '(' + this._x + ',' + this._y + ')';\n\t\t}\n\t}, {\n\t\tkey: 'x',\n\t\tget: function get() {\n\t\t\treturn this._x;\n\t\t},\n\t\tset: function set() {\n\t\t\tvar value = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : 0;\n\n\t\t\tthis._x = value;\n\t\t\treturn this;\n\t\t}\n\t}, {\n\t\tkey: 'y',\n\t\tget: function get() {\n\t\t\treturn this._y;\n\t\t},\n\t\tset: function set() {\n\t\t\tvar value = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : 0;\n\n\t\t\tthis._y = value;\n\t\t\treturn this;\n\t\t}\n\t}]);\n\n\treturn Coordinate;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (Coordinate);\n\n/***/ }),\n/* 5 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\nObject.defineProperty(__webpack_exports__, \"__esModule\", { value: true });\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_0__utils_index__ = __webpack_require__(6);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_1__utils_css__ = __webpack_require__(0);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_2__utils_type__ = __webpack_require__(2);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_3__utils_eventEmitter__ = __webpack_require__(3);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_4__components_autoplay__ = __webpack_require__(7);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_5__components_breakpoint__ = __webpack_require__(9);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_6__components_infinite__ = __webpack_require__(10);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_7__components_loop__ = __webpack_require__(11);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_8__components_navigation__ = __webpack_require__(13);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_9__components_pagination__ = __webpack_require__(15);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_10__components_swipe__ = __webpack_require__(18);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_11__components_transitioner__ = __webpack_require__(19);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_12__defaultOptions__ = __webpack_require__(22);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_13__templates__ = __webpack_require__(23);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_14__templates_item__ = __webpack_require__(24);\nvar _extends = Object.assign || function (target) { for (var i = 1; i < arguments.length; i++) { var source = arguments[i]; for (var key in source) { if (Object.prototype.hasOwnProperty.call(source, key)) { target[key] = source[key]; } } } return target; };\n\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _defineProperty(obj, key, value) { if (key in obj) { Object.defineProperty(obj, key, { value: value, enumerable: true, configurable: true, writable: true }); } else { obj[key] = value; } return obj; }\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\nfunction _possibleConstructorReturn(self, call) { if (!self) { throw new ReferenceError(\"this hasn't been initialised - super() hasn't been called\"); } return call && (typeof call === \"object\" || typeof call === \"function\") ? call : self; }\n\nfunction _inherits(subClass, superClass) { if (typeof superClass !== \"function\" && superClass !== null) { throw new TypeError(\"Super expression must either be null or a function, not \" + typeof superClass); } subClass.prototype = Object.create(superClass && superClass.prototype, { constructor: { value: subClass, enumerable: false, writable: true, configurable: true } }); if (superClass) Object.setPrototypeOf ? Object.setPrototypeOf(subClass, superClass) : subClass.__proto__ = superClass; }\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nvar bulmaCarousel = function (_EventEmitter) {\n _inherits(bulmaCarousel, _EventEmitter);\n\n function bulmaCarousel(selector) {\n var options = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : {};\n\n _classCallCheck(this, bulmaCarousel);\n\n var _this = _possibleConstructorReturn(this, (bulmaCarousel.__proto__ || Object.getPrototypeOf(bulmaCarousel)).call(this));\n\n _this.element = Object(__WEBPACK_IMPORTED_MODULE_2__utils_type__[\"c\" /* isString */])(selector) ? document.querySelector(selector) : selector;\n // An invalid selector or non-DOM node has been provided.\n if (!_this.element) {\n throw new Error('An invalid selector or non-DOM node has been provided.');\n }\n _this._clickEvents = ['click', 'touch'];\n\n // Use Element dataset values to override options\n var elementConfig = _this.element.dataset ? Object.keys(_this.element.dataset).filter(function (key) {\n return Object.keys(__WEBPACK_IMPORTED_MODULE_12__defaultOptions__[\"a\" /* default */]).includes(key);\n }).reduce(function (obj, key) {\n return _extends({}, obj, _defineProperty({}, key, _this.element.dataset[key]));\n }, {}) : {};\n // Set default options - dataset attributes are master\n _this.options = _extends({}, __WEBPACK_IMPORTED_MODULE_12__defaultOptions__[\"a\" /* default */], options, elementConfig);\n\n _this._id = Object(__WEBPACK_IMPORTED_MODULE_0__utils_index__[\"a\" /* uuid */])('slider');\n\n _this.onShow = _this.onShow.bind(_this);\n\n // Initiate plugin\n _this._init();\n return _this;\n }\n\n /**\n * Initiate all DOM element containing datePicker class\n * @method\n * @return {Array} Array of all datePicker instances\n */\n\n\n _createClass(bulmaCarousel, [{\n key: '_init',\n\n\n /****************************************************\n * *\n * PRIVATE FUNCTIONS *\n * *\n ****************************************************/\n /**\n * Initiate plugin instance\n * @method _init\n * @return {Slider} Current plugin instance\n */\n value: function _init() {\n this._items = Array.from(this.element.children);\n\n // Load plugins\n this._breakpoint = new __WEBPACK_IMPORTED_MODULE_5__components_breakpoint__[\"a\" /* default */](this);\n this._autoplay = new __WEBPACK_IMPORTED_MODULE_4__components_autoplay__[\"a\" /* default */](this);\n this._navigation = new __WEBPACK_IMPORTED_MODULE_8__components_navigation__[\"a\" /* default */](this);\n this._pagination = new __WEBPACK_IMPORTED_MODULE_9__components_pagination__[\"a\" /* default */](this);\n this._infinite = new __WEBPACK_IMPORTED_MODULE_6__components_infinite__[\"a\" /* default */](this);\n this._loop = new __WEBPACK_IMPORTED_MODULE_7__components_loop__[\"a\" /* default */](this);\n this._swipe = new __WEBPACK_IMPORTED_MODULE_10__components_swipe__[\"a\" /* default */](this);\n\n this._build();\n\n if (Object(__WEBPACK_IMPORTED_MODULE_2__utils_type__[\"a\" /* isFunction */])(this.options.onReady)) {\n this.options.onReady(this);\n }\n\n return this;\n }\n\n /**\n * Build Slider HTML component and append it to the DOM\n * @method _build\n */\n\n }, {\n key: '_build',\n value: function _build() {\n var _this2 = this;\n\n // Generate HTML Fragment of template\n this.node = document.createRange().createContextualFragment(Object(__WEBPACK_IMPORTED_MODULE_13__templates__[\"a\" /* default */])(this.id));\n // Save pointers to template parts\n this._ui = {\n wrapper: this.node.firstChild,\n container: this.node.querySelector('.slider-container')\n\n // Add slider to DOM\n };this.element.appendChild(this.node);\n this._ui.wrapper.classList.add('is-loading');\n this._ui.container.style.opacity = 0;\n\n this._transitioner = new __WEBPACK_IMPORTED_MODULE_11__components_transitioner__[\"a\" /* default */](this);\n\n // Wrap all items by slide element\n this._slides = this._items.map(function (item, index) {\n return _this2._createSlide(item, index);\n });\n\n this.reset();\n\n this._bindEvents();\n\n this._ui.container.style.opacity = 1;\n this._ui.wrapper.classList.remove('is-loading');\n }\n\n /**\n * Bind all events\n * @method _bindEvents\n * @return {void}\n */\n\n }, {\n key: '_bindEvents',\n value: function _bindEvents() {\n this.on('show', this.onShow);\n }\n }, {\n key: '_unbindEvents',\n value: function _unbindEvents() {\n this.off('show', this.onShow);\n }\n }, {\n key: '_createSlide',\n value: function _createSlide(item, index) {\n var slide = document.createRange().createContextualFragment(Object(__WEBPACK_IMPORTED_MODULE_14__templates_item__[\"a\" /* default */])()).firstChild;\n slide.dataset.sliderIndex = index;\n slide.appendChild(item);\n return slide;\n }\n\n /**\n * Calculate slider dimensions\n */\n\n }, {\n key: '_setDimensions',\n value: function _setDimensions() {\n var _this3 = this;\n\n if (!this.options.vertical) {\n if (this.options.centerMode) {\n this._ui.wrapper.style.padding = '0px ' + this.options.centerPadding;\n }\n } else {\n this._ui.wrapper.style.height = Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"c\" /* outerHeight */])(this._slides[0]) * this.slidesToShow;\n if (this.options.centerMode) {\n this._ui.wrapper.style.padding = this.options.centerPadding + ' 0px';\n }\n }\n\n this._wrapperWidth = Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"e\" /* width */])(this._ui.wrapper);\n this._wrapperHeight = Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"c\" /* outerHeight */])(this._ui.wrapper);\n\n if (!this.options.vertical) {\n this._slideWidth = Math.ceil(this._wrapperWidth / this.slidesToShow);\n this._containerWidth = Math.ceil(this._slideWidth * this._slides.length);\n this._ui.container.style.width = this._containerWidth + 'px';\n } else {\n this._slideWidth = Math.ceil(this._wrapperWidth);\n this._containerHeight = Math.ceil(Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"c\" /* outerHeight */])(this._slides[0]) * this._slides.length);\n this._ui.container.style.height = this._containerHeight + 'px';\n }\n\n this._slides.forEach(function (slide) {\n slide.style.width = _this3._slideWidth + 'px';\n });\n }\n }, {\n key: '_setHeight',\n value: function _setHeight() {\n if (this.options.effect !== 'translate') {\n this._ui.container.style.height = Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"c\" /* outerHeight */])(this._slides[this.state.index]) + 'px';\n }\n }\n\n // Update slides classes\n\n }, {\n key: '_setClasses',\n value: function _setClasses() {\n var _this4 = this;\n\n this._slides.forEach(function (slide) {\n Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"d\" /* removeClasses */])(slide, 'is-active is-current is-slide-previous is-slide-next');\n if (Math.abs((_this4.state.index - 1) % _this4.state.length) === parseInt(slide.dataset.sliderIndex, 10)) {\n slide.classList.add('is-slide-previous');\n }\n if (Math.abs(_this4.state.index % _this4.state.length) === parseInt(slide.dataset.sliderIndex, 10)) {\n slide.classList.add('is-current');\n }\n if (Math.abs((_this4.state.index + 1) % _this4.state.length) === parseInt(slide.dataset.sliderIndex, 10)) {\n slide.classList.add('is-slide-next');\n }\n });\n }\n\n /****************************************************\n * *\n * GETTERS and SETTERS *\n * *\n ****************************************************/\n\n /**\n * Get id of current datePicker\n */\n\n }, {\n key: 'onShow',\n\n\n /****************************************************\n * *\n * EVENTS FUNCTIONS *\n * *\n ****************************************************/\n value: function onShow(e) {\n this._navigation.refresh();\n this._pagination.refresh();\n this._setClasses();\n }\n\n /****************************************************\n * *\n * PUBLIC FUNCTIONS *\n * *\n ****************************************************/\n\n }, {\n key: 'next',\n value: function next() {\n if (!this.options.loop && !this.options.infinite && this.state.index + this.slidesToScroll > this.state.length - this.slidesToShow && !this.options.centerMode) {\n this.state.next = this.state.index;\n } else {\n this.state.next = this.state.index + this.slidesToScroll;\n }\n this.show();\n }\n }, {\n key: 'previous',\n value: function previous() {\n if (!this.options.loop && !this.options.infinite && this.state.index === 0) {\n this.state.next = this.state.index;\n } else {\n this.state.next = this.state.index - this.slidesToScroll;\n }\n this.show();\n }\n }, {\n key: 'start',\n value: function start() {\n this._autoplay.start();\n }\n }, {\n key: 'pause',\n value: function pause() {\n this._autoplay.pause();\n }\n }, {\n key: 'stop',\n value: function stop() {\n this._autoplay.stop();\n }\n }, {\n key: 'show',\n value: function show(index) {\n var force = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : false;\n\n // If all slides are already visible then return\n if (!this.state.length || this.state.length <= this.slidesToShow) {\n return;\n }\n\n if (typeof index === 'Number') {\n this.state.next = index;\n }\n\n if (this.options.loop) {\n this._loop.apply();\n }\n if (this.options.infinite) {\n this._infinite.apply();\n }\n\n // If new slide is already the current one then return\n if (this.state.index === this.state.next) {\n return;\n }\n\n this.emit('before:show', this.state);\n this._transitioner.apply(force, this._setHeight.bind(this));\n this.emit('after:show', this.state);\n\n this.emit('show', this);\n }\n }, {\n key: 'reset',\n value: function reset() {\n var _this5 = this;\n\n this.state = {\n length: this._items.length,\n index: Math.abs(this.options.initialSlide),\n next: Math.abs(this.options.initialSlide),\n prev: undefined\n };\n\n // Fix options\n if (this.options.loop && this.options.infinite) {\n this.options.loop = false;\n }\n if (this.options.slidesToScroll > this.options.slidesToShow) {\n this.options.slidesToScroll = this.slidesToShow;\n }\n this._breakpoint.init();\n\n if (this.state.index >= this.state.length && this.state.index !== 0) {\n this.state.index = this.state.index - this.slidesToScroll;\n }\n if (this.state.length <= this.slidesToShow) {\n this.state.index = 0;\n }\n\n this._ui.wrapper.appendChild(this._navigation.init().render());\n this._ui.wrapper.appendChild(this._pagination.init().render());\n\n if (this.options.navigationSwipe) {\n this._swipe.bindEvents();\n } else {\n this._swipe._bindEvents();\n }\n\n this._breakpoint.apply();\n // Move all created slides into slider\n this._slides.forEach(function (slide) {\n return _this5._ui.container.appendChild(slide);\n });\n this._transitioner.init().apply(true, this._setHeight.bind(this));\n\n if (this.options.autoplay) {\n this._autoplay.init().start();\n }\n }\n\n /**\n * Destroy Slider\n * @method destroy\n */\n\n }, {\n key: 'destroy',\n value: function destroy() {\n var _this6 = this;\n\n this._unbindEvents();\n this._items.forEach(function (item) {\n _this6.element.appendChild(item);\n });\n this.node.remove();\n }\n }, {\n key: 'id',\n get: function get() {\n return this._id;\n }\n }, {\n key: 'index',\n set: function set(index) {\n this._index = index;\n },\n get: function get() {\n return this._index;\n }\n }, {\n key: 'length',\n set: function set(length) {\n this._length = length;\n },\n get: function get() {\n return this._length;\n }\n }, {\n key: 'slides',\n get: function get() {\n return this._slides;\n },\n set: function set(slides) {\n this._slides = slides;\n }\n }, {\n key: 'slidesToScroll',\n get: function get() {\n return this.options.effect === 'translate' ? this._breakpoint.getSlidesToScroll() : 1;\n }\n }, {\n key: 'slidesToShow',\n get: function get() {\n return this.options.effect === 'translate' ? this._breakpoint.getSlidesToShow() : 1;\n }\n }, {\n key: 'direction',\n get: function get() {\n return this.element.dir.toLowerCase() === 'rtl' || this.element.style.direction === 'rtl' ? 'rtl' : 'ltr';\n }\n }, {\n key: 'wrapper',\n get: function get() {\n return this._ui.wrapper;\n }\n }, {\n key: 'wrapperWidth',\n get: function get() {\n return this._wrapperWidth || 0;\n }\n }, {\n key: 'container',\n get: function get() {\n return this._ui.container;\n }\n }, {\n key: 'containerWidth',\n get: function get() {\n return this._containerWidth || 0;\n }\n }, {\n key: 'slideWidth',\n get: function get() {\n return this._slideWidth || 0;\n }\n }, {\n key: 'transitioner',\n get: function get() {\n return this._transitioner;\n }\n }], [{\n key: 'attach',\n value: function attach() {\n var _this7 = this;\n\n var selector = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : '.slider';\n var options = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : {};\n\n var instances = new Array();\n\n var elements = Object(__WEBPACK_IMPORTED_MODULE_2__utils_type__[\"c\" /* isString */])(selector) ? document.querySelectorAll(selector) : Array.isArray(selector) ? selector : [selector];\n [].forEach.call(elements, function (element) {\n if (typeof element[_this7.constructor.name] === 'undefined') {\n var instance = new bulmaCarousel(element, options);\n element[_this7.constructor.name] = instance;\n instances.push(instance);\n } else {\n instances.push(element[_this7.constructor.name]);\n }\n });\n\n return instances;\n }\n }]);\n\n return bulmaCarousel;\n}(__WEBPACK_IMPORTED_MODULE_3__utils_eventEmitter__[\"a\" /* default */]);\n\n/* harmony default export */ __webpack_exports__[\"default\"] = (bulmaCarousel);\n\n/***/ }),\n/* 6 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"a\", function() { return uuid; });\n/* unused harmony export isRtl */\n/* unused harmony export defer */\n/* unused harmony export getNodeIndex */\n/* unused harmony export camelize */\nfunction _toConsumableArray(arr) { if (Array.isArray(arr)) { for (var i = 0, arr2 = Array(arr.length); i < arr.length; i++) { arr2[i] = arr[i]; } return arr2; } else { return Array.from(arr); } }\n\nvar uuid = function uuid() {\n\tvar prefix = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : '';\n\treturn prefix + ([1e7] + -1e3 + -4e3 + -8e3 + -1e11).replace(/[018]/g, function (c) {\n\t\treturn (c ^ crypto.getRandomValues(new Uint8Array(1))[0] & 15 >> c / 4).toString(16);\n\t});\n};\nvar isRtl = function isRtl() {\n\treturn document.documentElement.getAttribute('dir') === 'rtl';\n};\n\nvar defer = function defer() {\n\tthis.promise = new Promise(function (resolve, reject) {\n\t\tthis.resolve = resolve;\n\t\tthis.reject = reject;\n\t}.bind(this));\n\n\tthis.then = this.promise.then.bind(this.promise);\n\tthis.catch = this.promise.catch.bind(this.promise);\n};\n\nvar getNodeIndex = function getNodeIndex(node) {\n\treturn [].concat(_toConsumableArray(node.parentNode.children)).indexOf(node);\n};\nvar camelize = function camelize(str) {\n\treturn str.replace(/-(\\w)/g, toUpper);\n};\n\n/***/ }),\n/* 7 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_0__utils_eventEmitter__ = __webpack_require__(3);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_1__utils_device__ = __webpack_require__(8);\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\nfunction _possibleConstructorReturn(self, call) { if (!self) { throw new ReferenceError(\"this hasn't been initialised - super() hasn't been called\"); } return call && (typeof call === \"object\" || typeof call === \"function\") ? call : self; }\n\nfunction _inherits(subClass, superClass) { if (typeof superClass !== \"function\" && superClass !== null) { throw new TypeError(\"Super expression must either be null or a function, not \" + typeof superClass); } subClass.prototype = Object.create(superClass && superClass.prototype, { constructor: { value: subClass, enumerable: false, writable: true, configurable: true } }); if (superClass) Object.setPrototypeOf ? Object.setPrototypeOf(subClass, superClass) : subClass.__proto__ = superClass; }\n\n\n\n\nvar onVisibilityChange = Symbol('onVisibilityChange');\nvar onMouseEnter = Symbol('onMouseEnter');\nvar onMouseLeave = Symbol('onMouseLeave');\n\nvar defaultOptions = {\n\tautoplay: false,\n\tautoplaySpeed: 3000\n};\n\nvar Autoplay = function (_EventEmitter) {\n\t_inherits(Autoplay, _EventEmitter);\n\n\tfunction Autoplay(slider) {\n\t\t_classCallCheck(this, Autoplay);\n\n\t\tvar _this = _possibleConstructorReturn(this, (Autoplay.__proto__ || Object.getPrototypeOf(Autoplay)).call(this));\n\n\t\t_this.slider = slider;\n\n\t\t_this.onVisibilityChange = _this.onVisibilityChange.bind(_this);\n\t\t_this.onMouseEnter = _this.onMouseEnter.bind(_this);\n\t\t_this.onMouseLeave = _this.onMouseLeave.bind(_this);\n\t\treturn _this;\n\t}\n\n\t_createClass(Autoplay, [{\n\t\tkey: 'init',\n\t\tvalue: function init() {\n\t\t\tthis._bindEvents();\n\t\t\treturn this;\n\t\t}\n\t}, {\n\t\tkey: '_bindEvents',\n\t\tvalue: function _bindEvents() {\n\t\t\tdocument.addEventListener('visibilitychange', this.onVisibilityChange);\n\t\t\tif (this.slider.options.pauseOnHover) {\n\t\t\t\tthis.slider.container.addEventListener(__WEBPACK_IMPORTED_MODULE_1__utils_device__[\"a\" /* pointerEnter */], this.onMouseEnter);\n\t\t\t\tthis.slider.container.addEventListener(__WEBPACK_IMPORTED_MODULE_1__utils_device__[\"b\" /* pointerLeave */], this.onMouseLeave);\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: '_unbindEvents',\n\t\tvalue: function _unbindEvents() {\n\t\t\tdocument.removeEventListener('visibilitychange', this.onVisibilityChange);\n\t\t\tthis.slider.container.removeEventListener(__WEBPACK_IMPORTED_MODULE_1__utils_device__[\"a\" /* pointerEnter */], this.onMouseEnter);\n\t\t\tthis.slider.container.removeEventListener(__WEBPACK_IMPORTED_MODULE_1__utils_device__[\"b\" /* pointerLeave */], this.onMouseLeave);\n\t\t}\n\t}, {\n\t\tkey: 'start',\n\t\tvalue: function start() {\n\t\t\tvar _this2 = this;\n\n\t\t\tthis.stop();\n\t\t\tif (this.slider.options.autoplay) {\n\t\t\t\tthis.emit('start', this);\n\t\t\t\tthis._interval = setInterval(function () {\n\t\t\t\t\tif (!(_this2._hovering && _this2.slider.options.pauseOnHover)) {\n\t\t\t\t\t\tif (!_this2.slider.options.centerMode && _this2.slider.state.next >= _this2.slider.state.length - _this2.slider.slidesToShow && !_this2.slider.options.loop && !_this2.slider.options.infinite) {\n\t\t\t\t\t\t\t_this2.stop();\n\t\t\t\t\t\t} else {\n\t\t\t\t\t\t\t_this2.slider.next();\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t}, this.slider.options.autoplaySpeed);\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'stop',\n\t\tvalue: function stop() {\n\t\t\tthis._interval = clearInterval(this._interval);\n\t\t\tthis.emit('stop', this);\n\t\t}\n\t}, {\n\t\tkey: 'pause',\n\t\tvalue: function pause() {\n\t\t\tvar _this3 = this;\n\n\t\t\tvar speed = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : 0;\n\n\t\t\tif (this.paused) {\n\t\t\t\treturn;\n\t\t\t}\n\t\t\tif (this.timer) {\n\t\t\t\tthis.stop();\n\t\t\t}\n\t\t\tthis.paused = true;\n\t\t\tif (speed === 0) {\n\t\t\t\tthis.paused = false;\n\t\t\t\tthis.start();\n\t\t\t} else {\n\t\t\t\tthis.slider.on('transition:end', function () {\n\t\t\t\t\tif (!_this3) {\n\t\t\t\t\t\treturn;\n\t\t\t\t\t}\n\t\t\t\t\t_this3.paused = false;\n\t\t\t\t\tif (!_this3.run) {\n\t\t\t\t\t\t_this3.stop();\n\t\t\t\t\t} else {\n\t\t\t\t\t\t_this3.start();\n\t\t\t\t\t}\n\t\t\t\t});\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'onVisibilityChange',\n\t\tvalue: function onVisibilityChange(e) {\n\t\t\tif (document.hidden) {\n\t\t\t\tthis.stop();\n\t\t\t} else {\n\t\t\t\tthis.start();\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'onMouseEnter',\n\t\tvalue: function onMouseEnter(e) {\n\t\t\tthis._hovering = true;\n\t\t\tif (this.slider.options.pauseOnHover) {\n\t\t\t\tthis.pause();\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'onMouseLeave',\n\t\tvalue: function onMouseLeave(e) {\n\t\t\tthis._hovering = false;\n\t\t\tif (this.slider.options.pauseOnHover) {\n\t\t\t\tthis.pause();\n\t\t\t}\n\t\t}\n\t}]);\n\n\treturn Autoplay;\n}(__WEBPACK_IMPORTED_MODULE_0__utils_eventEmitter__[\"a\" /* default */]);\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (Autoplay);\n\n/***/ }),\n/* 8 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* unused harmony export isIE */\n/* unused harmony export isIETouch */\n/* unused harmony export isAndroid */\n/* unused harmony export isiPad */\n/* unused harmony export isiPod */\n/* unused harmony export isiPhone */\n/* unused harmony export isSafari */\n/* unused harmony export isUiWebView */\n/* unused harmony export supportsTouchEvents */\n/* unused harmony export supportsPointerEvents */\n/* unused harmony export supportsTouch */\n/* unused harmony export pointerDown */\n/* unused harmony export pointerMove */\n/* unused harmony export pointerUp */\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"a\", function() { return pointerEnter; });\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"b\", function() { return pointerLeave; });\nvar isIE = window.navigator.pointerEnabled || window.navigator.msPointerEnabled;\nvar isIETouch = window.navigator.msPointerEnabled && window.navigator.msMaxTouchPoints > 1 || window.navigator.pointerEnabled && window.navigator.maxTouchPoints > 1;\nvar isAndroid = navigator.userAgent.match(/(Android);?[\\s\\/]+([\\d.]+)?/);\nvar isiPad = navigator.userAgent.match(/(iPad).*OS\\s([\\d_]+)/);\nvar isiPod = navigator.userAgent.match(/(iPod)(.*OS\\s([\\d_]+))?/);\nvar isiPhone = !navigator.userAgent.match(/(iPad).*OS\\s([\\d_]+)/) && navigator.userAgent.match(/(iPhone\\sOS)\\s([\\d_]+)/);\nvar isSafari = navigator.userAgent.toLowerCase().indexOf('safari') >= 0 && navigator.userAgent.toLowerCase().indexOf('chrome') < 0 && navigator.userAgent.toLowerCase().indexOf('android') < 0;\nvar isUiWebView = /(iPhone|iPod|iPad).*AppleWebKit(?!.*Safari)/i.test(navigator.userAgent);\n\nvar supportsTouchEvents = !!('ontouchstart' in window);\nvar supportsPointerEvents = !!('PointerEvent' in window);\nvar supportsTouch = supportsTouchEvents || window.DocumentTouch && document instanceof DocumentTouch || navigator.maxTouchPoints; // IE >=11\nvar pointerDown = !supportsTouch ? 'mousedown' : 'mousedown ' + (supportsTouchEvents ? 'touchstart' : 'pointerdown');\nvar pointerMove = !supportsTouch ? 'mousemove' : 'mousemove ' + (supportsTouchEvents ? 'touchmove' : 'pointermove');\nvar pointerUp = !supportsTouch ? 'mouseup' : 'mouseup ' + (supportsTouchEvents ? 'touchend' : 'pointerup');\nvar pointerEnter = supportsTouch && supportsPointerEvents ? 'pointerenter' : 'mouseenter';\nvar pointerLeave = supportsTouch && supportsPointerEvents ? 'pointerleave' : 'mouseleave';\n\n/***/ }),\n/* 9 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\nvar onResize = Symbol('onResize');\n\nvar Breakpoints = function () {\n\tfunction Breakpoints(slider) {\n\t\t_classCallCheck(this, Breakpoints);\n\n\t\tthis.slider = slider;\n\t\tthis.options = slider.options;\n\n\t\tthis[onResize] = this[onResize].bind(this);\n\n\t\tthis._bindEvents();\n\t}\n\n\t_createClass(Breakpoints, [{\n\t\tkey: 'init',\n\t\tvalue: function init() {\n\t\t\tthis._defaultBreakpoint = {\n\t\t\t\tslidesToShow: this.options.slidesToShow,\n\t\t\t\tslidesToScroll: this.options.slidesToScroll\n\t\t\t};\n\t\t\tthis.options.breakpoints.sort(function (a, b) {\n\t\t\t\treturn parseInt(a.changePoint, 10) > parseInt(b.changePoint, 10);\n\t\t\t});\n\t\t\tthis._currentBreakpoint = this._getActiveBreakpoint();\n\n\t\t\treturn this;\n\t\t}\n\t}, {\n\t\tkey: 'destroy',\n\t\tvalue: function destroy() {\n\t\t\tthis._unbindEvents();\n\t\t}\n\t}, {\n\t\tkey: '_bindEvents',\n\t\tvalue: function _bindEvents() {\n\t\t\twindow.addEventListener('resize', this[onResize]);\n\t\t\twindow.addEventListener('orientationchange', this[onResize]);\n\t\t}\n\t}, {\n\t\tkey: '_unbindEvents',\n\t\tvalue: function _unbindEvents() {\n\t\t\twindow.removeEventListener('resize', this[onResize]);\n\t\t\twindow.removeEventListener('orientationchange', this[onResize]);\n\t\t}\n\t}, {\n\t\tkey: '_getActiveBreakpoint',\n\t\tvalue: function _getActiveBreakpoint() {\n\t\t\t//Get breakpoint for window width\n\t\t\tvar _iteratorNormalCompletion = true;\n\t\t\tvar _didIteratorError = false;\n\t\t\tvar _iteratorError = undefined;\n\n\t\t\ttry {\n\t\t\t\tfor (var _iterator = this.options.breakpoints[Symbol.iterator](), _step; !(_iteratorNormalCompletion = (_step = _iterator.next()).done); _iteratorNormalCompletion = true) {\n\t\t\t\t\tvar point = _step.value;\n\n\t\t\t\t\tif (point.changePoint >= window.innerWidth) {\n\t\t\t\t\t\treturn point;\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t} catch (err) {\n\t\t\t\t_didIteratorError = true;\n\t\t\t\t_iteratorError = err;\n\t\t\t} finally {\n\t\t\t\ttry {\n\t\t\t\t\tif (!_iteratorNormalCompletion && _iterator.return) {\n\t\t\t\t\t\t_iterator.return();\n\t\t\t\t\t}\n\t\t\t\t} finally {\n\t\t\t\t\tif (_didIteratorError) {\n\t\t\t\t\t\tthrow _iteratorError;\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t}\n\n\t\t\treturn this._defaultBreakpoint;\n\t\t}\n\t}, {\n\t\tkey: 'getSlidesToShow',\n\t\tvalue: function getSlidesToShow() {\n\t\t\treturn this._currentBreakpoint ? this._currentBreakpoint.slidesToShow : this._defaultBreakpoint.slidesToShow;\n\t\t}\n\t}, {\n\t\tkey: 'getSlidesToScroll',\n\t\tvalue: function getSlidesToScroll() {\n\t\t\treturn this._currentBreakpoint ? this._currentBreakpoint.slidesToScroll : this._defaultBreakpoint.slidesToScroll;\n\t\t}\n\t}, {\n\t\tkey: 'apply',\n\t\tvalue: function apply() {\n\t\t\tif (this.slider.state.index >= this.slider.state.length && this.slider.state.index !== 0) {\n\t\t\t\tthis.slider.state.index = this.slider.state.index - this._currentBreakpoint.slidesToScroll;\n\t\t\t}\n\t\t\tif (this.slider.state.length <= this._currentBreakpoint.slidesToShow) {\n\t\t\t\tthis.slider.state.index = 0;\n\t\t\t}\n\n\t\t\tif (this.options.loop) {\n\t\t\t\tthis.slider._loop.init().apply();\n\t\t\t}\n\n\t\t\tif (this.options.infinite) {\n\t\t\t\tthis.slider._infinite.init().apply();\n\t\t\t}\n\n\t\t\tthis.slider._setDimensions();\n\t\t\tthis.slider._transitioner.init().apply(true, this.slider._setHeight.bind(this.slider));\n\t\t\tthis.slider._setClasses();\n\n\t\t\tthis.slider._navigation.refresh();\n\t\t\tthis.slider._pagination.refresh();\n\t\t}\n\t}, {\n\t\tkey: onResize,\n\t\tvalue: function value(e) {\n\t\t\tvar newBreakPoint = this._getActiveBreakpoint();\n\t\t\tif (newBreakPoint.slidesToShow !== this._currentBreakpoint.slidesToShow) {\n\t\t\t\tthis._currentBreakpoint = newBreakPoint;\n\t\t\t\tthis.apply();\n\t\t\t}\n\t\t}\n\t}]);\n\n\treturn Breakpoints;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (Breakpoints);\n\n/***/ }),\n/* 10 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _toConsumableArray(arr) { if (Array.isArray(arr)) { for (var i = 0, arr2 = Array(arr.length); i < arr.length; i++) { arr2[i] = arr[i]; } return arr2; } else { return Array.from(arr); } }\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\nvar Infinite = function () {\n\tfunction Infinite(slider) {\n\t\t_classCallCheck(this, Infinite);\n\n\t\tthis.slider = slider;\n\t}\n\n\t_createClass(Infinite, [{\n\t\tkey: 'init',\n\t\tvalue: function init() {\n\t\t\tif (this.slider.options.infinite && this.slider.options.effect === 'translate') {\n\t\t\t\tif (this.slider.options.centerMode) {\n\t\t\t\t\tthis._infiniteCount = Math.ceil(this.slider.slidesToShow + this.slider.slidesToShow / 2);\n\t\t\t\t} else {\n\t\t\t\t\tthis._infiniteCount = this.slider.slidesToShow;\n\t\t\t\t}\n\n\t\t\t\tvar frontClones = [];\n\t\t\t\tvar slideIndex = 0;\n\t\t\t\tfor (var i = this.slider.state.length; i > this.slider.state.length - 1 - this._infiniteCount; i -= 1) {\n\t\t\t\t\tslideIndex = i - 1;\n\t\t\t\t\tfrontClones.unshift(this._cloneSlide(this.slider.slides[slideIndex], slideIndex - this.slider.state.length));\n\t\t\t\t}\n\n\t\t\t\tvar backClones = [];\n\t\t\t\tfor (var _i = 0; _i < this._infiniteCount + this.slider.state.length; _i += 1) {\n\t\t\t\t\tbackClones.push(this._cloneSlide(this.slider.slides[_i % this.slider.state.length], _i + this.slider.state.length));\n\t\t\t\t}\n\n\t\t\t\tthis.slider.slides = [].concat(frontClones, _toConsumableArray(this.slider.slides), backClones);\n\t\t\t}\n\t\t\treturn this;\n\t\t}\n\t}, {\n\t\tkey: 'apply',\n\t\tvalue: function apply() {}\n\t}, {\n\t\tkey: 'onTransitionEnd',\n\t\tvalue: function onTransitionEnd(e) {\n\t\t\tif (this.slider.options.infinite) {\n\t\t\t\tif (this.slider.state.next >= this.slider.state.length) {\n\t\t\t\t\tthis.slider.state.index = this.slider.state.next = this.slider.state.next - this.slider.state.length;\n\t\t\t\t\tthis.slider.transitioner.apply(true);\n\t\t\t\t} else if (this.slider.state.next < 0) {\n\t\t\t\t\tthis.slider.state.index = this.slider.state.next = this.slider.state.length + this.slider.state.next;\n\t\t\t\t\tthis.slider.transitioner.apply(true);\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: '_cloneSlide',\n\t\tvalue: function _cloneSlide(slide, index) {\n\t\t\tvar newSlide = slide.cloneNode(true);\n\t\t\tnewSlide.dataset.sliderIndex = index;\n\t\t\tnewSlide.dataset.cloned = true;\n\t\t\tvar ids = newSlide.querySelectorAll('[id]') || [];\n\t\t\tids.forEach(function (id) {\n\t\t\t\tid.setAttribute('id', '');\n\t\t\t});\n\t\t\treturn newSlide;\n\t\t}\n\t}]);\n\n\treturn Infinite;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (Infinite);\n\n/***/ }),\n/* 11 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_0__utils_dom__ = __webpack_require__(12);\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\n\n\nvar Loop = function () {\n\tfunction Loop(slider) {\n\t\t_classCallCheck(this, Loop);\n\n\t\tthis.slider = slider;\n\t}\n\n\t_createClass(Loop, [{\n\t\tkey: \"init\",\n\t\tvalue: function init() {\n\t\t\treturn this;\n\t\t}\n\t}, {\n\t\tkey: \"apply\",\n\t\tvalue: function apply() {\n\t\t\tif (this.slider.options.loop) {\n\t\t\t\tif (this.slider.state.next > 0) {\n\t\t\t\t\tif (this.slider.state.next < this.slider.state.length) {\n\t\t\t\t\t\tif (this.slider.state.next > this.slider.state.length - this.slider.slidesToShow && Object(__WEBPACK_IMPORTED_MODULE_0__utils_dom__[\"a\" /* isInViewport */])(this.slider._slides[this.slider.state.length - 1], this.slider.wrapper)) {\n\t\t\t\t\t\t\tthis.slider.state.next = 0;\n\t\t\t\t\t\t} else {\n\t\t\t\t\t\t\tthis.slider.state.next = Math.min(Math.max(this.slider.state.next, 0), this.slider.state.length - this.slider.slidesToShow);\n\t\t\t\t\t\t}\n\t\t\t\t\t} else {\n\t\t\t\t\t\tthis.slider.state.next = 0;\n\t\t\t\t\t}\n\t\t\t\t} else {\n\t\t\t\t\tif (this.slider.state.next <= 0 - this.slider.slidesToScroll) {\n\t\t\t\t\t\tthis.slider.state.next = this.slider.state.length - this.slider.slidesToShow;\n\t\t\t\t\t} else {\n\t\t\t\t\t\tthis.slider.state.next = 0;\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}]);\n\n\treturn Loop;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (Loop);\n\n/***/ }),\n/* 12 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"a\", function() { return isInViewport; });\nvar isInViewport = function isInViewport(element, html) {\n\tvar rect = element.getBoundingClientRect();\n\thtml = html || document.documentElement;\n\treturn rect.top >= 0 && rect.left >= 0 && rect.bottom <= (window.innerHeight || html.clientHeight) && rect.right <= (window.innerWidth || html.clientWidth);\n};\n\n/***/ }),\n/* 13 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_0__templates_navigation__ = __webpack_require__(14);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_1__utils_detect_supportsPassive__ = __webpack_require__(1);\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\n\n\n\nvar Navigation = function () {\n\tfunction Navigation(slider) {\n\t\t_classCallCheck(this, Navigation);\n\n\t\tthis.slider = slider;\n\n\t\tthis._clickEvents = ['click', 'touch'];\n\t\tthis._supportsPassive = Object(__WEBPACK_IMPORTED_MODULE_1__utils_detect_supportsPassive__[\"a\" /* default */])();\n\n\t\tthis.onPreviousClick = this.onPreviousClick.bind(this);\n\t\tthis.onNextClick = this.onNextClick.bind(this);\n\t\tthis.onKeyUp = this.onKeyUp.bind(this);\n\t}\n\n\t_createClass(Navigation, [{\n\t\tkey: 'init',\n\t\tvalue: function init() {\n\t\t\tthis.node = document.createRange().createContextualFragment(Object(__WEBPACK_IMPORTED_MODULE_0__templates_navigation__[\"a\" /* default */])(this.slider.options.icons));\n\t\t\tthis._ui = {\n\t\t\t\tprevious: this.node.querySelector('.slider-navigation-previous'),\n\t\t\t\tnext: this.node.querySelector('.slider-navigation-next')\n\t\t\t};\n\n\t\t\tthis._unbindEvents();\n\t\t\tthis._bindEvents();\n\n\t\t\tthis.refresh();\n\n\t\t\treturn this;\n\t\t}\n\t}, {\n\t\tkey: 'destroy',\n\t\tvalue: function destroy() {\n\t\t\tthis._unbindEvents();\n\t\t}\n\t}, {\n\t\tkey: '_bindEvents',\n\t\tvalue: function _bindEvents() {\n\t\t\tvar _this = this;\n\n\t\t\tthis.slider.wrapper.addEventListener('keyup', this.onKeyUp);\n\t\t\tthis._clickEvents.forEach(function (clickEvent) {\n\t\t\t\t_this._ui.previous.addEventListener(clickEvent, _this.onPreviousClick);\n\t\t\t\t_this._ui.next.addEventListener(clickEvent, _this.onNextClick);\n\t\t\t});\n\t\t}\n\t}, {\n\t\tkey: '_unbindEvents',\n\t\tvalue: function _unbindEvents() {\n\t\t\tvar _this2 = this;\n\n\t\t\tthis.slider.wrapper.removeEventListener('keyup', this.onKeyUp);\n\t\t\tthis._clickEvents.forEach(function (clickEvent) {\n\t\t\t\t_this2._ui.previous.removeEventListener(clickEvent, _this2.onPreviousClick);\n\t\t\t\t_this2._ui.next.removeEventListener(clickEvent, _this2.onNextClick);\n\t\t\t});\n\t\t}\n\t}, {\n\t\tkey: 'onNextClick',\n\t\tvalue: function onNextClick(e) {\n\t\t\tif (!this._supportsPassive) {\n\t\t\t\te.preventDefault();\n\t\t\t}\n\n\t\t\tif (this.slider.options.navigation) {\n\t\t\t\tthis.slider.next();\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'onPreviousClick',\n\t\tvalue: function onPreviousClick(e) {\n\t\t\tif (!this._supportsPassive) {\n\t\t\t\te.preventDefault();\n\t\t\t}\n\n\t\t\tif (this.slider.options.navigation) {\n\t\t\t\tthis.slider.previous();\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'onKeyUp',\n\t\tvalue: function onKeyUp(e) {\n\t\t\tif (this.slider.options.keyNavigation) {\n\t\t\t\tif (e.key === 'ArrowRight' || e.key === 'Right') {\n\t\t\t\t\tthis.slider.next();\n\t\t\t\t} else if (e.key === 'ArrowLeft' || e.key === 'Left') {\n\t\t\t\t\tthis.slider.previous();\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'refresh',\n\t\tvalue: function refresh() {\n\t\t\t// let centerOffset = Math.floor(this.options.slidesToShow / 2);\n\t\t\tif (!this.slider.options.loop && !this.slider.options.infinite) {\n\t\t\t\tif (this.slider.options.navigation && this.slider.state.length > this.slider.slidesToShow) {\n\t\t\t\t\tthis._ui.previous.classList.remove('is-hidden');\n\t\t\t\t\tthis._ui.next.classList.remove('is-hidden');\n\t\t\t\t\tif (this.slider.state.next === 0) {\n\t\t\t\t\t\tthis._ui.previous.classList.add('is-hidden');\n\t\t\t\t\t\tthis._ui.next.classList.remove('is-hidden');\n\t\t\t\t\t} else if (this.slider.state.next >= this.slider.state.length - this.slider.slidesToShow && !this.slider.options.centerMode) {\n\t\t\t\t\t\tthis._ui.previous.classList.remove('is-hidden');\n\t\t\t\t\t\tthis._ui.next.classList.add('is-hidden');\n\t\t\t\t\t} else if (this.slider.state.next >= this.slider.state.length - 1 && this.slider.options.centerMode) {\n\t\t\t\t\t\tthis._ui.previous.classList.remove('is-hidden');\n\t\t\t\t\t\tthis._ui.next.classList.add('is-hidden');\n\t\t\t\t\t}\n\t\t\t\t} else {\n\t\t\t\t\tthis._ui.previous.classList.add('is-hidden');\n\t\t\t\t\tthis._ui.next.classList.add('is-hidden');\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'render',\n\t\tvalue: function render() {\n\t\t\treturn this.node;\n\t\t}\n\t}]);\n\n\treturn Navigation;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (Navigation);\n\n/***/ }),\n/* 14 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony default export */ __webpack_exports__[\"a\"] = (function (icons) {\n\treturn \"
\" + icons.previous + \"
\\n
\" + icons.next + \"
\";\n});\n\n/***/ }),\n/* 15 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_0__templates_pagination__ = __webpack_require__(16);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_1__templates_pagination_page__ = __webpack_require__(17);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_2__utils_detect_supportsPassive__ = __webpack_require__(1);\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\n\n\n\n\nvar Pagination = function () {\n\tfunction Pagination(slider) {\n\t\t_classCallCheck(this, Pagination);\n\n\t\tthis.slider = slider;\n\n\t\tthis._clickEvents = ['click', 'touch'];\n\t\tthis._supportsPassive = Object(__WEBPACK_IMPORTED_MODULE_2__utils_detect_supportsPassive__[\"a\" /* default */])();\n\n\t\tthis.onPageClick = this.onPageClick.bind(this);\n\t\tthis.onResize = this.onResize.bind(this);\n\t}\n\n\t_createClass(Pagination, [{\n\t\tkey: 'init',\n\t\tvalue: function init() {\n\t\t\tthis._pages = [];\n\t\t\tthis.node = document.createRange().createContextualFragment(Object(__WEBPACK_IMPORTED_MODULE_0__templates_pagination__[\"a\" /* default */])());\n\t\t\tthis._ui = {\n\t\t\t\tcontainer: this.node.firstChild\n\t\t\t};\n\n\t\t\tthis._count = Math.ceil((this.slider.state.length - this.slider.slidesToShow) / this.slider.slidesToScroll);\n\n\t\t\tthis._draw();\n\t\t\tthis.refresh();\n\n\t\t\treturn this;\n\t\t}\n\t}, {\n\t\tkey: 'destroy',\n\t\tvalue: function destroy() {\n\t\t\tthis._unbindEvents();\n\t\t}\n\t}, {\n\t\tkey: '_bindEvents',\n\t\tvalue: function _bindEvents() {\n\t\t\tvar _this = this;\n\n\t\t\twindow.addEventListener('resize', this.onResize);\n\t\t\twindow.addEventListener('orientationchange', this.onResize);\n\n\t\t\tthis._clickEvents.forEach(function (clickEvent) {\n\t\t\t\t_this._pages.forEach(function (page) {\n\t\t\t\t\treturn page.addEventListener(clickEvent, _this.onPageClick);\n\t\t\t\t});\n\t\t\t});\n\t\t}\n\t}, {\n\t\tkey: '_unbindEvents',\n\t\tvalue: function _unbindEvents() {\n\t\t\tvar _this2 = this;\n\n\t\t\twindow.removeEventListener('resize', this.onResize);\n\t\t\twindow.removeEventListener('orientationchange', this.onResize);\n\n\t\t\tthis._clickEvents.forEach(function (clickEvent) {\n\t\t\t\t_this2._pages.forEach(function (page) {\n\t\t\t\t\treturn page.removeEventListener(clickEvent, _this2.onPageClick);\n\t\t\t\t});\n\t\t\t});\n\t\t}\n\t}, {\n\t\tkey: '_draw',\n\t\tvalue: function _draw() {\n\t\t\tthis._ui.container.innerHTML = '';\n\t\t\tif (this.slider.options.pagination && this.slider.state.length > this.slider.slidesToShow) {\n\t\t\t\tfor (var i = 0; i <= this._count; i++) {\n\t\t\t\t\tvar newPageNode = document.createRange().createContextualFragment(Object(__WEBPACK_IMPORTED_MODULE_1__templates_pagination_page__[\"a\" /* default */])()).firstChild;\n\t\t\t\t\tnewPageNode.dataset.index = i * this.slider.slidesToScroll;\n\t\t\t\t\tthis._pages.push(newPageNode);\n\t\t\t\t\tthis._ui.container.appendChild(newPageNode);\n\t\t\t\t}\n\t\t\t\tthis._bindEvents();\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'onPageClick',\n\t\tvalue: function onPageClick(e) {\n\t\t\tif (!this._supportsPassive) {\n\t\t\t\te.preventDefault();\n\t\t\t}\n\n\t\t\tthis.slider.state.next = e.currentTarget.dataset.index;\n\t\t\tthis.slider.show();\n\t\t}\n\t}, {\n\t\tkey: 'onResize',\n\t\tvalue: function onResize() {\n\t\t\tthis._draw();\n\t\t}\n\t}, {\n\t\tkey: 'refresh',\n\t\tvalue: function refresh() {\n\t\t\tvar _this3 = this;\n\n\t\t\tvar newCount = void 0;\n\n\t\t\tif (this.slider.options.infinite) {\n\t\t\t\tnewCount = Math.ceil(this.slider.state.length - 1 / this.slider.slidesToScroll);\n\t\t\t} else {\n\t\t\t\tnewCount = Math.ceil((this.slider.state.length - this.slider.slidesToShow) / this.slider.slidesToScroll);\n\t\t\t}\n\t\t\tif (newCount !== this._count) {\n\t\t\t\tthis._count = newCount;\n\t\t\t\tthis._draw();\n\t\t\t}\n\n\t\t\tthis._pages.forEach(function (page) {\n\t\t\t\tpage.classList.remove('is-active');\n\t\t\t\tif (parseInt(page.dataset.index, 10) === _this3.slider.state.next % _this3.slider.state.length) {\n\t\t\t\t\tpage.classList.add('is-active');\n\t\t\t\t}\n\t\t\t});\n\t\t}\n\t}, {\n\t\tkey: 'render',\n\t\tvalue: function render() {\n\t\t\treturn this.node;\n\t\t}\n\t}]);\n\n\treturn Pagination;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (Pagination);\n\n/***/ }),\n/* 16 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony default export */ __webpack_exports__[\"a\"] = (function () {\n\treturn \"
\";\n});\n\n/***/ }),\n/* 17 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony default export */ __webpack_exports__[\"a\"] = (function () {\n return \"
\";\n});\n\n/***/ }),\n/* 18 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_0__utils_coordinate__ = __webpack_require__(4);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_1__utils_detect_supportsPassive__ = __webpack_require__(1);\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\n\n\n\nvar Swipe = function () {\n\tfunction Swipe(slider) {\n\t\t_classCallCheck(this, Swipe);\n\n\t\tthis.slider = slider;\n\n\t\tthis._supportsPassive = Object(__WEBPACK_IMPORTED_MODULE_1__utils_detect_supportsPassive__[\"a\" /* default */])();\n\n\t\tthis.onStartDrag = this.onStartDrag.bind(this);\n\t\tthis.onMoveDrag = this.onMoveDrag.bind(this);\n\t\tthis.onStopDrag = this.onStopDrag.bind(this);\n\n\t\tthis._init();\n\t}\n\n\t_createClass(Swipe, [{\n\t\tkey: '_init',\n\t\tvalue: function _init() {}\n\t}, {\n\t\tkey: 'bindEvents',\n\t\tvalue: function bindEvents() {\n\t\t\tvar _this = this;\n\n\t\t\tthis.slider.container.addEventListener('dragstart', function (e) {\n\t\t\t\tif (!_this._supportsPassive) {\n\t\t\t\t\te.preventDefault();\n\t\t\t\t}\n\t\t\t});\n\t\t\tthis.slider.container.addEventListener('mousedown', this.onStartDrag);\n\t\t\tthis.slider.container.addEventListener('touchstart', this.onStartDrag);\n\n\t\t\twindow.addEventListener('mousemove', this.onMoveDrag);\n\t\t\twindow.addEventListener('touchmove', this.onMoveDrag);\n\n\t\t\twindow.addEventListener('mouseup', this.onStopDrag);\n\t\t\twindow.addEventListener('touchend', this.onStopDrag);\n\t\t\twindow.addEventListener('touchcancel', this.onStopDrag);\n\t\t}\n\t}, {\n\t\tkey: 'unbindEvents',\n\t\tvalue: function unbindEvents() {\n\t\t\tvar _this2 = this;\n\n\t\t\tthis.slider.container.removeEventListener('dragstart', function (e) {\n\t\t\t\tif (!_this2._supportsPassive) {\n\t\t\t\t\te.preventDefault();\n\t\t\t\t}\n\t\t\t});\n\t\t\tthis.slider.container.removeEventListener('mousedown', this.onStartDrag);\n\t\t\tthis.slider.container.removeEventListener('touchstart', this.onStartDrag);\n\n\t\t\twindow.removeEventListener('mousemove', this.onMoveDrag);\n\t\t\twindow.removeEventListener('touchmove', this.onMoveDrag);\n\n\t\t\twindow.removeEventListener('mouseup', this.onStopDrag);\n\t\t\twindow.removeEventListener('mouseup', this.onStopDrag);\n\t\t\twindow.removeEventListener('touchcancel', this.onStopDrag);\n\t\t}\n\n\t\t/**\n * @param {MouseEvent|TouchEvent}\n */\n\n\t}, {\n\t\tkey: 'onStartDrag',\n\t\tvalue: function onStartDrag(e) {\n\t\t\tif (e.touches) {\n\t\t\t\tif (e.touches.length > 1) {\n\t\t\t\t\treturn;\n\t\t\t\t} else {\n\t\t\t\t\te = e.touches[0];\n\t\t\t\t}\n\t\t\t}\n\n\t\t\tthis._origin = new __WEBPACK_IMPORTED_MODULE_0__utils_coordinate__[\"a\" /* default */](e.screenX, e.screenY);\n\t\t\tthis.width = this.slider.wrapperWidth;\n\t\t\tthis.slider.transitioner.disable();\n\t\t}\n\n\t\t/**\n * @param {MouseEvent|TouchEvent}\n */\n\n\t}, {\n\t\tkey: 'onMoveDrag',\n\t\tvalue: function onMoveDrag(e) {\n\t\t\tif (this._origin) {\n\t\t\t\tvar point = e.touches ? e.touches[0] : e;\n\t\t\t\tthis._lastTranslate = new __WEBPACK_IMPORTED_MODULE_0__utils_coordinate__[\"a\" /* default */](point.screenX - this._origin.x, point.screenY - this._origin.y);\n\t\t\t\tif (e.touches) {\n\t\t\t\t\tif (Math.abs(this._lastTranslate.x) > Math.abs(this._lastTranslate.y)) {\n\t\t\t\t\t\tif (!this._supportsPassive) {\n\t\t\t\t\t\t\te.preventDefault();\n\t\t\t\t\t\t}\n\t\t\t\t\t\te.stopPropagation();\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\n\t\t/**\n * @param {MouseEvent|TouchEvent}\n */\n\n\t}, {\n\t\tkey: 'onStopDrag',\n\t\tvalue: function onStopDrag(e) {\n\t\t\tif (this._origin && this._lastTranslate) {\n\t\t\t\tif (Math.abs(this._lastTranslate.x) > 0.2 * this.width) {\n\t\t\t\t\tif (this._lastTranslate.x < 0) {\n\t\t\t\t\t\tthis.slider.next();\n\t\t\t\t\t} else {\n\t\t\t\t\t\tthis.slider.previous();\n\t\t\t\t\t}\n\t\t\t\t} else {\n\t\t\t\t\tthis.slider.show(true);\n\t\t\t\t}\n\t\t\t}\n\t\t\tthis._origin = null;\n\t\t\tthis._lastTranslate = null;\n\t\t}\n\t}]);\n\n\treturn Swipe;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (Swipe);\n\n/***/ }),\n/* 19 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_0__transitions_fade__ = __webpack_require__(20);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_1__transitions_translate__ = __webpack_require__(21);\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\n\n\n\nvar Transitioner = function () {\n\tfunction Transitioner(slider) {\n\t\t_classCallCheck(this, Transitioner);\n\n\t\tthis.slider = slider;\n\t\tthis.options = slider.options;\n\n\t\tthis._animating = false;\n\t\tthis._animation = undefined;\n\n\t\tthis._translate = new __WEBPACK_IMPORTED_MODULE_1__transitions_translate__[\"a\" /* default */](this, slider, slider.options);\n\t\tthis._fade = new __WEBPACK_IMPORTED_MODULE_0__transitions_fade__[\"a\" /* default */](this, slider, slider.options);\n\t}\n\n\t_createClass(Transitioner, [{\n\t\tkey: 'init',\n\t\tvalue: function init() {\n\t\t\tthis._fade.init();\n\t\t\tthis._translate.init();\n\t\t\treturn this;\n\t\t}\n\t}, {\n\t\tkey: 'isAnimating',\n\t\tvalue: function isAnimating() {\n\t\t\treturn this._animating;\n\t\t}\n\t}, {\n\t\tkey: 'enable',\n\t\tvalue: function enable() {\n\t\t\tthis._animation && this._animation.enable();\n\t\t}\n\t}, {\n\t\tkey: 'disable',\n\t\tvalue: function disable() {\n\t\t\tthis._animation && this._animation.disable();\n\t\t}\n\t}, {\n\t\tkey: 'apply',\n\t\tvalue: function apply(force, callback) {\n\t\t\t// If we don't force refresh and animation in progress then return\n\t\t\tif (this._animating && !force) {\n\t\t\t\treturn;\n\t\t\t}\n\n\t\t\tswitch (this.options.effect) {\n\t\t\t\tcase 'fade':\n\t\t\t\t\tthis._animation = this._fade;\n\t\t\t\t\tbreak;\n\t\t\t\tcase 'translate':\n\t\t\t\tdefault:\n\t\t\t\t\tthis._animation = this._translate;\n\t\t\t\t\tbreak;\n\t\t\t}\n\n\t\t\tthis._animationCallback = callback;\n\n\t\t\tif (force) {\n\t\t\t\tthis._animation && this._animation.disable();\n\t\t\t} else {\n\t\t\t\tthis._animation && this._animation.enable();\n\t\t\t\tthis._animating = true;\n\t\t\t}\n\n\t\t\tthis._animation && this._animation.apply();\n\n\t\t\tif (force) {\n\t\t\t\tthis.end();\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'end',\n\t\tvalue: function end() {\n\t\t\tthis._animating = false;\n\t\t\tthis._animation = undefined;\n\t\t\tthis.slider.state.index = this.slider.state.next;\n\t\t\tif (this._animationCallback) {\n\t\t\t\tthis._animationCallback();\n\t\t\t}\n\t\t}\n\t}]);\n\n\treturn Transitioner;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (Transitioner);\n\n/***/ }),\n/* 20 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_0__utils_css__ = __webpack_require__(0);\nvar _extends = Object.assign || function (target) { for (var i = 1; i < arguments.length; i++) { var source = arguments[i]; for (var key in source) { if (Object.prototype.hasOwnProperty.call(source, key)) { target[key] = source[key]; } } } return target; };\n\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\n\n\nvar Fade = function () {\n\tfunction Fade(transitioner, slider) {\n\t\tvar options = arguments.length > 2 && arguments[2] !== undefined ? arguments[2] : {};\n\n\t\t_classCallCheck(this, Fade);\n\n\t\tthis.transitioner = transitioner;\n\t\tthis.slider = slider;\n\t\tthis.options = _extends({}, options);\n\t}\n\n\t_createClass(Fade, [{\n\t\tkey: 'init',\n\t\tvalue: function init() {\n\t\t\tvar _this = this;\n\n\t\t\tif (this.options.effect === 'fade') {\n\t\t\t\tthis.slider.slides.forEach(function (slide, index) {\n\t\t\t\t\tObject(__WEBPACK_IMPORTED_MODULE_0__utils_css__[\"a\" /* css */])(slide, {\n\t\t\t\t\t\tposition: 'absolute',\n\t\t\t\t\t\tleft: 0,\n\t\t\t\t\t\ttop: 0,\n\t\t\t\t\t\tbottom: 0,\n\t\t\t\t\t\t'z-index': slide.dataset.sliderIndex == _this.slider.state.index ? 0 : -2,\n\t\t\t\t\t\topacity: slide.dataset.sliderIndex == _this.slider.state.index ? 1 : 0\n\t\t\t\t\t});\n\t\t\t\t});\n\t\t\t}\n\t\t\treturn this;\n\t\t}\n\t}, {\n\t\tkey: 'enable',\n\t\tvalue: function enable() {\n\t\t\tvar _this2 = this;\n\n\t\t\tthis._oldSlide = this.slider.slides.filter(function (slide) {\n\t\t\t\treturn slide.dataset.sliderIndex == _this2.slider.state.index;\n\t\t\t})[0];\n\t\t\tthis._newSlide = this.slider.slides.filter(function (slide) {\n\t\t\t\treturn slide.dataset.sliderIndex == _this2.slider.state.next;\n\t\t\t})[0];\n\t\t\tif (this._newSlide) {\n\t\t\t\tthis._newSlide.addEventListener('transitionend', this.onTransitionEnd.bind(this));\n\t\t\t\tthis._newSlide.style.transition = this.options.duration + 'ms ' + this.options.timing;\n\t\t\t\tif (this._oldSlide) {\n\t\t\t\t\tthis._oldSlide.addEventListener('transitionend', this.onTransitionEnd.bind(this));\n\t\t\t\t\tthis._oldSlide.style.transition = this.options.duration + 'ms ' + this.options.timing;\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'disable',\n\t\tvalue: function disable() {\n\t\t\tvar _this3 = this;\n\n\t\t\tthis._oldSlide = this.slider.slides.filter(function (slide) {\n\t\t\t\treturn slide.dataset.sliderIndex == _this3.slider.state.index;\n\t\t\t})[0];\n\t\t\tthis._newSlide = this.slider.slides.filter(function (slide) {\n\t\t\t\treturn slide.dataset.sliderIndex == _this3.slider.state.next;\n\t\t\t})[0];\n\t\t\tif (this._newSlide) {\n\t\t\t\tthis._newSlide.removeEventListener('transitionend', this.onTransitionEnd.bind(this));\n\t\t\t\tthis._newSlide.style.transition = 'none';\n\t\t\t\tif (this._oldSlide) {\n\t\t\t\t\tthis._oldSlide.removeEventListener('transitionend', this.onTransitionEnd.bind(this));\n\t\t\t\t\tthis._oldSlide.style.transition = 'none';\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'apply',\n\t\tvalue: function apply(force) {\n\t\t\tvar _this4 = this;\n\n\t\t\tthis._oldSlide = this.slider.slides.filter(function (slide) {\n\t\t\t\treturn slide.dataset.sliderIndex == _this4.slider.state.index;\n\t\t\t})[0];\n\t\t\tthis._newSlide = this.slider.slides.filter(function (slide) {\n\t\t\t\treturn slide.dataset.sliderIndex == _this4.slider.state.next;\n\t\t\t})[0];\n\n\t\t\tif (this._oldSlide && this._newSlide) {\n\t\t\t\tObject(__WEBPACK_IMPORTED_MODULE_0__utils_css__[\"a\" /* css */])(this._oldSlide, {\n\t\t\t\t\topacity: 0\n\t\t\t\t});\n\t\t\t\tObject(__WEBPACK_IMPORTED_MODULE_0__utils_css__[\"a\" /* css */])(this._newSlide, {\n\t\t\t\t\topacity: 1,\n\t\t\t\t\t'z-index': force ? 0 : -1\n\t\t\t\t});\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'onTransitionEnd',\n\t\tvalue: function onTransitionEnd(e) {\n\t\t\tif (this.options.effect === 'fade') {\n\t\t\t\tif (this.transitioner.isAnimating() && e.target == this._newSlide) {\n\t\t\t\t\tif (this._newSlide) {\n\t\t\t\t\t\tObject(__WEBPACK_IMPORTED_MODULE_0__utils_css__[\"a\" /* css */])(this._newSlide, {\n\t\t\t\t\t\t\t'z-index': 0\n\t\t\t\t\t\t});\n\t\t\t\t\t\tthis._newSlide.removeEventListener('transitionend', this.onTransitionEnd.bind(this));\n\t\t\t\t\t}\n\t\t\t\t\tif (this._oldSlide) {\n\t\t\t\t\t\tObject(__WEBPACK_IMPORTED_MODULE_0__utils_css__[\"a\" /* css */])(this._oldSlide, {\n\t\t\t\t\t\t\t'z-index': -2\n\t\t\t\t\t\t});\n\t\t\t\t\t\tthis._oldSlide.removeEventListener('transitionend', this.onTransitionEnd.bind(this));\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t\tthis.transitioner.end();\n\t\t\t}\n\t\t}\n\t}]);\n\n\treturn Fade;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (Fade);\n\n/***/ }),\n/* 21 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_0__utils_coordinate__ = __webpack_require__(4);\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_1__utils_css__ = __webpack_require__(0);\nvar _extends = Object.assign || function (target) { for (var i = 1; i < arguments.length; i++) { var source = arguments[i]; for (var key in source) { if (Object.prototype.hasOwnProperty.call(source, key)) { target[key] = source[key]; } } } return target; };\n\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\n\n\n\nvar Translate = function () {\n\tfunction Translate(transitioner, slider) {\n\t\tvar options = arguments.length > 2 && arguments[2] !== undefined ? arguments[2] : {};\n\n\t\t_classCallCheck(this, Translate);\n\n\t\tthis.transitioner = transitioner;\n\t\tthis.slider = slider;\n\t\tthis.options = _extends({}, options);\n\n\t\tthis.onTransitionEnd = this.onTransitionEnd.bind(this);\n\t}\n\n\t_createClass(Translate, [{\n\t\tkey: 'init',\n\t\tvalue: function init() {\n\t\t\tthis._position = new __WEBPACK_IMPORTED_MODULE_0__utils_coordinate__[\"a\" /* default */](this.slider.container.offsetLeft, this.slider.container.offsetTop);\n\t\t\tthis._bindEvents();\n\t\t\treturn this;\n\t\t}\n\t}, {\n\t\tkey: 'destroy',\n\t\tvalue: function destroy() {\n\t\t\tthis._unbindEvents();\n\t\t}\n\t}, {\n\t\tkey: '_bindEvents',\n\t\tvalue: function _bindEvents() {\n\t\t\tthis.slider.container.addEventListener('transitionend', this.onTransitionEnd);\n\t\t}\n\t}, {\n\t\tkey: '_unbindEvents',\n\t\tvalue: function _unbindEvents() {\n\t\t\tthis.slider.container.removeEventListener('transitionend', this.onTransitionEnd);\n\t\t}\n\t}, {\n\t\tkey: 'enable',\n\t\tvalue: function enable() {\n\t\t\tthis.slider.container.style.transition = this.options.duration + 'ms ' + this.options.timing;\n\t\t}\n\t}, {\n\t\tkey: 'disable',\n\t\tvalue: function disable() {\n\t\t\tthis.slider.container.style.transition = 'none';\n\t\t}\n\t}, {\n\t\tkey: 'apply',\n\t\tvalue: function apply() {\n\t\t\tvar _this = this;\n\n\t\t\tvar maxOffset = void 0;\n\t\t\tif (this.options.effect === 'translate') {\n\t\t\t\tvar slide = this.slider.slides.filter(function (slide) {\n\t\t\t\t\treturn slide.dataset.sliderIndex == _this.slider.state.next;\n\t\t\t\t})[0];\n\t\t\t\tvar slideOffset = new __WEBPACK_IMPORTED_MODULE_0__utils_coordinate__[\"a\" /* default */](slide.offsetLeft, slide.offsetTop);\n\t\t\t\tif (this.options.centerMode) {\n\t\t\t\t\tmaxOffset = new __WEBPACK_IMPORTED_MODULE_0__utils_coordinate__[\"a\" /* default */](Math.round(Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"e\" /* width */])(this.slider.container)), Math.round(Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"b\" /* height */])(this.slider.container)));\n\t\t\t\t} else {\n\t\t\t\t\tmaxOffset = new __WEBPACK_IMPORTED_MODULE_0__utils_coordinate__[\"a\" /* default */](Math.round(Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"e\" /* width */])(this.slider.container) - Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"e\" /* width */])(this.slider.wrapper)), Math.round(Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"b\" /* height */])(this.slider.container) - Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"b\" /* height */])(this.slider.wrapper)));\n\t\t\t\t}\n\t\t\t\tvar nextOffset = new __WEBPACK_IMPORTED_MODULE_0__utils_coordinate__[\"a\" /* default */](Math.min(Math.max(slideOffset.x * -1, maxOffset.x * -1), 0), Math.min(Math.max(slideOffset.y * -1, maxOffset.y * -1), 0));\n\t\t\t\tif (this.options.loop) {\n\t\t\t\t\tif (!this.options.vertical && Math.abs(this._position.x) > maxOffset.x) {\n\t\t\t\t\t\tnextOffset.x = 0;\n\t\t\t\t\t\tthis.slider.state.next = 0;\n\t\t\t\t\t} else if (this.options.vertical && Math.abs(this._position.y) > maxOffset.y) {\n\t\t\t\t\t\tnextOffset.y = 0;\n\t\t\t\t\t\tthis.slider.state.next = 0;\n\t\t\t\t\t}\n\t\t\t\t}\n\n\t\t\t\tthis._position.x = nextOffset.x;\n\t\t\t\tthis._position.y = nextOffset.y;\n\t\t\t\tif (this.options.centerMode) {\n\t\t\t\t\tthis._position.x = this._position.x + this.slider.wrapperWidth / 2 - Object(__WEBPACK_IMPORTED_MODULE_1__utils_css__[\"e\" /* width */])(slide) / 2;\n\t\t\t\t}\n\n\t\t\t\tif (this.slider.direction === 'rtl') {\n\t\t\t\t\tthis._position.x = -this._position.x;\n\t\t\t\t\tthis._position.y = -this._position.y;\n\t\t\t\t}\n\t\t\t\tthis.slider.container.style.transform = 'translate3d(' + this._position.x + 'px, ' + this._position.y + 'px, 0)';\n\n\t\t\t\t/**\n * update the index with the nextIndex only if\n * the offset of the nextIndex is in the range of the maxOffset\n */\n\t\t\t\tif (slideOffset.x > maxOffset.x) {\n\t\t\t\t\tthis.slider.transitioner.end();\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}, {\n\t\tkey: 'onTransitionEnd',\n\t\tvalue: function onTransitionEnd(e) {\n\t\t\tif (this.options.effect === 'translate') {\n\n\t\t\t\tif (this.transitioner.isAnimating() && e.target == this.slider.container) {\n\t\t\t\t\tif (this.options.infinite) {\n\t\t\t\t\t\tthis.slider._infinite.onTransitionEnd(e);\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t\tthis.transitioner.end();\n\t\t\t}\n\t\t}\n\t}]);\n\n\treturn Translate;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (Translate);\n\n/***/ }),\n/* 22 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\nvar defaultOptions = {\n initialSlide: 0,\n slidesToScroll: 1,\n slidesToShow: 1,\n\n navigation: true,\n navigationKeys: true,\n navigationSwipe: true,\n\n pagination: true,\n\n loop: false,\n infinite: false,\n\n effect: 'translate',\n duration: 300,\n timing: 'ease',\n\n autoplay: false,\n autoplaySpeed: 3000,\n pauseOnHover: true,\n breakpoints: [{\n changePoint: 480,\n slidesToShow: 1,\n slidesToScroll: 1\n }, {\n changePoint: 640,\n slidesToShow: 2,\n slidesToScroll: 2\n }, {\n changePoint: 768,\n slidesToShow: 3,\n slidesToScroll: 3\n }],\n\n onReady: null,\n icons: {\n 'previous': '\\n \\n ',\n 'next': '\\n \\n '\n }\n};\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (defaultOptions);\n\n/***/ }),\n/* 23 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony default export */ __webpack_exports__[\"a\"] = (function (id) {\n return \"
\\n
\\n
\";\n});\n\n/***/ }),\n/* 24 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\n/* harmony default export */ __webpack_exports__[\"a\"] = (function () {\n return \"
\";\n});\n\n/***/ })\n/******/ ])[\"default\"];\n});" }, { "path": "docs/static/js/bulma-slider.js", "content": "(function webpackUniversalModuleDefinition(root, factory) {\n\tif(typeof exports === 'object' && typeof module === 'object')\n\t\tmodule.exports = factory();\n\telse if(typeof define === 'function' && define.amd)\n\t\tdefine([], factory);\n\telse if(typeof exports === 'object')\n\t\texports[\"bulmaSlider\"] = factory();\n\telse\n\t\troot[\"bulmaSlider\"] = factory();\n})(typeof self !== 'undefined' ? self : this, function() {\nreturn /******/ (function(modules) { // webpackBootstrap\n/******/ \t// The module cache\n/******/ \tvar installedModules = {};\n/******/\n/******/ \t// The require function\n/******/ \tfunction __webpack_require__(moduleId) {\n/******/\n/******/ \t\t// Check if module is in cache\n/******/ \t\tif(installedModules[moduleId]) {\n/******/ \t\t\treturn installedModules[moduleId].exports;\n/******/ \t\t}\n/******/ \t\t// Create a new module (and put it into the cache)\n/******/ \t\tvar module = installedModules[moduleId] = {\n/******/ \t\t\ti: moduleId,\n/******/ \t\t\tl: false,\n/******/ \t\t\texports: {}\n/******/ \t\t};\n/******/\n/******/ \t\t// Execute the module function\n/******/ \t\tmodules[moduleId].call(module.exports, module, module.exports, __webpack_require__);\n/******/\n/******/ \t\t// Flag the module as loaded\n/******/ \t\tmodule.l = true;\n/******/\n/******/ \t\t// Return the exports of the module\n/******/ \t\treturn module.exports;\n/******/ \t}\n/******/\n/******/\n/******/ \t// expose the modules object (__webpack_modules__)\n/******/ \t__webpack_require__.m = modules;\n/******/\n/******/ \t// expose the module cache\n/******/ \t__webpack_require__.c = installedModules;\n/******/\n/******/ \t// define getter function for harmony exports\n/******/ \t__webpack_require__.d = function(exports, name, getter) {\n/******/ \t\tif(!__webpack_require__.o(exports, name)) {\n/******/ \t\t\tObject.defineProperty(exports, name, {\n/******/ \t\t\t\tconfigurable: false,\n/******/ \t\t\t\tenumerable: true,\n/******/ \t\t\t\tget: getter\n/******/ \t\t\t});\n/******/ \t\t}\n/******/ \t};\n/******/\n/******/ \t// getDefaultExport function for compatibility with non-harmony modules\n/******/ \t__webpack_require__.n = function(module) {\n/******/ \t\tvar getter = module && module.__esModule ?\n/******/ \t\t\tfunction getDefault() { return module['default']; } :\n/******/ \t\t\tfunction getModuleExports() { return module; };\n/******/ \t\t__webpack_require__.d(getter, 'a', getter);\n/******/ \t\treturn getter;\n/******/ \t};\n/******/\n/******/ \t// Object.prototype.hasOwnProperty.call\n/******/ \t__webpack_require__.o = function(object, property) { return Object.prototype.hasOwnProperty.call(object, property); };\n/******/\n/******/ \t// __webpack_public_path__\n/******/ \t__webpack_require__.p = \"\";\n/******/\n/******/ \t// Load entry module and return exports\n/******/ \treturn __webpack_require__(__webpack_require__.s = 0);\n/******/ })\n/************************************************************************/\n/******/ ([\n/* 0 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\nObject.defineProperty(__webpack_exports__, \"__esModule\", { value: true });\n/* harmony export (binding) */ __webpack_require__.d(__webpack_exports__, \"isString\", function() { return isString; });\n/* harmony import */ var __WEBPACK_IMPORTED_MODULE_0__events__ = __webpack_require__(1);\nvar _extends = Object.assign || function (target) { for (var i = 1; i < arguments.length; i++) { var source = arguments[i]; for (var key in source) { if (Object.prototype.hasOwnProperty.call(source, key)) { target[key] = source[key]; } } } return target; };\n\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nvar _typeof = typeof Symbol === \"function\" && typeof Symbol.iterator === \"symbol\" ? function (obj) { return typeof obj; } : function (obj) { return obj && typeof Symbol === \"function\" && obj.constructor === Symbol && obj !== Symbol.prototype ? \"symbol\" : typeof obj; };\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\nfunction _possibleConstructorReturn(self, call) { if (!self) { throw new ReferenceError(\"this hasn't been initialised - super() hasn't been called\"); } return call && (typeof call === \"object\" || typeof call === \"function\") ? call : self; }\n\nfunction _inherits(subClass, superClass) { if (typeof superClass !== \"function\" && superClass !== null) { throw new TypeError(\"Super expression must either be null or a function, not \" + typeof superClass); } subClass.prototype = Object.create(superClass && superClass.prototype, { constructor: { value: subClass, enumerable: false, writable: true, configurable: true } }); if (superClass) Object.setPrototypeOf ? Object.setPrototypeOf(subClass, superClass) : subClass.__proto__ = superClass; }\n\n\n\nvar isString = function isString(unknown) {\n return typeof unknown === 'string' || !!unknown && (typeof unknown === 'undefined' ? 'undefined' : _typeof(unknown)) === 'object' && Object.prototype.toString.call(unknown) === '[object String]';\n};\n\nvar bulmaSlider = function (_EventEmitter) {\n _inherits(bulmaSlider, _EventEmitter);\n\n function bulmaSlider(selector) {\n var options = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : {};\n\n _classCallCheck(this, bulmaSlider);\n\n var _this = _possibleConstructorReturn(this, (bulmaSlider.__proto__ || Object.getPrototypeOf(bulmaSlider)).call(this));\n\n _this.element = typeof selector === 'string' ? document.querySelector(selector) : selector;\n // An invalid selector or non-DOM node has been provided.\n if (!_this.element) {\n throw new Error('An invalid selector or non-DOM node has been provided.');\n }\n\n _this._clickEvents = ['click'];\n /// Set default options and merge with instance defined\n _this.options = _extends({}, options);\n\n _this.onSliderInput = _this.onSliderInput.bind(_this);\n\n _this.init();\n return _this;\n }\n\n /**\n * Initiate all DOM element containing selector\n * @method\n * @return {Array} Array of all slider instances\n */\n\n\n _createClass(bulmaSlider, [{\n key: 'init',\n\n\n /**\n * Initiate plugin\n * @method init\n * @return {void}\n */\n value: function init() {\n this._id = 'bulmaSlider' + new Date().getTime() + Math.floor(Math.random() * Math.floor(9999));\n this.output = this._findOutputForSlider();\n\n this._bindEvents();\n\n if (this.output) {\n if (this.element.classList.contains('has-output-tooltip')) {\n // Get new output position\n var newPosition = this._getSliderOutputPosition();\n\n // Set output position\n this.output.style['left'] = newPosition.position;\n }\n }\n\n this.emit('bulmaslider:ready', this.element.value);\n }\n }, {\n key: '_findOutputForSlider',\n value: function _findOutputForSlider() {\n var _this2 = this;\n\n var result = null;\n var outputs = document.getElementsByTagName('output') || [];\n\n Array.from(outputs).forEach(function (output) {\n if (output.htmlFor == _this2.element.getAttribute('id')) {\n result = output;\n return true;\n }\n });\n return result;\n }\n }, {\n key: '_getSliderOutputPosition',\n value: function _getSliderOutputPosition() {\n // Update output position\n var newPlace, minValue;\n\n var style = window.getComputedStyle(this.element, null);\n // Measure width of range input\n var sliderWidth = parseInt(style.getPropertyValue('width'), 10);\n\n // Figure out placement percentage between left and right of input\n if (!this.element.getAttribute('min')) {\n minValue = 0;\n } else {\n minValue = this.element.getAttribute('min');\n }\n var newPoint = (this.element.value - minValue) / (this.element.getAttribute('max') - minValue);\n\n // Prevent bubble from going beyond left or right (unsupported browsers)\n if (newPoint < 0) {\n newPlace = 0;\n } else if (newPoint > 1) {\n newPlace = sliderWidth;\n } else {\n newPlace = sliderWidth * newPoint;\n }\n\n return {\n 'position': newPlace + 'px'\n };\n }\n\n /**\n * Bind all events\n * @method _bindEvents\n * @return {void}\n */\n\n }, {\n key: '_bindEvents',\n value: function _bindEvents() {\n if (this.output) {\n // Add event listener to update output when slider value change\n this.element.addEventListener('input', this.onSliderInput, false);\n }\n }\n }, {\n key: 'onSliderInput',\n value: function onSliderInput(e) {\n e.preventDefault();\n\n if (this.element.classList.contains('has-output-tooltip')) {\n // Get new output position\n var newPosition = this._getSliderOutputPosition();\n\n // Set output position\n this.output.style['left'] = newPosition.position;\n }\n\n // Check for prefix and postfix\n var prefix = this.output.hasAttribute('data-prefix') ? this.output.getAttribute('data-prefix') : '';\n var postfix = this.output.hasAttribute('data-postfix') ? this.output.getAttribute('data-postfix') : '';\n\n // Update output with slider value\n this.output.value = prefix + this.element.value + postfix;\n\n this.emit('bulmaslider:ready', this.element.value);\n }\n }], [{\n key: 'attach',\n value: function attach() {\n var _this3 = this;\n\n var selector = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : 'input[type=\"range\"].slider';\n var options = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : {};\n\n var instances = new Array();\n\n var elements = isString(selector) ? document.querySelectorAll(selector) : Array.isArray(selector) ? selector : [selector];\n elements.forEach(function (element) {\n if (typeof element[_this3.constructor.name] === 'undefined') {\n var instance = new bulmaSlider(element, options);\n element[_this3.constructor.name] = instance;\n instances.push(instance);\n } else {\n instances.push(element[_this3.constructor.name]);\n }\n });\n\n return instances;\n }\n }]);\n\n return bulmaSlider;\n}(__WEBPACK_IMPORTED_MODULE_0__events__[\"a\" /* default */]);\n\n/* harmony default export */ __webpack_exports__[\"default\"] = (bulmaSlider);\n\n/***/ }),\n/* 1 */\n/***/ (function(module, __webpack_exports__, __webpack_require__) {\n\n\"use strict\";\nvar _createClass = function () { function defineProperties(target, props) { for (var i = 0; i < props.length; i++) { var descriptor = props[i]; descriptor.enumerable = descriptor.enumerable || false; descriptor.configurable = true; if (\"value\" in descriptor) descriptor.writable = true; Object.defineProperty(target, descriptor.key, descriptor); } } return function (Constructor, protoProps, staticProps) { if (protoProps) defineProperties(Constructor.prototype, protoProps); if (staticProps) defineProperties(Constructor, staticProps); return Constructor; }; }();\n\nfunction _classCallCheck(instance, Constructor) { if (!(instance instanceof Constructor)) { throw new TypeError(\"Cannot call a class as a function\"); } }\n\nvar EventEmitter = function () {\n function EventEmitter() {\n var listeners = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : [];\n\n _classCallCheck(this, EventEmitter);\n\n this._listeners = new Map(listeners);\n this._middlewares = new Map();\n }\n\n _createClass(EventEmitter, [{\n key: \"listenerCount\",\n value: function listenerCount(eventName) {\n if (!this._listeners.has(eventName)) {\n return 0;\n }\n\n var eventListeners = this._listeners.get(eventName);\n return eventListeners.length;\n }\n }, {\n key: \"removeListeners\",\n value: function removeListeners() {\n var _this = this;\n\n var eventName = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : null;\n var middleware = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : false;\n\n if (eventName !== null) {\n if (Array.isArray(eventName)) {\n name.forEach(function (e) {\n return _this.removeListeners(e, middleware);\n });\n } else {\n this._listeners.delete(eventName);\n\n if (middleware) {\n this.removeMiddleware(eventName);\n }\n }\n } else {\n this._listeners = new Map();\n }\n }\n }, {\n key: \"middleware\",\n value: function middleware(eventName, fn) {\n var _this2 = this;\n\n if (Array.isArray(eventName)) {\n name.forEach(function (e) {\n return _this2.middleware(e, fn);\n });\n } else {\n if (!Array.isArray(this._middlewares.get(eventName))) {\n this._middlewares.set(eventName, []);\n }\n\n this._middlewares.get(eventName).push(fn);\n }\n }\n }, {\n key: \"removeMiddleware\",\n value: function removeMiddleware() {\n var _this3 = this;\n\n var eventName = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : null;\n\n if (eventName !== null) {\n if (Array.isArray(eventName)) {\n name.forEach(function (e) {\n return _this3.removeMiddleware(e);\n });\n } else {\n this._middlewares.delete(eventName);\n }\n } else {\n this._middlewares = new Map();\n }\n }\n }, {\n key: \"on\",\n value: function on(name, callback) {\n var _this4 = this;\n\n var once = arguments.length > 2 && arguments[2] !== undefined ? arguments[2] : false;\n\n if (Array.isArray(name)) {\n name.forEach(function (e) {\n return _this4.on(e, callback);\n });\n } else {\n name = name.toString();\n var split = name.split(/,|, | /);\n\n if (split.length > 1) {\n split.forEach(function (e) {\n return _this4.on(e, callback);\n });\n } else {\n if (!Array.isArray(this._listeners.get(name))) {\n this._listeners.set(name, []);\n }\n\n this._listeners.get(name).push({ once: once, callback: callback });\n }\n }\n }\n }, {\n key: \"once\",\n value: function once(name, callback) {\n this.on(name, callback, true);\n }\n }, {\n key: \"emit\",\n value: function emit(name, data) {\n var _this5 = this;\n\n var silent = arguments.length > 2 && arguments[2] !== undefined ? arguments[2] : false;\n\n name = name.toString();\n var listeners = this._listeners.get(name);\n var middlewares = null;\n var doneCount = 0;\n var execute = silent;\n\n if (Array.isArray(listeners)) {\n listeners.forEach(function (listener, index) {\n // Start Middleware checks unless we're doing a silent emit\n if (!silent) {\n middlewares = _this5._middlewares.get(name);\n // Check and execute Middleware\n if (Array.isArray(middlewares)) {\n middlewares.forEach(function (middleware) {\n middleware(data, function () {\n var newData = arguments.length > 0 && arguments[0] !== undefined ? arguments[0] : null;\n\n if (newData !== null) {\n data = newData;\n }\n doneCount++;\n }, name);\n });\n\n if (doneCount >= middlewares.length) {\n execute = true;\n }\n } else {\n execute = true;\n }\n }\n\n // If Middleware checks have been passed, execute\n if (execute) {\n if (listener.once) {\n listeners[index] = null;\n }\n listener.callback(data);\n }\n });\n\n // Dirty way of removing used Events\n while (listeners.indexOf(null) !== -1) {\n listeners.splice(listeners.indexOf(null), 1);\n }\n }\n }\n }]);\n\n return EventEmitter;\n}();\n\n/* harmony default export */ __webpack_exports__[\"a\"] = (EventEmitter);\n\n/***/ })\n/******/ ])[\"default\"];\n});" }, { "path": "docs/static/js/index.js", "content": "window.HELP_IMPROVE_VIDEOJS = false;\n\nvar INTERP_BASE = \"./static/interpolation/stacked\";\nvar NUM_INTERP_FRAMES = 240;\n\nvar interp_images = [];\nfunction preloadInterpolationImages() {\n for (var i = 0; i < NUM_INTERP_FRAMES; i++) {\n var path = INTERP_BASE + '/' + String(i).padStart(6, '0') + '.jpg';\n interp_images[i] = new Image();\n interp_images[i].src = path;\n }\n}\n\nfunction setInterpolationImage(i) {\n var image = interp_images[i];\n image.ondragstart = function() { return false; };\n image.oncontextmenu = function() { return false; };\n $('#interpolation-image-wrapper').empty().append(image);\n}\n\n\n$(document).ready(function() {\n // Check for click events on the navbar burger icon\n $(\".navbar-burger\").click(function() {\n // Toggle the \"is-active\" class on both the \"navbar-burger\" and the \"navbar-menu\"\n $(\".navbar-burger\").toggleClass(\"is-active\");\n $(\".navbar-menu\").toggleClass(\"is-active\");\n\n });\n\n var options = {\n\t\t\tslidesToScroll: 1,\n\t\t\tslidesToShow: 3,\n\t\t\tloop: true,\n\t\t\tinfinite: true,\n\t\t\tautoplay: false,\n\t\t\tautoplaySpeed: 3000,\n }\n\n\t\t// Initialize all div with carousel class\n var carousels = bulmaCarousel.attach('.carousel', options);\n\n // Loop on each carousel initialized\n for(var i = 0; i < carousels.length; i++) {\n \t// Add listener to event\n \tcarousels[i].on('before:show', state => {\n \t\tconsole.log(state);\n \t});\n }\n\n // Access to bulmaCarousel instance of an element\n var element = document.querySelector('#my-element');\n if (element && element.bulmaCarousel) {\n \t// bulmaCarousel instance is available as element.bulmaCarousel\n \telement.bulmaCarousel.on('before-show', function(state) {\n \t\tconsole.log(state);\n \t});\n }\n\n /*var player = document.getElementById('interpolation-video');\n player.addEventListener('loadedmetadata', function() {\n $('#interpolation-slider').on('input', function(event) {\n console.log(this.value, player.duration);\n player.currentTime = player.duration / 100 * this.value;\n })\n }, false);*/\n preloadInterpolationImages();\n\n $('#interpolation-slider').on('input', function(event) {\n setInterpolationImage(this.value);\n });\n setInterpolationImage(0);\n $('#interpolation-slider').prop('max', NUM_INTERP_FRAMES - 1);\n\n bulmaSlider.attach();\n\n})\n" }, { "path": "eval.sh", "content": "python evaluate.py \\\n --base_method merge \\\n --comp_method composite \\\n --compos_num 2 \\\n --image_style anime \\\n --image_path output \\\n --save_path eval_result \\\n" }, { "path": "evaluate.py", "content": "import os\nimport re\nimport json\nimport openai\nimport base64\nimport logging\nimport requests\nimport argparse\nfrom tqdm import tqdm\nfrom os.path import join, exists\nfrom openai import OpenAI\n\nfrom utils import load_lora_info, generate_combinations\nfrom utils import get_eval_prompt\n\nclass GPT4V:\n def comparative_evaluate(\n self, prompt, image_1, image_2, max_tokens=2048, temperature=1.0, max_retries=5, **kwargs\n ):\n self.api_key = os.environ.get('OPENAI_API_KEY', None)\n self.client = OpenAI(api_key=self.api_key)\n retry_interval_exp = 1 \n retry_count = 0\n while retry_count < max_retries:\n try:\n response = self.client.chat.completions.create(\n model=\"gpt-4-vision-preview\",\n messages=[\n {\n \"role\": \"user\",\n \"content\": [\n {\n \"type\": \"text\",\n \"text\": prompt,\n },\n {\n \"type\": \"image_url\",\n \"image_url\": {\n \"url\": f\"data:image/png;base64,{image_1}\"\n }\n },\n {\n \"type\": \"image_url\",\n \"image_url\": {\n \"url\": f\"data:image/png;base64,{image_2}\"\n }\n },\n ]\n }\n ],\n max_tokens=max_tokens,\n temperature=temperature,\n )\n return response.choices[0].message.content\n except openai.RateLimitError:\n logging.warning(\"OpenAI rate limit error. Retry!\")\n except openai.APIConnectionError:\n logging.warning(\"OpenAI API connection error. Retry!\")\n except openai.APITimeoutError:\n logging.warning(\"OpenAI timeout error. Retry!\")\n except Exception as e:\n logging.error(f\"Unexpected error: {e}\")\n break\n\n # Simple backoff mechanism\n time.sleep(min(60, 0.5 * (2 ** retry_interval_exp)))\n retry_interval_exp += 1\n retry_count += 1\n\n return \"An error occurred while processing the request.\"\n\n# Function to encode the image\ndef encode_image(image_path):\n with open(image_path, \"rb\") as image_file:\n return base64.b64encode(image_file.read()).decode('utf-8')\n\ndef parse_scores(text):\n # Regular expression pattern to match scores, \n pattern = r\"image (\\d): composition quality: ([\\d\\.]+)\\/10.*?image quality: ([\\d\\.]+)\\/10\"\n \n # Find all matches in the evaluation text, case-insensitive\n matches = re.findall(pattern, text, re.IGNORECASE)\n\n # Check if exactly two images are present\n if len(matches) != 2:\n return False, \"Expected scores for exactly two images\"\n\n results = {}\n for match in matches:\n image_number, comp_quality, image_quality = match\n comp_quality = float(comp_quality)\n image_quality = float(image_quality)\n\n # Check if scores are within the valid range\n if not (0 <= comp_quality <= 10) or not (0 <= image_quality <= 10):\n return False, \"Scores must be between 0 and 10\"\n\n results[f'image {image_number}'] = {\n 'composition quality': comp_quality,\n 'image quality': image_quality\n }\n\n return True, results\n\ndef evaluate(args):\n\n image_path = f\"{args.image_path}_{args.image_style}\"\n image_path = join(image_path, f'{args.compos_num}_elements')\n\n # load all the information of LoRAs\n lora_info = load_lora_info(args.image_style, args.lora_info_path)\n\n # generate all combinations that can be composed\n combinations = generate_combinations(lora_info, args.compos_num)\n\n # comparative evaluation\n gpt4v = GPT4V()\n all_eval = []\n for combo in tqdm(combinations):\n # get the image path\n elements = '_'.join([lora['id'] for lora in combo])\n image_1_path = join(image_path, args.base_method + '_' + elements + '.png')\n image_2_path = join(image_path, args.comp_method + '_' + elements + '.png')\n if not exists(image_1_path) or not exists(image_2_path):\n print(f\"Can't find the generate images for {elements}\")\n continue\n \n # encode the images\n image_1 = encode_image(image_1_path)\n image_2 = encode_image(image_2_path)\n\n # get the prompt for the comparative evaluation\n prompt = get_eval_prompt(combo)\n # print(prompt)\n\n # comparative evaluation\n # If the scores cannot be parsed from the evaluation result, then retry\n retry_cnt = 0\n max_retries = 10\n while retry_cnt < max_retries:\n result = gpt4v.comparative_evaluate(prompt, image_1, image_2)\n print(result)\n valid, scores = parse_scores(result)\n if valid == True:\n cur_eval = {}\n cur_eval['elements'] = elements\n cur_eval['method 1'] = args.base_method\n cur_eval['method 2'] = args.comp_method\n cur_eval['eval'] = result\n cur_eval['scores'] = scores\n all_eval.append(cur_eval)\n break\n else:\n print(scores)\n print(f\"Retry for {elements}\")\n retry_cnt += 1\n if retry_cnt == max_retries:\n print(f\"Can't get evaluation scores for {elements}!\")\n\n # save the evaluation results\n if not exists(args.save_path):\n os.makedirs(args.save_path)\n save_path = join(args.save_path, f'{args.image_style}_{args.compos_num}_elements_{args.base_method}_vs_{args.comp_method}.json')\n with open(save_path, 'w') as f:\n json.dump(all_eval, f, indent=4, ensure_ascii=False)\n\nif __name__ == \"__main__\":\n parser = argparse.ArgumentParser(\n description='Evaluate the generated images based on composition efficacy and image quality'\n )\n\n parser.add_argument('--image_path', default='output',\n help='path to store the generated image', type=str)\n parser.add_argument('--save_path', default='eval_result',\n help='path to save the evaluation results', type=str)\n parser.add_argument('--base_method', default='merge',\n choices=['merge', 'switch', 'composite'],\n help='the first method used for comparative evaluation', type=str)\n parser.add_argument('--comp_method', default='composite',\n choices=['merge', 'switch', 'composite'],\n help='the first method used for comparative evaluation', type=str)\n parser.add_argument('--compos_num', default=2,\n help='number of elements to be evaluated in a single image', type=int)\n parser.add_argument('--image_style', default='reality',\n choices=['anime', 'reality'],\n help='styles of images to be evaluated', type=str)\n parser.add_argument('--lora_info_path', default='lora_info.json',\n help='path to stroe all LoRA information', type=str)\n\n args = parser.parse_args()\n\n evaluate(args)\n" }, { "path": "example.py", "content": "import torch\nimport argparse\nfrom diffusers import DiffusionPipeline, StableDiffusionPipeline, AutoencoderKL\nfrom diffusers import DPMSolverMultistepScheduler\nfrom callbacks import make_callback\n\ndef get_example_prompt():\n prompt = \"RAW photo, subject, 8k uhd, dslr, high quality, Fujifilm XT3, half-length portrait from knees up, scarlett, short red hair, blue eyes, school uniform, white shirt, red tie, blue pleated microskirt\"\n negative_prompt = \"extra heads, nsfw, deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, text, cropped, out of frame, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck\"\n return prompt, negative_prompt\n\ndef main(args):\n\n # set the prompts for image generation\n prompt, negative_prompt = get_example_prompt()\n\n # base model for the realistic style example\n model_name = 'SG161222/Realistic_Vision_V5.1_noVAE'\n\n # set base model\n pipeline = DiffusionPipeline.from_pretrained(\n model_name,\n custom_pipeline=\"./pipelines/sd1.5_0.26.3\",\n use_safetensors=True\n ).to(\"cuda\")\n\n # set vae\n vae = AutoencoderKL.from_pretrained(\n \"stabilityai/sd-vae-ft-mse\",\n ).to(\"cuda\")\n pipeline.vae = vae\n\n # set scheduler\n schedule_config = dict(pipeline.scheduler.config)\n schedule_config[\"algorithm_type\"] = \"dpmsolver++\"\n pipeline.scheduler = DPMSolverMultistepScheduler.from_config(schedule_config)\n\n # initialize LoRAs\n # This example shows the composition of a character LoRA and a clothing LoRA\n pipeline.load_lora_weights(args.lora_path, weight_name=\"character_2.safetensors\", adapter_name=\"character\")\n pipeline.load_lora_weights(args.lora_path, weight_name=\"clothing_2.safetensors\", adapter_name=\"clothing\")\n cur_loras = [\"character\", \"clothing\"]\n\n # select the method for the composition\n if args.method == \"merge\":\n pipeline.set_adapters(cur_loras)\n switch_callback = None\n elif args.method == \"switch\":\n pipeline.set_adapters([cur_loras[0]])\n switch_callback = make_callback(switch_step=args.switch_step, loras=cur_loras)\n else:\n pipeline.set_adapters(cur_loras)\n switch_callback = None\n\n image = pipeline(\n prompt=prompt, \n negative_prompt=negative_prompt,\n height=args.height,\n width=args.width,\n num_inference_steps=args.denoise_steps,\n guidance_scale=args.cfg_scale,\n generator=args.generator,\n cross_attention_kwargs={\"scale\": args.lora_scale},\n callback_on_step_end=switch_callback,\n lora_composite=True if args.method == \"composite\" else False\n ).images[0]\n\n image.save(args.save_path)\n\nif __name__ == \"__main__\":\n parser = argparse.ArgumentParser(\n description='Example code for multi-LoRA composition'\n )\n\n # Arguments for composing LoRAs\n parser.add_argument('--method', default='switch',\n choices=['merge', 'switch', 'composite'],\n help='methods for combining LoRAs', type=str)\n parser.add_argument('--save_path', default='example.png',\n help='path to save the generated image', type=str)\n parser.add_argument('--lora_path', default='models/lora/reality',\n help='path to store all LoRAs', type=str)\n parser.add_argument('--lora_scale', default=0.8,\n help='scale of each LoRA when generating images', type=float)\n parser.add_argument('--switch_step', default=5,\n help='number of steps to switch LoRA during denoising, applicable only in the switch method', type=int)\n\n # Arguments for generating images\n parser.add_argument('--height', default=1024,\n help='height of the generated images', type=int)\n parser.add_argument('--width', default=768,\n help='width of the generated images', type=int)\n parser.add_argument('--denoise_steps', default=50,\n help='number of the denoising steps', type=int)\n parser.add_argument('--cfg_scale', default=7,\n help='scale for classifier-free guidance', type=float)\n parser.add_argument('--seed', default=11,\n help='seed for generating images', type=int)\n\n args = parser.parse_args()\n args.generator = torch.manual_seed(args.seed)\n\n main(args)" }, { "path": "human_eval/README.md", "content": "# Human Evaluations of Generated Images\n\nWe conduct detailed human evaluations on a subset of images generated by different methods. If you're interested in exploring these images and their associated data, follow the instructions and information provided below.\n\n## Accessing the Generated Images\nYou can download a sample of 120 generated images for evaluation from the following link: [Download Generated Images](https://drive.google.com/file/d/12Prm1cJRiOMSOEVF3H6-rIfWyZbsMASN/view?usp=sharing). Please unzip the file in this folder for the following analysis.\n\nAlongside each image, we provide:\n\n- Detailed prompts used for generating and evaluting the image\n- Human evaluation scores for composition and image quality\n- CLIPScore for each image\n- Scores on composition and image quality based on GPT-4V evaluations\n\nAll the detailed information is available in `results.json`.\n\n## Evaluation Results by Method\nWe evaluate the images generated by different methods using several metrics. Here are the summarized results:\n```\n+-----------+-------------------+-------------+-----------+-------------------+-------------+\n| Methods | Human-Composition | Human-Image | CLIPScore | GPT4V-Composition | GPT4V-Image |\n+-----------+-------------------+-------------+-----------+-------------------+-------------+\n| merge | 3.14 | 2.94 | 33.42 | 7.02 | 9.24 |\n| switch | 3.91 | 4.15 | 35.52 | 7.22 | 9.53 |\n| composite | 3.77 | 4.35 | 34.85 | 7.13 | 9.55 |\n+-----------+-------------------+-------------+-----------+-------------------+-------------+\n```\n\n## Correlations Between Metrics and Human Judgements\nTo further understand the effectiveness of each metric, we calculate their correlations with human judgements. The correlation scores are as follows:\n```\n+-----------------------------------------+---------+----------+---------+\n| Metrics | Pearson | Spearman | Kendall |\n+-----------------------------------------+---------+----------+---------+\n| CLIPScore --- Human Composition | -0.006 | 0.024 | 0.019 |\n| CLIPScore --- Human Image Quliaty | 0.083 | 0.08 | 0.055 |\n| GPT4V Composition --- Human Composition | 0.454 | 0.449 | 0.337 |\n| GPT4V Image --- Human Image | 0.457 | 0.426 | 0.337 |\n+-----------------------------------------+---------+----------+---------+\n```" }, { "path": "human_eval/calculate_correlations.py", "content": "import json\nfrom prettytable import PrettyTable\nfrom scipy.stats import spearmanr, pearsonr, kendalltau\n\ndef init_metric():\n scores = ['human_compos', 'human_image', 'clip_score', 'gpt4v_compos', 'gpt4v_image']\n methods = ['merge', 'switch', 'composite']\n return {method: {score: 0 for score in scores} for method in methods}\n\ndef print_metric(metric):\n table = PrettyTable(['Methods','Human-Composition', 'Human-Image', 'CLIPScore', 'GPT4V-Composition', 'GPT4V-Image'])\n for method, scores in metric.items():\n row = [method] + [round(scores[score], 2) for score in scores]\n table.add_row(row)\n print(table)\n\ndef calculate_correlation(pred_score, human_score):\n assert len(pred_score) == len(human_score)\n correlations = [pearsonr(pred_score, human_score)[0], spearmanr(pred_score, human_score)[0], kendalltau(pred_score, human_score)[0]]\n return correlations\n\ndef print_correlation(correlation_results):\n table = PrettyTable(['Metrics', 'Pearson', 'Spearman', 'Kendall'])\n for name, scores in correlation_results.items():\n table.add_row([name] + [round(score, 3) for score in scores])\n print(table)\n\ndef main(results):\n metric = init_metric()\n counts = {method: 0 for method in metric.keys()}\n\n clip_score, human_compos, human_image, gpt_compos, gpt_image = [], [], [], [], []\n\n for i, result in enumerate(results):\n method = ['merge', 'switch', 'composite'][i % 3]\n metric[method]['human_compos'] += result['avg_human_score']['composition']\n metric[method]['human_image'] += result['avg_human_score']['image']\n metric[method]['clip_score'] += result['clipscore']['score']\n metric[method]['gpt4v_compos'] += result['gpt4v']['composition']\n metric[method]['gpt4v_image'] += result['gpt4v']['image']\n counts[method] += 1\n\n human_compos.append(result['avg_human_score']['composition'])\n human_image.append(result['avg_human_score']['image'])\n clip_score.append(result['clipscore']['score'])\n gpt_compos.append(result['gpt4v']['composition'])\n gpt_image.append(result['gpt4v']['image'])\n\n for method in metric:\n for score in metric[method]:\n metric[method][score] /= counts[method]\n \n print('Results for each method:')\n print_metric(metric)\n\n correlation_results = {\n 'CLIPScore --- Human Composition': calculate_correlation(clip_score, human_compos),\n 'CLIPScore --- Human Image Quliaty': calculate_correlation(clip_score, human_image),\n 'GPT4V Composition --- Human Composition': calculate_correlation(gpt_compos, human_compos),\n 'GPT4V Image --- Human Image': calculate_correlation(gpt_image, human_image)\n }\n\n print('Correlations between different metrics with human judgements:')\n print_correlation(correlation_results)\n\nif __name__ == \"__main__\":\n with open('results.json') as f:\n results = json.loads(f.read())\n main(results)\n" }, { "path": "human_eval/clipscore.py", "content": "import json\nimport torch\nimport numpy as np\nfrom tqdm import tqdm\nfrom os.path import join\nfrom functools import partial\nfrom diffusers.utils import load_image\nfrom torchmetrics.functional.multimodal import clip_score\n\nclip_score_fn = partial(clip_score, model_name_or_path=\"openai/clip-vit-base-patch16\")\nimage_n = 120 # number of images to evaluate\n\ndef calculate_clip_score(images, prompts):\n images_int = (images * 255).astype(\"uint8\")\n clip_score = clip_score_fn(torch.from_numpy(images_int).permute(0, 3, 1, 2), prompts).detach()\n return round(float(clip_score), 4)\n\n# Load images\n# The size of anime and realistic style images is different.\nimage_path = 'images'\nimages = []\nfor i in range(1, image_n + 1):\n cur_image = load_image(join(image_path, f'{i}.png'))\n # scale pixel values to [0, 1]\n cur_image = np.array(cur_image, dtype=np.float32) / 255.0\n images.append(cur_image)\n\n# Load prompts\nprompts = []\nwith open('image_info.json') as f:\n image_info = json.loads(f.read())\nfor i in range(len(image_info)):\n cur_prompt = ', '.join(content.split(': ')[1] for content in image_info[i]['prompt'])\n prompts.append(cur_prompt)\n\n# Calculate CLIP score\nscores = []\nfor i in tqdm(range(image_n)):\n cur_image = np.expand_dims(images[i], axis=0)\n cur_prompt = [prompts[i]]\n sd_clip_score = calculate_clip_score(cur_image, cur_prompt)\n scores.append(sd_clip_score)\n\n# Save results\nresult = []\nfor i in range(len(scores)):\n cur = {}\n cur['index'] = i + 1\n cur['clip_score'] = scores[i]\n result.append(cur)\nwith open('clip_results.json', 'w') as f:\n json.dump(result, f, indent=4, ensure_ascii=False)\n" }, { "path": "human_eval/gpt4v.py", "content": "import os\nimport re\nimport json\nimport copy\nimport openai\nimport base64\nimport logging\nimport requests\nimport argparse\nfrom tqdm import tqdm\nfrom os.path import join, exists\nfrom openai import OpenAI\n\nclass GPT4V:\n def comparative_evaluate(\n self, prompt, image_1, image_2, max_tokens=2048, temperature=1.0, max_retries=5, **kwargs\n ):\n self.api_key = os.environ.get('OPENAI_API_KEY', None)\n self.client = OpenAI(api_key=self.api_key)\n retry_interval_exp = 1 \n retry_count = 0\n while retry_count < max_retries:\n try:\n response = self.client.chat.completions.create(\n model=\"gpt-4-vision-preview\",\n messages=[\n {\n \"role\": \"user\",\n \"content\": [\n {\n \"type\": \"text\",\n \"text\": prompt,\n },\n {\n \"type\": \"image_url\",\n \"image_url\": {\n \"url\": f\"data:image/png;base64,{image_1}\"\n }\n },\n {\n \"type\": \"image_url\",\n \"image_url\": {\n \"url\": f\"data:image/png;base64,{image_2}\"\n }\n },\n ]\n }\n ],\n max_tokens=max_tokens,\n temperature=temperature,\n )\n return response.choices[0].message.content\n except openai.RateLimitError:\n logging.warning(\"OpenAI rate limit error. Retry!\")\n except openai.APIConnectionError:\n logging.warning(\"OpenAI API connection error. Retry!\")\n except openai.APITimeoutError:\n logging.warning(\"OpenAI timeout error. Retry!\")\n except Exception as e:\n logging.error(f\"Unexpected error: {e}\")\n break\n\n # Simple backoff mechanism\n time.sleep(min(60, 0.5 * (2 ** retry_interval_exp)))\n retry_interval_exp += 1\n retry_count += 1\n\n return \"An error occurred while processing the request.\"\n\n# Function to encode the image\ndef encode_image(image_path):\n with open(image_path, \"rb\") as image_file:\n return base64.b64encode(image_file.read()).decode('utf-8')\n\ndef get_eval_prompt(element_prompt):\n prompt = \"I need assistance in comparatively evaluating two text-to-image models based on their ability to compose different elements into a single image. The elements and their key features are as follows:\\n\\n\"\n prompt += element_prompt + '\\n\\n'\n prompt += \"Please help me rate both given images on the following evaluation dimensions and criteria:\\n\\nComposition quality:\\n- Score on a scale of 0 to 10, in 0.5 increments, where 10 is the best and 0 is the worst.\\n- Deduct 3 points if any element is missing or incorrectly depicted.\\n- Deduct 1 point for each missing or incorrect feature within an element.\\n- Deduct 1 point for minor inconsistencies or lack of harmony between elements.\\n- Additional deductions can be made for compositions that lack coherence, creativity, or realism.\\n\\nImage quality:\\n- Score on a scale of 0 to 10, in 0.5 increments, where 10 is the best and 0 is the worst.\\n- Deduct 3 points for each deformity in the image (e.g., extra limbs or fingers, distorted face, incorrect proportions).\\n- Deduct 2 points for noticeable issues with texture, lighting, or color.\\n- Deduct 1 point for each minor flaw or imperfection.\\n- Additional deductions can be made for any issues affecting the overall aesthetic or clarity of the image.\\n\\nPlease format the evaluation as follows:\\n\\nFor Image 1:\\n[Explanation of evaluation]\\n\\nFor Image 2:\\n[Explanation of evaluation]\\n\\nScores:\\nImage 1: Composition Quality: [score]/10, Image Quality: [score]/10\\nImage 2: Composition Quality: [score]/10, Image Quality: [score]/10\\n\\nBased on the above guidelines, help me to conduct a step-by-step comparative evaluation of the given images. The scoring should follow two principles:\\n1. Please evaluate critically.\\n2. Try not to let the two models end in a tie on both dimensions.\"\n return prompt\n\ndef parse_scores(text):\n # Regular expression pattern to match scores, \n pattern = r\"image (\\d): composition quality: ([\\d\\.]+)\\/10.*?image quality: ([\\d\\.]+)\\/10\"\n \n # Find all matches in the evaluation text, case-insensitive\n matches = re.findall(pattern, text, re.IGNORECASE)\n\n # Check if exactly two images are present\n if len(matches) != 2:\n return False, \"Expected scores for exactly two images\"\n\n results = {}\n for match in matches:\n image_number, comp_quality, image_quality = match\n comp_quality = float(comp_quality)\n image_quality = float(image_quality)\n\n # Check if scores are within the valid range\n if not (0 <= comp_quality <= 10) or not (0 <= image_quality <= 10):\n return False, \"Scores must be between 0 and 10\"\n\n results[f'image {image_number}'] = {\n 'composition quality': comp_quality,\n 'image quality': image_quality\n }\n\n return True, results\n\ndef evaluate():\n image_n = 120 # number of images to evaluate\n gpt4v = GPT4V()\n \n # Load images\n image_path = 'images'\n # Initialize the list with a placeholder to ensure indexing starts at 1.\n images = [None]\n for i in range(1, image_n + 1):\n cur_image = encode_image(join(image_path, f'{i}.png'))\n images.append(cur_image)\n \n # Load prompts\n prompts = [None]\n with open('image_info.json') as f:\n image_info = json.loads(f.read())\n for i in range(len(image_info)):\n cur_prompt = '\\n'.join(image_info[i]['prompt'])\n prompts.append(cur_prompt)\n\n # Comparative evaluation\n gpt4v = GPT4V()\n gpt4v_scores = [{} for _ in range(image_n + 1)]\n # i: merge\n # i + 1: switch\n # i + 2: composite\n for i in tqdm(range(1, image_n + 1, 3)):\n merge_image = images[i]\n switch_image = images[i + 1]\n composite_image = images[i + 2]\n\n cur_prompt = get_eval_prompt(prompts[i])\n print(i)\n \n print(\"merge (image 1) vs switch (image 2)\")\n retry_cnt = 0\n max_retries = 20\n while retry_cnt < max_retries:\n result = gpt4v.comparative_evaluate(cur_prompt, merge_image, switch_image)\n valid, scores = parse_scores(result)\n if valid == True:\n print(scores)\n # merge\n if len(gpt4v_scores[i]) == 0:\n gpt4v_scores[i] = copy.deepcopy(scores['image 1'])\n else:\n gpt4v_scores[i]['composition quality'] += scores['image 1']['composition quality']\n gpt4v_scores[i]['image quality'] += scores['image 1']['image quality']\n # switch\n if len(gpt4v_scores[i + 1]) == 0:\n gpt4v_scores[i + 1] = copy.deepcopy(scores['image 2'])\n else:\n gpt4v_scores[i + 1]['composition quality'] += scores['image 2']['composition quality']\n gpt4v_scores[i + 1]['image quality'] += scores['image 2']['image quality']\n break\n else:\n print(f\"Retry for {i}.png\")\n retry_cnt += 1\n if retry_cnt == max_retries:\n print(f\"Can't get evaluation scores for {i}.png!\")\n\n print(\"switch (image 1) vs merge (image 2)\")\n retry_cnt = 0\n max_retries = 20\n while retry_cnt < max_retries:\n result = gpt4v.comparative_evaluate(cur_prompt, switch_image, merge_image)\n valid, scores = parse_scores(result)\n if valid == True:\n print(scores)\n # merge\n if len(gpt4v_scores[i]) == 0:\n gpt4v_scores[i] = copy.deepcopy(scores['image 2'])\n else:\n gpt4v_scores[i]['composition quality'] += scores['image 2']['composition quality']\n gpt4v_scores[i]['image quality'] += scores['image 2']['image quality']\n # switch\n if len(gpt4v_scores[i + 1]) == 0:\n gpt4v_scores[i + 1] = copy.deepcopy(scores['image 1'])\n else:\n gpt4v_scores[i + 1]['composition quality'] += scores['image 1']['composition quality']\n gpt4v_scores[i + 1]['image quality'] += scores['image 1']['image quality']\n break\n else:\n print(f\"Retry for {i}.png\")\n retry_cnt += 1\n if retry_cnt == max_retries:\n print(f\"Can't get evaluation scores for {i}.png!\")\n\n print(\"merge (image 1) vs composite (image 2)\")\n retry_cnt = 0\n max_retries = 20\n while retry_cnt < max_retries:\n result = gpt4v.comparative_evaluate(cur_prompt, merge_image, composite_image)\n valid, scores = parse_scores(result)\n if valid == True:\n print(scores)\n # merge\n if len(gpt4v_scores[i]) == 0:\n gpt4v_scores[i] = copy.deepcopy(scores['image 1'])\n else:\n gpt4v_scores[i]['composition quality'] += scores['image 1']['composition quality']\n gpt4v_scores[i]['image quality'] += scores['image 1']['image quality']\n # switch\n if len(gpt4v_scores[i + 2]) == 0:\n gpt4v_scores[i + 2] = copy.deepcopy(scores['image 2'])\n else:\n gpt4v_scores[i + 2]['composition quality'] += scores['image 2']['composition quality']\n gpt4v_scores[i + 2]['image quality'] += scores['image 2']['image quality']\n break\n else:\n print(f\"Retry for {i}.png\")\n retry_cnt += 1\n if retry_cnt == max_retries:\n print(f\"Can't get evaluation scores for {i}.png!\")\n \n print(\"composite (image 1) vs merge (image 2)\")\n retry_cnt = 0\n max_retries = 20\n while retry_cnt < max_retries:\n result = gpt4v.comparative_evaluate(cur_prompt, composite_image, merge_image)\n valid, scores = parse_scores(result)\n if valid == True:\n print(scores)\n # merge\n if len(gpt4v_scores[i]) == 0:\n gpt4v_scores[i] = copy.deepcopy(scores['image 2'])\n else:\n gpt4v_scores[i]['composition quality'] += scores['image 2']['composition quality']\n gpt4v_scores[i]['image quality'] += scores['image 2']['image quality']\n # switch\n if len(gpt4v_scores[i + 2]) == 0:\n gpt4v_scores[i + 2] = copy.deepcopy(scores['image 1'])\n else:\n gpt4v_scores[i + 2]['composition quality'] += scores['image 1']['composition quality']\n gpt4v_scores[i + 2]['image quality'] += scores['image 1']['image quality']\n break\n else:\n print(f\"Retry for {i}.png\")\n retry_cnt += 1\n if retry_cnt == max_retries:\n print(f\"Can't get evaluation scores for {i}.png!\")\n\n results = []\n for i in range(1, image_n + 1):\n cur = {}\n cur['index'] = i\n # merge\n if i % 3 == 1:\n cur['composition quality'] = gpt4v_scores[i]['composition quality'] / 4\n cur['image quality'] = gpt4v_scores[i]['image quality'] / 4\n else:\n cur['composition quality'] = gpt4v_scores[i]['composition quality'] / 2\n cur['image quality'] = gpt4v_scores[i]['image quality'] / 2\n results.append(cur)\n with open('gpt4v_results.json', 'w') as f:\n json.dump(results, f, indent=4, ensure_ascii=False)\n\nif __name__ == \"__main__\":\n evaluate()" }, { "path": "human_eval/image_info.json", "content": "[\n {\n \"index\": 1,\n \"path\": \"output_anime/2_elements/merge_character_1_style_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\"\n ]\n },\n {\n \"index\": 2,\n \"path\": \"output_anime/2_elements/switch_character_1_style_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\"\n ]\n },\n {\n \"index\": 3,\n \"path\": \"output_anime/2_elements/composite_character_1_style_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\"\n ]\n },\n {\n \"index\": 4,\n \"path\": \"output_anime/4_elements/merge_character_2_clothing_1_style_2_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\"\n ]\n },\n {\n \"index\": 5,\n \"path\": \"output_anime/4_elements/switch_character_2_clothing_1_style_2_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\"\n ]\n },\n {\n \"index\": 6,\n \"path\": \"output_anime/4_elements/composite_character_2_clothing_1_style_2_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\"\n ]\n },\n {\n \"index\": 7,\n \"path\": \"output_anime/5_elements/merge_character_3_clothing_1_style_1_background_1_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 8,\n \"path\": \"output_anime/5_elements/switch_character_3_clothing_1_style_1_background_1_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 9,\n \"path\": \"output_anime/5_elements/composite_character_3_clothing_1_style_1_background_1_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 10,\n \"path\": \"output_anime/3_elements/merge_character_3_clothing_2_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ]\n },\n {\n \"index\": 11,\n \"path\": \"output_anime/3_elements/switch_character_3_clothing_2_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ]\n },\n {\n \"index\": 12,\n \"path\": \"output_anime/3_elements/composite_character_3_clothing_2_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ]\n },\n {\n \"index\": 13,\n \"path\": \"output_anime/5_elements/merge_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 14,\n \"path\": \"output_anime/5_elements/switch_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 15,\n \"path\": \"output_anime/5_elements/composite_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 16,\n \"path\": \"output_anime/5_elements/merge_character_2_clothing_2_style_1_background_1_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 17,\n \"path\": \"output_anime/5_elements/switch_character_2_clothing_2_style_1_background_1_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 18,\n \"path\": \"output_anime/5_elements/composite_character_2_clothing_2_style_1_background_1_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 19,\n \"path\": \"output_anime/3_elements/merge_character_2_clothing_2_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ]\n },\n {\n \"index\": 20,\n \"path\": \"output_anime/3_elements/switch_character_2_clothing_2_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ]\n },\n {\n \"index\": 21,\n \"path\": \"output_anime/3_elements/composite_character_2_clothing_2_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ]\n },\n {\n \"index\": 22,\n \"path\": \"output_anime/4_elements/merge_character_1_clothing_2_background_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 23,\n \"path\": \"output_anime/4_elements/switch_character_1_clothing_2_background_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 24,\n \"path\": \"output_anime/4_elements/composite_character_1_clothing_2_background_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 25,\n \"path\": \"output_anime/4_elements/merge_character_1_clothing_2_background_2_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 26,\n \"path\": \"output_anime/4_elements/switch_character_1_clothing_2_background_2_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 27,\n \"path\": \"output_anime/4_elements/composite_character_1_clothing_2_background_2_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 28,\n \"path\": \"output_anime/4_elements/merge_character_2_clothing_2_style_2_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 29,\n \"path\": \"output_anime/4_elements/switch_character_2_clothing_2_style_2_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 30,\n \"path\": \"output_anime/4_elements/composite_character_2_clothing_2_style_2_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 31,\n \"path\": \"output_anime/3_elements/merge_character_3_style_2_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 32,\n \"path\": \"output_anime/3_elements/switch_character_3_style_2_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 33,\n \"path\": \"output_anime/3_elements/composite_character_3_style_2_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 34,\n \"path\": \"output_anime/4_elements/merge_character_1_clothing_1_style_2_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\"\n ]\n },\n {\n \"index\": 35,\n \"path\": \"output_anime/4_elements/switch_character_1_clothing_1_style_2_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\"\n ]\n },\n {\n \"index\": 36,\n \"path\": \"output_anime/4_elements/composite_character_1_clothing_1_style_2_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\"\n ]\n },\n {\n \"index\": 37,\n \"path\": \"output_anime/4_elements/merge_character_3_clothing_2_background_1_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 38,\n \"path\": \"output_anime/4_elements/switch_character_3_clothing_2_background_1_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 39,\n \"path\": \"output_anime/4_elements/composite_character_3_clothing_2_background_1_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 40,\n \"path\": \"output_anime/4_elements/merge_character_1_clothing_1_style_2_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 41,\n \"path\": \"output_anime/4_elements/switch_character_1_clothing_1_style_2_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 42,\n \"path\": \"output_anime/4_elements/composite_character_1_clothing_1_style_2_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 43,\n \"path\": \"output_anime/4_elements/merge_character_3_clothing_2_style_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 44,\n \"path\": \"output_anime/4_elements/switch_character_3_clothing_2_style_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 45,\n \"path\": \"output_anime/4_elements/composite_character_3_clothing_2_style_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 46,\n \"path\": \"output_anime/5_elements/merge_character_3_clothing_2_style_1_background_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 47,\n \"path\": \"output_anime/5_elements/switch_character_3_clothing_2_style_1_background_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 48,\n \"path\": \"output_anime/5_elements/composite_character_3_clothing_2_style_1_background_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 49,\n \"path\": \"output_anime/4_elements/merge_character_1_style_2_background_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 50,\n \"path\": \"output_anime/4_elements/switch_character_1_style_2_background_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 51,\n \"path\": \"output_anime/4_elements/composite_character_1_style_2_background_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 52,\n \"path\": \"output_anime/5_elements/merge_character_2_clothing_2_style_2_background_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 53,\n \"path\": \"output_anime/5_elements/switch_character_2_clothing_2_style_2_background_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 54,\n \"path\": \"output_anime/5_elements/composite_character_2_clothing_2_style_2_background_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ]\n },\n {\n \"index\": 55,\n \"path\": \"output_anime/4_elements/merge_character_1_clothing_2_style_2_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 56,\n \"path\": \"output_anime/4_elements/switch_character_1_clothing_2_style_2_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 57,\n \"path\": \"output_anime/4_elements/composite_character_1_clothing_2_style_2_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 58,\n \"path\": \"output_anime/4_elements/merge_character_2_clothing_2_style_1_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 59,\n \"path\": \"output_anime/4_elements/switch_character_2_clothing_2_style_1_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 60,\n \"path\": \"output_anime/4_elements/composite_character_2_clothing_2_style_1_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ]\n },\n {\n \"index\": 61,\n \"path\": \"output_reality/3_elements/merge_character_3_clothing_2_style_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\"\n ]\n },\n {\n \"index\": 62,\n \"path\": \"output_reality/3_elements/switch_character_3_clothing_2_style_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\"\n ]\n },\n {\n \"index\": 63,\n \"path\": \"output_reality/3_elements/composite_character_3_clothing_2_style_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\"\n ]\n },\n {\n \"index\": 64,\n \"path\": \"output_reality/2_elements/merge_character_1_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Forest Background): slg, river, forest\"\n ]\n },\n {\n \"index\": 65,\n \"path\": \"output_reality/2_elements/switch_character_1_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Forest Background): slg, river, forest\"\n ]\n },\n {\n \"index\": 66,\n \"path\": \"output_reality/2_elements/composite_character_1_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Forest Background): slg, river, forest\"\n ]\n },\n {\n \"index\": 67,\n \"path\": \"output_reality/4_elements/merge_character_2_clothing_2_style_1_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 68,\n \"path\": \"output_reality/4_elements/switch_character_2_clothing_2_style_1_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 69,\n \"path\": \"output_reality/4_elements/composite_character_2_clothing_2_style_1_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 70,\n \"path\": \"output_reality/4_elements/merge_character_1_clothing_1_style_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 71,\n \"path\": \"output_reality/4_elements/switch_character_1_clothing_1_style_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 72,\n \"path\": \"output_reality/4_elements/composite_character_1_clothing_1_style_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 73,\n \"path\": \"output_reality/4_elements/merge_character_1_clothing_1_background_1_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"4. object (Umbrella): transparent umbrella\"\n ]\n },\n {\n \"index\": 74,\n \"path\": \"output_reality/4_elements/switch_character_1_clothing_1_background_1_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"4. object (Umbrella): transparent umbrella\"\n ]\n },\n {\n \"index\": 75,\n \"path\": \"output_reality/4_elements/composite_character_1_clothing_1_background_1_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"4. object (Umbrella): transparent umbrella\"\n ]\n },\n {\n \"index\": 76,\n \"path\": \"output_reality/4_elements/merge_character_2_clothing_1_style_2_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Forest Background): slg, river, forest\"\n ]\n },\n {\n \"index\": 77,\n \"path\": \"output_reality/4_elements/switch_character_2_clothing_1_style_2_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Forest Background): slg, river, forest\"\n ]\n },\n {\n \"index\": 78,\n \"path\": \"output_reality/4_elements/composite_character_2_clothing_1_style_2_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Forest Background): slg, river, forest\"\n ]\n },\n {\n \"index\": 79,\n \"path\": \"output_reality/4_elements/merge_character_3_clothing_1_style_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 80,\n \"path\": \"output_reality/4_elements/switch_character_3_clothing_1_style_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 81,\n \"path\": \"output_reality/4_elements/composite_character_3_clothing_1_style_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 82,\n \"path\": \"output_reality/3_elements/merge_character_1_clothing_2_style_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\"\n ]\n },\n {\n \"index\": 83,\n \"path\": \"output_reality/3_elements/switch_character_1_clothing_2_style_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\"\n ]\n },\n {\n \"index\": 84,\n \"path\": \"output_reality/3_elements/composite_character_1_clothing_2_style_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\"\n ]\n },\n {\n \"index\": 85,\n \"path\": \"output_reality/3_elements/merge_character_2_clothing_1_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 86,\n \"path\": \"output_reality/3_elements/switch_character_2_clothing_1_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 87,\n \"path\": \"output_reality/3_elements/composite_character_2_clothing_1_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 88,\n \"path\": \"output_reality/5_elements/merge_character_3_clothing_1_style_2_background_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 89,\n \"path\": \"output_reality/5_elements/switch_character_3_clothing_1_style_2_background_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 90,\n \"path\": \"output_reality/5_elements/composite_character_3_clothing_1_style_2_background_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 91,\n \"path\": \"output_reality/3_elements/merge_character_1_clothing_2_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 92,\n \"path\": \"output_reality/3_elements/switch_character_1_clothing_2_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 93,\n \"path\": \"output_reality/3_elements/composite_character_1_clothing_2_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 94,\n \"path\": \"output_reality/3_elements/merge_character_2_clothing_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 95,\n \"path\": \"output_reality/3_elements/switch_character_2_clothing_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 96,\n \"path\": \"output_reality/3_elements/composite_character_2_clothing_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 97,\n \"path\": \"output_reality/3_elements/merge_character_1_background_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 98,\n \"path\": \"output_reality/3_elements/switch_character_1_background_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 99,\n \"path\": \"output_reality/3_elements/composite_character_1_background_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 100,\n \"path\": \"output_reality/5_elements/merge_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Forest Background): slg, river, forest\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 101,\n \"path\": \"output_reality/5_elements/switch_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Forest Background): slg, river, forest\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 102,\n \"path\": \"output_reality/5_elements/composite_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Forest Background): slg, river, forest\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 103,\n \"path\": \"output_reality/3_elements/merge_character_1_style_1_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 104,\n \"path\": \"output_reality/3_elements/switch_character_1_style_1_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 105,\n \"path\": \"output_reality/3_elements/composite_character_1_style_1_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 106,\n \"path\": \"output_reality/5_elements/merge_character_1_clothing_2_style_2_background_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 107,\n \"path\": \"output_reality/5_elements/switch_character_1_clothing_2_style_2_background_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 108,\n \"path\": \"output_reality/5_elements/composite_character_1_clothing_2_style_2_background_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 109,\n \"path\": \"output_reality/4_elements/merge_character_1_style_2_background_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 110,\n \"path\": \"output_reality/4_elements/switch_character_1_style_2_background_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 111,\n \"path\": \"output_reality/4_elements/composite_character_1_style_2_background_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ]\n },\n {\n \"index\": 112,\n \"path\": \"output_reality/4_elements/merge_character_2_clothing_2_style_2_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 113,\n \"path\": \"output_reality/4_elements/switch_character_2_clothing_2_style_2_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 114,\n \"path\": \"output_reality/4_elements/composite_character_2_clothing_2_style_2_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ]\n },\n {\n \"index\": 115,\n \"path\": \"output_reality/3_elements/merge_character_1_style_2_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\"\n ]\n },\n {\n \"index\": 116,\n \"path\": \"output_reality/3_elements/switch_character_1_style_2_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\"\n ]\n },\n {\n \"index\": 117,\n \"path\": \"output_reality/3_elements/composite_character_1_style_2_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\"\n ]\n },\n {\n \"index\": 118,\n \"path\": \"output_reality/2_elements/merge_character_3_style_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\"\n ]\n },\n {\n \"index\": 119,\n \"path\": \"output_reality/2_elements/switch_character_3_style_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\"\n ]\n },\n {\n \"index\": 120,\n \"path\": \"output_reality/2_elements/composite_character_3_style_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\"\n ]\n }\n]" }, { "path": "human_eval/results.json", "content": "[\n {\n \"index\": 1,\n \"path\": \"output_anime/2_elements/merge_character_1_style_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 34.57\n },\n \"gpt4v\": {\n \"composition\": 7.75,\n \"image\": 9.0\n }\n },\n {\n \"index\": 2,\n \"path\": \"output_anime/2_elements/switch_character_1_style_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\"\n ],\n \"human_1\": {\n \"composition\": 2.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 36.1757\n },\n \"gpt4v\": {\n \"composition\": 6.0,\n \"image\": 8.5\n }\n },\n {\n \"index\": 3,\n \"path\": \"output_anime/2_elements/composite_character_1_style_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 36.2511\n },\n \"gpt4v\": {\n \"composition\": 8.75,\n \"image\": 9.25\n }\n },\n {\n \"index\": 4,\n \"path\": \"output_anime/4_elements/merge_character_2_clothing_1_style_2_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 3.0\n },\n \"clipscore\": {\n \"score\": 34.0178\n },\n \"gpt4v\": {\n \"composition\": 7.5,\n \"image\": 8.875\n }\n },\n {\n \"index\": 5,\n \"path\": \"output_anime/4_elements/switch_character_2_clothing_1_style_2_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 33.3053\n },\n \"gpt4v\": {\n \"composition\": 7.75,\n \"image\": 9.25\n }\n },\n {\n \"index\": 6,\n \"path\": \"output_anime/4_elements/composite_character_2_clothing_1_style_2_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 31.7461\n },\n \"gpt4v\": {\n \"composition\": 8.25,\n \"image\": 9.5\n }\n },\n {\n \"index\": 7,\n \"path\": \"output_anime/5_elements/merge_character_3_clothing_1_style_1_background_1_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 1.0,\n \"image\": 1.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 1.5,\n \"image\": 1.0\n },\n \"clipscore\": {\n \"score\": 38.0065\n },\n \"gpt4v\": {\n \"composition\": 8.0,\n \"image\": 9.5\n }\n },\n {\n \"index\": 8,\n \"path\": \"output_anime/5_elements/switch_character_3_clothing_1_style_1_background_1_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"clipscore\": {\n \"score\": 40.1243\n },\n \"gpt4v\": {\n \"composition\": 8.75,\n \"image\": 9.75\n }\n },\n {\n \"index\": 9,\n \"path\": \"output_anime/5_elements/composite_character_3_clothing_1_style_1_background_1_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 39.4737\n },\n \"gpt4v\": {\n \"composition\": 8.0,\n \"image\": 9.75\n }\n },\n {\n \"index\": 10,\n \"path\": \"output_anime/3_elements/merge_character_3_clothing_2_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 2.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 2.5\n },\n \"clipscore\": {\n \"score\": 37.5053\n },\n \"gpt4v\": {\n \"composition\": 8.625,\n \"image\": 9.25\n }\n },\n {\n \"index\": 11,\n \"path\": \"output_anime/3_elements/switch_character_3_clothing_2_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 35.5424\n },\n \"gpt4v\": {\n \"composition\": 7.5,\n \"image\": 9.0\n }\n },\n {\n \"index\": 12,\n \"path\": \"output_anime/3_elements/composite_character_3_clothing_2_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 38.1935\n },\n \"gpt4v\": {\n \"composition\": 9.75,\n \"image\": 9.5\n }\n },\n {\n \"index\": 13,\n \"path\": \"output_anime/5_elements/merge_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 33.5435\n },\n \"gpt4v\": {\n \"composition\": 6.5,\n \"image\": 9.75\n }\n },\n {\n \"index\": 14,\n \"path\": \"output_anime/5_elements/switch_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 36.0033\n },\n \"gpt4v\": {\n \"composition\": 7.75,\n \"image\": 9.75\n }\n },\n {\n \"index\": 15,\n \"path\": \"output_anime/5_elements/composite_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 32.1798\n },\n \"gpt4v\": {\n \"composition\": 8.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 16,\n \"path\": \"output_anime/5_elements/merge_character_2_clothing_2_style_1_background_1_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 2.0,\n \"image\": 2.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 2.5\n },\n \"clipscore\": {\n \"score\": 29.6089\n },\n \"gpt4v\": {\n \"composition\": 7.5,\n \"image\": 9.75\n }\n },\n {\n \"index\": 17,\n \"path\": \"output_anime/5_elements/switch_character_2_clothing_2_style_1_background_1_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 34.6888\n },\n \"gpt4v\": {\n \"composition\": 9.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 18,\n \"path\": \"output_anime/5_elements/composite_character_2_clothing_2_style_1_background_1_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 29.4748\n },\n \"gpt4v\": {\n \"composition\": 9.0,\n \"image\": 9.75\n }\n },\n {\n \"index\": 19,\n \"path\": \"output_anime/3_elements/merge_character_2_clothing_2_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 2.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 2.5\n },\n \"clipscore\": {\n \"score\": 38.4343\n },\n \"gpt4v\": {\n \"composition\": 8.625,\n \"image\": 9.375\n }\n },\n {\n \"index\": 20,\n \"path\": \"output_anime/3_elements/switch_character_2_clothing_2_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 36.5482\n },\n \"gpt4v\": {\n \"composition\": 8.25,\n \"image\": 10.0\n }\n },\n {\n \"index\": 21,\n \"path\": \"output_anime/3_elements/composite_character_2_clothing_2_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 35.3934\n },\n \"gpt4v\": {\n \"composition\": 8.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 22,\n \"path\": \"output_anime/4_elements/merge_character_1_clothing_2_background_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"clipscore\": {\n \"score\": 40.0848\n },\n \"gpt4v\": {\n \"composition\": 8.125,\n \"image\": 9.875\n }\n },\n {\n \"index\": 23,\n \"path\": \"output_anime/4_elements/switch_character_1_clothing_2_background_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 36.855\n },\n \"gpt4v\": {\n \"composition\": 7.75,\n \"image\": 10.0\n }\n },\n {\n \"index\": 24,\n \"path\": \"output_anime/4_elements/composite_character_1_clothing_2_background_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 36.9936\n },\n \"gpt4v\": {\n \"composition\": 6.5,\n \"image\": 9.75\n }\n },\n {\n \"index\": 25,\n \"path\": \"output_anime/4_elements/merge_character_1_clothing_2_background_2_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 1.0,\n \"image\": 1.0\n },\n \"human_2\": {\n \"composition\": 1.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 1.0,\n \"image\": 1.0\n },\n \"clipscore\": {\n \"score\": 36.696\n },\n \"gpt4v\": {\n \"composition\": 7.5,\n \"image\": 8.625\n }\n },\n {\n \"index\": 26,\n \"path\": \"output_anime/4_elements/switch_character_1_clothing_2_background_2_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 43.1276\n },\n \"gpt4v\": {\n \"composition\": 8.25,\n \"image\": 10.0\n }\n },\n {\n \"index\": 27,\n \"path\": \"output_anime/4_elements/composite_character_1_clothing_2_background_2_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Auroral Background): auroral, starry sky, outdoors\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 42.7789\n },\n \"gpt4v\": {\n \"composition\": 7.25,\n \"image\": 9.5\n }\n },\n {\n \"index\": 28,\n \"path\": \"output_anime/4_elements/merge_character_2_clothing_2_style_2_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 2.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 2.0,\n \"image\": 2.0\n },\n \"clipscore\": {\n \"score\": 30.5346\n },\n \"gpt4v\": {\n \"composition\": 4.25,\n \"image\": 9.125\n }\n },\n {\n \"index\": 29,\n \"path\": \"output_anime/4_elements/switch_character_2_clothing_2_style_2_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 36.1002\n },\n \"gpt4v\": {\n \"composition\": 4.25,\n \"image\": 9.5\n }\n },\n {\n \"index\": 30,\n \"path\": \"output_anime/4_elements/composite_character_2_clothing_2_style_2_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 34.4661\n },\n \"gpt4v\": {\n \"composition\": 5.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 31,\n \"path\": \"output_anime/3_elements/merge_character_3_style_2_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 2.0\n },\n \"clipscore\": {\n \"score\": 35.0985\n },\n \"gpt4v\": {\n \"composition\": 7.75,\n \"image\": 8.875\n }\n },\n {\n \"index\": 32,\n \"path\": \"output_anime/3_elements/switch_character_3_style_2_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 34.5847\n },\n \"gpt4v\": {\n \"composition\": 5.0,\n \"image\": 9.0\n }\n },\n {\n \"index\": 33,\n \"path\": \"output_anime/3_elements/composite_character_3_style_2_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 38.6079\n },\n \"gpt4v\": {\n \"composition\": 6.75,\n \"image\": 9.0\n }\n },\n {\n \"index\": 34,\n \"path\": \"output_anime/4_elements/merge_character_1_clothing_1_style_2_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\"\n ],\n \"human_1\": {\n \"composition\": 1.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 2.0\n },\n \"avg_human_score\": {\n \"composition\": 1.5,\n \"image\": 2.5\n },\n \"clipscore\": {\n \"score\": 33.2259\n },\n \"gpt4v\": {\n \"composition\": 4.375,\n \"image\": 8.25\n }\n },\n {\n \"index\": 35,\n \"path\": \"output_anime/4_elements/switch_character_1_clothing_1_style_2_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 37.2583\n },\n \"gpt4v\": {\n \"composition\": 7.75,\n \"image\": 9.0\n }\n },\n {\n \"index\": 36,\n \"path\": \"output_anime/4_elements/composite_character_1_clothing_1_style_2_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\"\n ],\n \"human_1\": {\n \"composition\": 2.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 34.5118\n },\n \"gpt4v\": {\n \"composition\": 5.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 37,\n \"path\": \"output_anime/4_elements/merge_character_3_clothing_2_background_1_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 2.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 2.0,\n \"image\": 2.5\n },\n \"clipscore\": {\n \"score\": 34.9788\n },\n \"gpt4v\": {\n \"composition\": 6.0,\n \"image\": 9.25\n }\n },\n {\n \"index\": 38,\n \"path\": \"output_anime/4_elements/switch_character_3_clothing_2_background_1_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 2.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"clipscore\": {\n \"score\": 34.8866\n },\n \"gpt4v\": {\n \"composition\": 5.5,\n \"image\": 8.5\n }\n },\n {\n \"index\": 39,\n \"path\": \"output_anime/4_elements/composite_character_3_clothing_2_background_1_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 2.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 36.994\n },\n \"gpt4v\": {\n \"composition\": 7.25,\n \"image\": 9.75\n }\n },\n {\n \"index\": 40,\n \"path\": \"output_anime/4_elements/merge_character_1_clothing_1_style_2_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 1.0,\n \"image\": 2.0\n },\n \"human_2\": {\n \"composition\": 1.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 1.0,\n \"image\": 1.5\n },\n \"clipscore\": {\n \"score\": 29.6\n },\n \"gpt4v\": {\n \"composition\": 4.875,\n \"image\": 9.125\n }\n },\n {\n \"index\": 41,\n \"path\": \"output_anime/4_elements/switch_character_1_clothing_1_style_2_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 38.3351\n },\n \"gpt4v\": {\n \"composition\": 6.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 42,\n \"path\": \"output_anime/4_elements/composite_character_1_clothing_1_style_2_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Garreg Mach Monastery Uniform): gmuniform, blue thighhighs, long sleeves\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 38.7396\n },\n \"gpt4v\": {\n \"composition\": 6.0,\n \"image\": 9.5\n }\n },\n {\n \"index\": 43,\n \"path\": \"output_anime/4_elements/merge_character_3_clothing_2_style_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 33.0407\n },\n \"gpt4v\": {\n \"composition\": 8.25,\n \"image\": 9.125\n }\n },\n {\n \"index\": 44,\n \"path\": \"output_anime/4_elements/switch_character_3_clothing_2_style_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 32.3786\n },\n \"gpt4v\": {\n \"composition\": 6.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 45,\n \"path\": \"output_anime/4_elements/composite_character_3_clothing_2_style_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 32.9488\n },\n \"gpt4v\": {\n \"composition\": 7.0,\n \"image\": 9.5\n }\n },\n {\n \"index\": 46,\n \"path\": \"output_anime/5_elements/merge_character_3_clothing_2_style_1_background_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 2.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 2.0\n },\n \"avg_human_score\": {\n \"composition\": 2.0,\n \"image\": 3.0\n },\n \"clipscore\": {\n \"score\": 36.2348\n },\n \"gpt4v\": {\n \"composition\": 7.125,\n \"image\": 8.75\n }\n },\n {\n \"index\": 47,\n \"path\": \"output_anime/5_elements/switch_character_3_clothing_2_style_1_background_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 2.0\n },\n \"clipscore\": {\n \"score\": 34.9156\n },\n \"gpt4v\": {\n \"composition\": 5.5,\n \"image\": 9.0\n }\n },\n {\n \"index\": 48,\n \"path\": \"output_anime/5_elements/composite_character_3_clothing_2_style_1_background_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Son Goku): son goku, spiked hair, muscular male, wristband\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"5. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 33.1048\n },\n \"gpt4v\": {\n \"composition\": 6.25,\n \"image\": 9.5\n }\n },\n {\n \"index\": 49,\n \"path\": \"output_anime/4_elements/merge_character_1_style_2_background_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"clipscore\": {\n \"score\": 36.6768\n },\n \"gpt4v\": {\n \"composition\": 8.25,\n \"image\": 9.625\n }\n },\n {\n \"index\": 50,\n \"path\": \"output_anime/4_elements/switch_character_1_style_2_background_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 40.792\n },\n \"gpt4v\": {\n \"composition\": 9.75,\n \"image\": 9.75\n }\n },\n {\n \"index\": 51,\n \"path\": \"output_anime/4_elements/composite_character_1_style_2_background_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"3. background (Bamboolight Background): bamboolight, outdoors, bamboo\",\n \"4. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 38.5719\n },\n \"gpt4v\": {\n \"composition\": 8.75,\n \"image\": 9.5\n }\n },\n {\n \"index\": 52,\n \"path\": \"output_anime/5_elements/merge_character_2_clothing_2_style_2_background_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 2.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 2.0,\n \"image\": 2.0\n },\n \"clipscore\": {\n \"score\": 29.867\n },\n \"gpt4v\": {\n \"composition\": 4.375,\n \"image\": 9.375\n }\n },\n {\n \"index\": 53,\n \"path\": \"output_anime/5_elements/switch_character_2_clothing_2_style_2_background_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 34.3121\n },\n \"gpt4v\": {\n \"composition\": 6.0,\n \"image\": 9.25\n }\n },\n {\n \"index\": 54,\n \"path\": \"output_anime/5_elements/composite_character_2_clothing_2_style_2_background_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. background (Auroral Background): auroral, starry sky, outdoors\",\n \"5. object (Toast): toast, toast in mouth\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 31.6278\n },\n \"gpt4v\": {\n \"composition\": 5.75,\n \"image\": 9.75\n }\n },\n {\n \"index\": 55,\n \"path\": \"output_anime/4_elements/merge_character_1_clothing_2_style_2_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 2.0\n },\n \"clipscore\": {\n \"score\": 40.51\n },\n \"gpt4v\": {\n \"composition\": 6.75,\n \"image\": 9.0\n }\n },\n {\n \"index\": 56,\n \"path\": \"output_anime/4_elements/switch_character_1_clothing_2_style_2_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 2.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 37.5231\n },\n \"gpt4v\": {\n \"composition\": 5.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 57,\n \"path\": \"output_anime/4_elements/composite_character_1_clothing_2_style_2_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Kamado Nezuko): kamado nezuko, black hair, pink eyes, forehead\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Shuimobysim (Chinese Ink Wash Style)): shuimobysim, traditional chinese ink painting\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 37.4707\n },\n \"gpt4v\": {\n \"composition\": 6.0,\n \"image\": 9.0\n }\n },\n {\n \"index\": 58,\n \"path\": \"output_anime/4_elements/merge_character_2_clothing_2_style_1_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 2.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 3.0\n },\n \"clipscore\": {\n \"score\": 30.7663\n },\n \"gpt4v\": {\n \"composition\": 6.375,\n \"image\": 9.0\n }\n },\n {\n \"index\": 59,\n \"path\": \"output_anime/4_elements/switch_character_2_clothing_2_style_1_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 30.6962\n },\n \"gpt4v\": {\n \"composition\": 6.75,\n \"image\": 9.25\n }\n },\n {\n \"index\": 60,\n \"path\": \"output_anime/4_elements/composite_character_2_clothing_2_style_1_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Texas the Omertosa in Arknights): omertosa, 1girl, wolf ears, long hair\",\n \"2. clothing (Zero Suit (Metroid)): zero suit, blue gloves, high heels\",\n \"3. style (Hand-drawn Style): lineart, hand-drawn style\",\n \"4. object (Huge Two-Handed Burger): two-handed burger, holding a huge burger with both hands\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 36.6772\n },\n \"gpt4v\": {\n \"composition\": 6.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 61,\n \"path\": \"output_reality/3_elements/merge_character_3_clothing_2_style_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 31.7792\n },\n \"gpt4v\": {\n \"composition\": 8.125,\n \"image\": 9.0\n }\n },\n {\n \"index\": 62,\n \"path\": \"output_reality/3_elements/switch_character_3_clothing_2_style_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 34.0626\n },\n \"gpt4v\": {\n \"composition\": 7.5,\n \"image\": 9.0\n }\n },\n {\n \"index\": 63,\n \"path\": \"output_reality/3_elements/composite_character_3_clothing_2_style_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 33.965\n },\n \"gpt4v\": {\n \"composition\": 7.5,\n \"image\": 9.0\n }\n },\n {\n \"index\": 64,\n \"path\": \"output_reality/2_elements/merge_character_1_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Forest Background): slg, river, forest\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 3.0\n },\n \"clipscore\": {\n \"score\": 26.6335\n },\n \"gpt4v\": {\n \"composition\": 8.25,\n \"image\": 9.75\n }\n },\n {\n \"index\": 65,\n \"path\": \"output_reality/2_elements/switch_character_1_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Forest Background): slg, river, forest\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 30.5736\n },\n \"gpt4v\": {\n \"composition\": 7.0,\n \"image\": 10.0\n }\n },\n {\n \"index\": 66,\n \"path\": \"output_reality/2_elements/composite_character_1_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Forest Background): slg, river, forest\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 30.1499\n },\n \"gpt4v\": {\n \"composition\": 7.0,\n \"image\": 9.5\n }\n },\n {\n \"index\": 67,\n \"path\": \"output_reality/4_elements/merge_character_2_clothing_2_style_1_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 37.2627\n },\n \"gpt4v\": {\n \"composition\": 7.0,\n \"image\": 9.125\n }\n },\n {\n \"index\": 68,\n \"path\": \"output_reality/4_elements/switch_character_2_clothing_2_style_1_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 38.5944\n },\n \"gpt4v\": {\n \"composition\": 7.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 69,\n \"path\": \"output_reality/4_elements/composite_character_2_clothing_2_style_1_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 35.552\n },\n \"gpt4v\": {\n \"composition\": 6.75,\n \"image\": 9.5\n }\n },\n {\n \"index\": 70,\n \"path\": \"output_reality/4_elements/merge_character_1_clothing_1_style_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 2.5\n },\n \"clipscore\": {\n \"score\": 29.8164\n },\n \"gpt4v\": {\n \"composition\": 7.875,\n \"image\": 9.25\n }\n },\n {\n \"index\": 71,\n \"path\": \"output_reality/4_elements/switch_character_1_clothing_1_style_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 37.9293\n },\n \"gpt4v\": {\n \"composition\": 6.0,\n \"image\": 9.5\n }\n },\n {\n \"index\": 72,\n \"path\": \"output_reality/4_elements/composite_character_1_clothing_1_style_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 36.2537\n },\n \"gpt4v\": {\n \"composition\": 5.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 73,\n \"path\": \"output_reality/4_elements/merge_character_1_clothing_1_background_1_object_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"4. object (Umbrella): transparent umbrella\"\n ],\n \"human_1\": {\n \"composition\": 2.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 2.0,\n \"image\": 2.5\n },\n \"clipscore\": {\n \"score\": 36.3731\n },\n \"gpt4v\": {\n \"composition\": 2.25,\n \"image\": 9.25\n }\n },\n {\n \"index\": 74,\n \"path\": \"output_reality/4_elements/switch_character_1_clothing_1_background_1_object_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"4. object (Umbrella): transparent umbrella\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 39.3619\n },\n \"gpt4v\": {\n \"composition\": 9.0,\n \"image\": 10.0\n }\n },\n {\n \"index\": 75,\n \"path\": \"output_reality/4_elements/composite_character_1_clothing_1_background_1_object_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"4. object (Umbrella): transparent umbrella\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 36.4537\n },\n \"gpt4v\": {\n \"composition\": 5.0,\n \"image\": 9.5\n }\n },\n {\n \"index\": 76,\n \"path\": \"output_reality/4_elements/merge_character_2_clothing_1_style_2_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Forest Background): slg, river, forest\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 2.0\n },\n \"avg_human_score\": {\n \"composition\": 3.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 35.4899\n },\n \"gpt4v\": {\n \"composition\": 9.0,\n \"image\": 9.375\n }\n },\n {\n \"index\": 77,\n \"path\": \"output_reality/4_elements/switch_character_2_clothing_1_style_2_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Forest Background): slg, river, forest\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 38.2703\n },\n \"gpt4v\": {\n \"composition\": 9.5,\n \"image\": 10.0\n }\n },\n {\n \"index\": 78,\n \"path\": \"output_reality/4_elements/composite_character_2_clothing_1_style_2_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Forest Background): slg, river, forest\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 36.787\n },\n \"gpt4v\": {\n \"composition\": 8.5,\n \"image\": 9.75\n }\n },\n {\n \"index\": 79,\n \"path\": \"output_reality/4_elements/merge_character_3_clothing_1_style_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 2.5\n },\n \"clipscore\": {\n \"score\": 31.6602\n },\n \"gpt4v\": {\n \"composition\": 6.375,\n \"image\": 9.625\n }\n },\n {\n \"index\": 80,\n \"path\": \"output_reality/4_elements/switch_character_3_clothing_1_style_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 36.4865\n },\n \"gpt4v\": {\n \"composition\": 8.0,\n \"image\": 9.5\n }\n },\n {\n \"index\": 81,\n \"path\": \"output_reality/4_elements/composite_character_3_clothing_1_style_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 34.3751\n },\n \"gpt4v\": {\n \"composition\": 6.5,\n \"image\": 9.75\n }\n },\n {\n \"index\": 82,\n \"path\": \"output_reality/3_elements/merge_character_1_clothing_2_style_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 34.7694\n },\n \"gpt4v\": {\n \"composition\": 9.0,\n \"image\": 9.625\n }\n },\n {\n \"index\": 83,\n \"path\": \"output_reality/3_elements/switch_character_1_clothing_2_style_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 37.2876\n },\n \"gpt4v\": {\n \"composition\": 8.0,\n \"image\": 9.75\n }\n },\n {\n \"index\": 84,\n \"path\": \"output_reality/3_elements/composite_character_1_clothing_2_style_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 35.6261\n },\n \"gpt4v\": {\n \"composition\": 8.5,\n \"image\": 9.0\n }\n },\n {\n \"index\": 85,\n \"path\": \"output_reality/3_elements/merge_character_2_clothing_1_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 36.0037\n },\n \"gpt4v\": {\n \"composition\": 9.0,\n \"image\": 9.75\n }\n },\n {\n \"index\": 86,\n \"path\": \"output_reality/3_elements/switch_character_2_clothing_1_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 36.511\n },\n \"gpt4v\": {\n \"composition\": 9.0,\n \"image\": 10.0\n }\n },\n {\n \"index\": 87,\n \"path\": \"output_reality/3_elements/composite_character_2_clothing_1_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 39.224\n },\n \"gpt4v\": {\n \"composition\": 5.0,\n \"image\": 9.5\n }\n },\n {\n \"index\": 88,\n \"path\": \"output_reality/5_elements/merge_character_3_clothing_1_style_2_background_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 2.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 3.0\n },\n \"clipscore\": {\n \"score\": 28.4802\n },\n \"gpt4v\": {\n \"composition\": 1.75,\n \"image\": 9.5\n }\n },\n {\n \"index\": 89,\n \"path\": \"output_reality/5_elements/switch_character_3_clothing_1_style_2_background_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 35.9777\n },\n \"gpt4v\": {\n \"composition\": 7.0,\n \"image\": 10.0\n }\n },\n {\n \"index\": 90,\n \"path\": \"output_reality/5_elements/composite_character_3_clothing_1_style_2_background_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 36.1203\n },\n \"gpt4v\": {\n \"composition\": 6.5,\n \"image\": 10.0\n }\n },\n {\n \"index\": 91,\n \"path\": \"output_reality/3_elements/merge_character_1_clothing_2_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 37.029\n },\n \"gpt4v\": {\n \"composition\": 7.75,\n \"image\": 9.375\n }\n },\n {\n \"index\": 92,\n \"path\": \"output_reality/3_elements/switch_character_1_clothing_2_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 38.7682\n },\n \"gpt4v\": {\n \"composition\": 9.0,\n \"image\": 10.0\n }\n },\n {\n \"index\": 93,\n \"path\": \"output_reality/3_elements/composite_character_1_clothing_2_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 37.642\n },\n \"gpt4v\": {\n \"composition\": 8.75,\n \"image\": 10.0\n }\n },\n {\n \"index\": 94,\n \"path\": \"output_reality/3_elements/merge_character_2_clothing_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 37.2169\n },\n \"gpt4v\": {\n \"composition\": 8.875,\n \"image\": 9.875\n }\n },\n {\n \"index\": 95,\n \"path\": \"output_reality/3_elements/switch_character_2_clothing_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 36.5079\n },\n \"gpt4v\": {\n \"composition\": 7.75,\n \"image\": 10.0\n }\n },\n {\n \"index\": 96,\n \"path\": \"output_reality/3_elements/composite_character_2_clothing_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 38.2253\n },\n \"gpt4v\": {\n \"composition\": 8.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 97,\n \"path\": \"output_reality/3_elements/merge_character_1_background_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 30.5479\n },\n \"gpt4v\": {\n \"composition\": 8.625,\n \"image\": 9.25\n }\n },\n {\n \"index\": 98,\n \"path\": \"output_reality/3_elements/switch_character_1_background_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 33.8878\n },\n \"gpt4v\": {\n \"composition\": 8.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 99,\n \"path\": \"output_reality/3_elements/composite_character_1_background_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"3. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 31.293\n },\n \"gpt4v\": {\n \"composition\": 6.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 100,\n \"path\": \"output_reality/5_elements/merge_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Forest Background): slg, river, forest\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 2.0\n },\n \"clipscore\": {\n \"score\": 29.54\n },\n \"gpt4v\": {\n \"composition\": 4.125,\n \"image\": 9.25\n }\n },\n {\n \"index\": 101,\n \"path\": \"output_reality/5_elements/switch_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Forest Background): slg, river, forest\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 30.1944\n },\n \"gpt4v\": {\n \"composition\": 5.0,\n \"image\": 9.25\n }\n },\n {\n \"index\": 102,\n \"path\": \"output_reality/5_elements/composite_character_3_clothing_1_style_1_background_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. clothing (Thai University Uniform): mahalaiuniform, white shirt short sleeves, black pencil skirt\",\n \"3. style (Japanese Film Color Style): film overlay, film grain\",\n \"4. background (Forest Background): slg, river, forest\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 30.875\n },\n \"gpt4v\": {\n \"composition\": 4.75,\n \"image\": 9.5\n }\n },\n {\n \"index\": 103,\n \"path\": \"output_reality/3_elements/merge_character_1_style_1_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 29.2703\n },\n \"gpt4v\": {\n \"composition\": 6.875,\n \"image\": 9.0\n }\n },\n {\n \"index\": 104,\n \"path\": \"output_reality/3_elements/switch_character_1_style_1_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 31.1808\n },\n \"gpt4v\": {\n \"composition\": 6.0,\n \"image\": 9.0\n }\n },\n {\n \"index\": 105,\n \"path\": \"output_reality/3_elements/composite_character_1_style_1_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\",\n \"3. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 31.265\n },\n \"gpt4v\": {\n \"composition\": 6.0,\n \"image\": 9.5\n }\n },\n {\n \"index\": 106,\n \"path\": \"output_reality/5_elements/merge_character_1_clothing_2_style_2_background_1_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 2.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 2.0\n },\n \"avg_human_score\": {\n \"composition\": 2.0,\n \"image\": 2.5\n },\n \"clipscore\": {\n \"score\": 35.6569\n },\n \"gpt4v\": {\n \"composition\": 7.5,\n \"image\": 8.0\n }\n },\n {\n \"index\": 107,\n \"path\": \"output_reality/5_elements/switch_character_1_clothing_2_style_2_background_1_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 3.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 38.4301\n },\n \"gpt4v\": {\n \"composition\": 5.75,\n \"image\": 10.0\n }\n },\n {\n \"index\": 108,\n \"path\": \"output_reality/5_elements/composite_character_1_clothing_2_style_2_background_1_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 5,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\",\n \"5. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 34.751\n },\n \"gpt4v\": {\n \"composition\": 6.0,\n \"image\": 9.5\n }\n },\n {\n \"index\": 109,\n \"path\": \"output_reality/4_elements/merge_character_1_style_2_background_2_object_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 3.0,\n \"image\": 3.0\n },\n \"human_2\": {\n \"composition\": 2.0,\n \"image\": 1.0\n },\n \"avg_human_score\": {\n \"composition\": 2.5,\n \"image\": 2.0\n },\n \"clipscore\": {\n \"score\": 25.1638\n },\n \"gpt4v\": {\n \"composition\": 6.625,\n \"image\": 7.75\n }\n },\n {\n \"index\": 110,\n \"path\": \"output_reality/4_elements/switch_character_1_style_2_background_2_object_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 28.4197\n },\n \"gpt4v\": {\n \"composition\": 8.25,\n \"image\": 9.25\n }\n },\n {\n \"index\": 111,\n \"path\": \"output_reality/4_elements/composite_character_1_style_2_background_2_object_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\",\n \"4. object (Bubble Gum): blow bubble gum\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 4.5,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 26.9113\n },\n \"gpt4v\": {\n \"composition\": 8.75,\n \"image\": 9.5\n }\n },\n {\n \"index\": 112,\n \"path\": \"output_reality/4_elements/merge_character_2_clothing_2_style_2_background_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 36.9875\n },\n \"gpt4v\": {\n \"composition\": 9.0,\n \"image\": 9.625\n }\n },\n {\n \"index\": 113,\n \"path\": \"output_reality/4_elements/switch_character_2_clothing_2_style_2_background_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 38.9775\n },\n \"gpt4v\": {\n \"composition\": 8.5,\n \"image\": 9.5\n }\n },\n {\n \"index\": 114,\n \"path\": \"output_reality/4_elements/composite_character_2_clothing_2_style_2_background_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 4,\n \"prompt\": [\n \"1. character (Scarlett Johansson): scarlett, short red hair, blue eyes\",\n \"2. clothing (School Dress): school uniform, white shirt, red tie, blue pleated microskirt\",\n \"3. style (Bright Style): bright lighting\",\n \"4. background (Library Bookshelf Background): lib_bg, library bookshelf\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 40.0983\n },\n \"gpt4v\": {\n \"composition\": 9.0,\n \"image\": 10.0\n }\n },\n {\n \"index\": 115,\n \"path\": \"output_reality/3_elements/merge_character_1_style_2_background_2.png\",\n \"method\": \"merge\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 3.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 3.5\n },\n \"clipscore\": {\n \"score\": 26.566\n },\n \"gpt4v\": {\n \"composition\": 7.75,\n \"image\": 9.875\n }\n },\n {\n \"index\": 116,\n \"path\": \"output_reality/3_elements/switch_character_1_style_2_background_2.png\",\n \"method\": \"switch\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\"\n ],\n \"human_1\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 30.3636\n },\n \"gpt4v\": {\n \"composition\": 4.5,\n \"image\": 10.0\n }\n },\n {\n \"index\": 117,\n \"path\": \"output_reality/3_elements/composite_character_1_style_2_background_2.png\",\n \"method\": \"composite\",\n \"number of elements\": 3,\n \"prompt\": [\n \"1. character (IU (Lee Ji Eun, Korean singer)): iu1, long straight black hair, hazel eyes, diamond stud earrings\",\n \"2. style (Bright Style): bright lighting\",\n \"3. background (Forest Background): slg, river, forest\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 3.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 4.0,\n \"image\": 4.5\n },\n \"clipscore\": {\n \"score\": 27.9653\n },\n \"gpt4v\": {\n \"composition\": 7.5,\n \"image\": 10.0\n }\n },\n {\n \"index\": 118,\n \"path\": \"output_reality/2_elements/merge_character_3_style_1.png\",\n \"method\": \"merge\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 4.0\n },\n \"clipscore\": {\n \"score\": 27.6391\n },\n \"gpt4v\": {\n \"composition\": 6.5,\n \"image\": 10.0\n }\n },\n {\n \"index\": 119,\n \"path\": \"output_reality/2_elements/switch_character_3_style_1.png\",\n \"method\": \"switch\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 25.042\n },\n \"gpt4v\": {\n \"composition\": 6.5,\n \"image\": 10.0\n }\n },\n {\n \"index\": 120,\n \"path\": \"output_reality/2_elements/composite_character_3_style_1.png\",\n \"method\": \"composite\",\n \"number of elements\": 2,\n \"prompt\": [\n \"1. character (The Rock (Dwayne Johnson)): th3r0ck with no hair, muscular male, serious look on his face\",\n \"2. style (Japanese Film Color Style): film overlay, film grain\"\n ],\n \"human_1\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"human_2\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"avg_human_score\": {\n \"composition\": 5.0,\n \"image\": 5.0\n },\n \"clipscore\": {\n \"score\": 24.2756\n },\n \"gpt4v\": {\n \"composition\": 7.0,\n \"image\": 10.0\n }\n }\n]" }, { "path": "lcm_example.py", "content": "import torch\nimport argparse\nfrom diffusers import DiffusionPipeline, StableDiffusionPipeline, AutoencoderKL\nfrom diffusers import LCMScheduler\nfrom callbacks import make_callback\n\ndef get_example_prompt():\n prompt = \"RAW photo, subject, 8k uhd, dslr, high quality, Fujifilm XT3, half-length portrait from knees up, scarlett, short red hair, blue eyes, school uniform, white shirt, red tie, blue pleated microskirt\"\n negative_prompt = \"extra heads, nsfw, deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, text, cropped, out of frame, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck\"\n return prompt, negative_prompt\n\ndef main(args):\n\n # set the prompts for image generation\n prompt, negative_prompt = get_example_prompt()\n\n # base model for the realistic style example\n model_name = 'SimianLuo/LCM_Dreamshaper_v7'\n\n # set base model\n pipeline = DiffusionPipeline.from_pretrained(\n model_name,\n custom_pipeline=\"./pipelines/sd1.5_0.26.3\",\n use_safetensors=True,\n safety_checker=None,\n requires_safety_checker=False\n ).to(\"cuda\")\n\n # set scheduler\n pipeline.scheduler = LCMScheduler.from_config(pipeline.scheduler.config)\n\n # initialize LoRAs\n # This example shows the composition of a character LoRA and a clothing LoRA\n pipeline.load_lora_weights(args.lora_path, weight_name=\"character_2.safetensors\", adapter_name=\"character\")\n pipeline.load_lora_weights(args.lora_path, weight_name=\"clothing_2.safetensors\", adapter_name=\"clothing\")\n cur_loras = [\"character\", \"clothing\"]\n\n # select the method for the composition\n if args.method == \"merge\":\n pipeline.set_adapters(cur_loras)\n switch_callback = None\n elif args.method == \"switch\":\n pipeline.set_adapters([cur_loras[0]])\n switch_callback = make_callback(switch_step=args.switch_step, loras=cur_loras)\n else:\n pipeline.set_adapters(cur_loras)\n switch_callback = None\n\n image = pipeline(\n prompt=prompt, \n negative_prompt=negative_prompt,\n height=args.height,\n width=args.width,\n num_inference_steps=args.denoise_steps,\n guidance_scale=args.cfg_scale,\n generator=args.generator,\n cross_attention_kwargs={\"scale\": args.lora_scale},\n callback_on_step_end=switch_callback,\n lora_composite=True if args.method == \"composite\" else False,\n ).images[0]\n\n image.save(args.save_path)\n\nif __name__ == \"__main__\":\n parser = argparse.ArgumentParser(\n description='Example code for multi-LoRA composition'\n )\n\n # Arguments for composing LoRAs\n parser.add_argument('--method', default='switch',\n choices=['merge', 'switch', 'composite'],\n help='methods for combining LoRAs', type=str)\n parser.add_argument('--save_path', default='example.png',\n help='path to save the generated image', type=str)\n parser.add_argument('--lora_path', default='models/lora/reality',\n help='path to store all LoRAs', type=str)\n parser.add_argument('--lora_scale', default=0.9,\n help='scale of each LoRA when generating images', type=float)\n parser.add_argument('--switch_step', default=2,\n help='number of steps to switch LoRA during denoising, applicable only in the switch method', type=int)\n\n # Arguments for generating images\n parser.add_argument('--height', default=1024,\n help='height of the generated images', type=int)\n parser.add_argument('--width', default=768,\n help='width of the generated images', type=int)\n parser.add_argument('--denoise_steps', default=8,\n help='number of the denoising steps', type=int)\n parser.add_argument('--cfg_scale', default=7.0,\n help='scale for classifier-free guidance', type=float)\n parser.add_argument('--seed', default=11,\n help='seed for generating images', type=int)\n\n args = parser.parse_args()\n args.generator = torch.manual_seed(args.seed)\n\n main(args)" }, { "path": "lcm_lora_example.py", "content": "import os\nimport torch\nimport argparse\nfrom diffusers import DiffusionPipeline, StableDiffusionPipeline, AutoencoderKL\nfrom diffusers import LCMScheduler\nfrom callbacks import make_callback\n\ndef get_example_prompt():\n prompt = \"RAW photo, subject, 8k uhd, dslr, high quality, Fujifilm XT3, half-length portrait from knees up, scarlett, short red hair, blue eyes, school uniform, white shirt, red tie, blue pleated microskirt\"\n negative_prompt = \"extra heads, nsfw, deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, text, cropped, out of frame, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck\"\n return prompt, negative_prompt\n\ndef main(args):\n\n # set the prompts for image generation\n prompt, negative_prompt = get_example_prompt()\n\n # base model for the realistic style example\n model_name = 'models/Realistic-LCM-LoRA'\n\n if not os.path.exists(model_name):\n # set base model\n pipeline = DiffusionPipeline.from_pretrained(\n 'SG161222/Realistic_Vision_V5.1_noVAE',\n custom_pipeline=\"./pipelines/sd1.5_0.26.3\",\n use_safetensors=True,\n safety_checker=None,\n requires_safety_checker=False\n ).to(\"cuda\")\n\n # set vae\n vae = AutoencoderKL.from_pretrained(\n \"stabilityai/sd-vae-ft-mse\",\n ).to(\"cuda\")\n pipeline.vae = vae\n\n # set scheduler\n pipeline.scheduler = LCMScheduler.from_config(pipeline.scheduler.config)\n\n # It seems that there is a bug when composing multiple LoRAs after fusing.\n # Therefore, we first save the models before proceeding with the composition.\n pipeline.load_lora_weights(\"latent-consistency/lcm-lora-sdv1-5\")\n pipeline.fuse_lora()\n pipeline.unload_lora_weights()\n pipeline.save_pretrained(\"models/Realistic-LCM-LoRA\")\n\n # set base model\n pipeline = DiffusionPipeline.from_pretrained(\n model_name,\n custom_pipeline=\"./pipelines/sd1.5_0.26.3\",\n use_safetensors=True,\n safety_checker=None,\n requires_safety_checker=False\n ).to(\"cuda\")\n\n # set scheduler\n pipeline.scheduler = LCMScheduler.from_config(pipeline.scheduler.config)\n\n # initialize LoRAs\n # This example shows the composition of a character LoRA and a clothing LoRA\n pipeline.load_lora_weights(args.lora_path, weight_name=\"character_2.safetensors\", adapter_name=\"character\")\n pipeline.load_lora_weights(args.lora_path, weight_name=\"clothing_2.safetensors\", adapter_name=\"clothing\")\n cur_loras = [\"character\", \"clothing\"]\n\n # select the method for the composition\n if args.method == \"merge\":\n pipeline.set_adapters(cur_loras)\n switch_callback = None\n elif args.method == \"switch\":\n pipeline.set_adapters([cur_loras[0]])\n switch_callback = make_callback(switch_step=args.switch_step, loras=cur_loras)\n else:\n pipeline.set_adapters(cur_loras)\n switch_callback = None\n\n image = pipeline(\n prompt=prompt, \n negative_prompt=negative_prompt,\n height=args.height,\n width=args.width,\n num_inference_steps=args.denoise_steps,\n guidance_scale=args.cfg_scale,\n generator=args.generator,\n cross_attention_kwargs={\"scale\": args.lora_scale},\n callback_on_step_end=switch_callback,\n lora_composite=True if args.method == \"composite\" else False\n ).images[0]\n\n image.save(args.save_path)\n\nif __name__ == \"__main__\":\n parser = argparse.ArgumentParser(\n description='Example code for multi-LoRA composition'\n )\n\n # Arguments for composing LoRAs\n parser.add_argument('--method', default='switch',\n choices=['merge', 'switch', 'composite'],\n help='methods for combining LoRAs', type=str)\n parser.add_argument('--save_path', default='example.png',\n help='path to save the generated image', type=str)\n parser.add_argument('--lora_path', default='models/lora/reality',\n help='path to store all LoRAs', type=str)\n parser.add_argument('--lora_scale', default=0.8,\n help='scale of each LoRA when generating images', type=float)\n parser.add_argument('--switch_step', default=2,\n help='number of steps to switch LoRA during denoising, applicable only in the switch method', type=int)\n\n # Arguments for generating images\n parser.add_argument('--height', default=1024,\n help='height of the generated images', type=int)\n parser.add_argument('--width', default=768,\n help='width of the generated images', type=int)\n parser.add_argument('--denoise_steps', default=8,\n help='number of the denoising steps', type=int)\n parser.add_argument('--cfg_scale', default=1.8,\n help='scale for classifier-free guidance', type=float)\n parser.add_argument('--seed', default=42,\n help='seed for generating images', type=int)\n\n args = parser.parse_args()\n args.generator = torch.manual_seed(args.seed)\n\n main(args)" }, { "path": "models/README.md", "content": "# Models in ComposLoRA\nOur **ComposLoRA** testbed comprises 22 pre-trained LoRAs, spanning characters, colors, styles, backgrounds, and objects. Download `ComposLoRA.zip` from [this link](https://drive.google.com/file/d/1SuwRgV1LtEud8dfjftnw-zxBMgzSCwIT/view?usp=sharing), and unzip it in this folder.\n" }, { "path": "pipelines/sd1.5_0.26.3/pipeline.py", "content": "# Copyright 2023 The HuggingFace Team. All rights reserved.\n#\n# Licensed under the Apache License, Version 2.0 (the \"License\");\n# you may not use this file except in compliance with the License.\n# You may obtain a copy of the License at\n#\n# http://www.apache.org/licenses/LICENSE-2.0\n#\n# Unless required by applicable law or agreed to in writing, software\n# distributed under the License is distributed on an \"AS IS\" BASIS,\n# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.\n\n# for diffusers version 0.26.3\n\nimport inspect\nfrom typing import Any, Callable, Dict, List, Optional, Union\n\nimport torch\nfrom packaging import version\nfrom transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection\n\nfrom diffusers.configuration_utils import FrozenDict\nfrom diffusers.image_processor import PipelineImageInput, VaeImageProcessor\nfrom diffusers.loaders import FromSingleFileMixin, IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin\nfrom diffusers.models import AutoencoderKL, ImageProjection, UNet2DConditionModel\nfrom diffusers.models.attention_processor import FusedAttnProcessor2_0\nfrom diffusers.models.lora import adjust_lora_scale_text_encoder\nfrom diffusers.schedulers import KarrasDiffusionSchedulers\nfrom diffusers.utils import (\n USE_PEFT_BACKEND,\n deprecate,\n logging,\n replace_example_docstring,\n scale_lora_layers,\n unscale_lora_layers,\n)\nfrom diffusers.utils.torch_utils import randn_tensor\nfrom diffusers.pipelines.pipeline_utils import DiffusionPipeline\nfrom diffusers.pipelines.stable_diffusion.pipeline_output import StableDiffusionPipelineOutput\nfrom diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker\n\n\nlogger = logging.get_logger(__name__) # pylint: disable=invalid-name\n\nEXAMPLE_DOC_STRING = \"\"\"\n Examples:\n ```py\n >>> import torch\n >>> from diffusers import StableDiffusionPipeline\n\n >>> pipe = StableDiffusionPipeline.from_pretrained(\"runwayml/stable-diffusion-v1-5\", torch_dtype=torch.float16)\n >>> pipe = pipe.to(\"cuda\")\n\n >>> prompt = \"a photo of an astronaut riding a horse on mars\"\n >>> image = pipe(prompt).images[0]\n ```\n\"\"\"\n\n\ndef rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):\n \"\"\"\n Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and\n Sample Steps are Flawed](https://arxiv.org/pdf/2305.08891.pdf). See Section 3.4\n \"\"\"\n std_text = noise_pred_text.std(dim=list(range(1, noise_pred_text.ndim)), keepdim=True)\n std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True)\n # rescale the results from guidance (fixes overexposure)\n noise_pred_rescaled = noise_cfg * (std_text / std_cfg)\n # mix with the original results from guidance by factor guidance_rescale to avoid \"plain looking\" images\n noise_cfg = guidance_rescale * noise_pred_rescaled + (1 - guidance_rescale) * noise_cfg\n return noise_cfg\n\n\ndef retrieve_timesteps(\n scheduler,\n num_inference_steps: Optional[int] = None,\n device: Optional[Union[str, torch.device]] = None,\n timesteps: Optional[List[int]] = None,\n **kwargs,\n):\n \"\"\"\n Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles\n custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.\n\n Args:\n scheduler (`SchedulerMixin`):\n The scheduler to get timesteps from.\n num_inference_steps (`int`):\n The number of diffusion steps used when generating samples with a pre-trained model. If used,\n `timesteps` must be `None`.\n device (`str` or `torch.device`, *optional*):\n The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.\n timesteps (`List[int]`, *optional*):\n Custom timesteps used to support arbitrary spacing between timesteps. If `None`, then the default\n timestep spacing strategy of the scheduler is used. If `timesteps` is passed, `num_inference_steps`\n must be `None`.\n\n Returns:\n `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the\n second element is the number of inference steps.\n \"\"\"\n if timesteps is not None:\n accepts_timesteps = \"timesteps\" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())\n if not accepts_timesteps:\n raise ValueError(\n f\"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom\"\n f\" timestep schedules. Please check whether you are using the correct scheduler.\"\n )\n scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)\n timesteps = scheduler.timesteps\n num_inference_steps = len(timesteps)\n else:\n scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)\n timesteps = scheduler.timesteps\n return timesteps, num_inference_steps\n\n\nclass StableDiffusionPipeline(\n DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin, IPAdapterMixin, FromSingleFileMixin\n):\n r\"\"\"\n Pipeline for text-to-image generation using Stable Diffusion.\n\n This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods\n implemented for all pipelines (downloading, saving, running on a particular device, etc.).\n\n The pipeline also inherits the following loading methods:\n - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings\n - [`~loaders.LoraLoaderMixin.load_lora_weights`] for loading LoRA weights\n - [`~loaders.LoraLoaderMixin.save_lora_weights`] for saving LoRA weights\n - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files\n - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters\n\n Args:\n vae ([`AutoencoderKL`]):\n Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations.\n text_encoder ([`~transformers.CLIPTextModel`]):\n Frozen text-encoder ([clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14)).\n tokenizer ([`~transformers.CLIPTokenizer`]):\n A `CLIPTokenizer` to tokenize text.\n unet ([`UNet2DConditionModel`]):\n A `UNet2DConditionModel` to denoise the encoded image latents.\n scheduler ([`SchedulerMixin`]):\n A scheduler to be used in combination with `unet` to denoise the encoded image latents. Can be one of\n [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].\n safety_checker ([`StableDiffusionSafetyChecker`]):\n Classification module that estimates whether generated images could be considered offensive or harmful.\n Please refer to the [model card](https://huggingface.co/runwayml/stable-diffusion-v1-5) for more details\n about a model's potential harms.\n feature_extractor ([`~transformers.CLIPImageProcessor`]):\n A `CLIPImageProcessor` to extract features from generated images; used as inputs to the `safety_checker`.\n \"\"\"\n\n model_cpu_offload_seq = \"text_encoder->image_encoder->unet->vae\"\n _optional_components = [\"safety_checker\", \"feature_extractor\", \"image_encoder\"]\n _exclude_from_cpu_offload = [\"safety_checker\"]\n _callback_tensor_inputs = [\"latents\", \"prompt_embeds\", \"negative_prompt_embeds\"]\n\n def __init__(\n self,\n vae: AutoencoderKL,\n text_encoder: CLIPTextModel,\n tokenizer: CLIPTokenizer,\n unet: UNet2DConditionModel,\n scheduler: KarrasDiffusionSchedulers,\n safety_checker: StableDiffusionSafetyChecker,\n feature_extractor: CLIPImageProcessor,\n image_encoder: CLIPVisionModelWithProjection = None,\n requires_safety_checker: bool = True,\n ):\n super().__init__()\n\n if hasattr(scheduler.config, \"steps_offset\") and scheduler.config.steps_offset != 1:\n deprecation_message = (\n f\"The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`\"\n f\" should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure \"\n \"to update the config accordingly as leaving `steps_offset` might led to incorrect results\"\n \" in future versions. If you have downloaded this checkpoint from the Hugging Face Hub,\"\n \" it would be very nice if you could open a Pull request for the `scheduler/scheduler_config.json`\"\n \" file\"\n )\n deprecate(\"steps_offset!=1\", \"1.0.0\", deprecation_message, standard_warn=False)\n new_config = dict(scheduler.config)\n new_config[\"steps_offset\"] = 1\n scheduler._internal_dict = FrozenDict(new_config)\n\n if hasattr(scheduler.config, \"clip_sample\") and scheduler.config.clip_sample is True:\n deprecation_message = (\n f\"The configuration file of this scheduler: {scheduler} has not set the configuration `clip_sample`.\"\n \" `clip_sample` should be set to False in the configuration file. Please make sure to update the\"\n \" config accordingly as not setting `clip_sample` in the config might lead to incorrect results in\"\n \" future versions. If you have downloaded this checkpoint from the Hugging Face Hub, it would be very\"\n \" nice if you could open a Pull request for the `scheduler/scheduler_config.json` file\"\n )\n deprecate(\"clip_sample not set\", \"1.0.0\", deprecation_message, standard_warn=False)\n new_config = dict(scheduler.config)\n new_config[\"clip_sample\"] = False\n scheduler._internal_dict = FrozenDict(new_config)\n\n if safety_checker is None and requires_safety_checker:\n logger.warning(\n f\"You have disabled the safety checker for {self.__class__} by passing `safety_checker=None`. Ensure\"\n \" that you abide to the conditions of the Stable Diffusion license and do not expose unfiltered\"\n \" results in services or applications open to the public. Both the diffusers team and Hugging Face\"\n \" strongly recommend to keep the safety filter enabled in all public facing circumstances, disabling\"\n \" it only for use-cases that involve analyzing network behavior or auditing its results. For more\"\n \" information, please have a look at https://github.com/huggingface/diffusers/pull/254 .\"\n )\n\n if safety_checker is not None and feature_extractor is None:\n raise ValueError(\n \"Make sure to define a feature extractor when loading {self.__class__} if you want to use the safety\"\n \" checker. If you do not want to use the safety checker, you can pass `'safety_checker=None'` instead.\"\n )\n\n is_unet_version_less_0_9_0 = hasattr(unet.config, \"_diffusers_version\") and version.parse(\n version.parse(unet.config._diffusers_version).base_version\n ) < version.parse(\"0.9.0.dev0\")\n is_unet_sample_size_less_64 = hasattr(unet.config, \"sample_size\") and unet.config.sample_size < 64\n if is_unet_version_less_0_9_0 and is_unet_sample_size_less_64:\n deprecation_message = (\n \"The configuration file of the unet has set the default `sample_size` to smaller than\"\n \" 64 which seems highly unlikely. If your checkpoint is a fine-tuned version of any of the\"\n \" following: \\n- CompVis/stable-diffusion-v1-4 \\n- CompVis/stable-diffusion-v1-3 \\n-\"\n \" CompVis/stable-diffusion-v1-2 \\n- CompVis/stable-diffusion-v1-1 \\n- runwayml/stable-diffusion-v1-5\"\n \" \\n- runwayml/stable-diffusion-inpainting \\n you should change 'sample_size' to 64 in the\"\n \" configuration file. Please make sure to update the config accordingly as leaving `sample_size=32`\"\n \" in the config might lead to incorrect results in future versions. If you have downloaded this\"\n \" checkpoint from the Hugging Face Hub, it would be very nice if you could open a Pull request for\"\n \" the `unet/config.json` file\"\n )\n deprecate(\"sample_size<64\", \"1.0.0\", deprecation_message, standard_warn=False)\n new_config = dict(unet.config)\n new_config[\"sample_size\"] = 64\n unet._internal_dict = FrozenDict(new_config)\n\n self.register_modules(\n vae=vae,\n text_encoder=text_encoder,\n tokenizer=tokenizer,\n unet=unet,\n scheduler=scheduler,\n safety_checker=safety_checker,\n feature_extractor=feature_extractor,\n image_encoder=image_encoder,\n )\n self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)\n self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)\n self.register_to_config(requires_safety_checker=requires_safety_checker)\n\n def enable_vae_slicing(self):\n r\"\"\"\n Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to\n compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.\n \"\"\"\n self.vae.enable_slicing()\n\n def disable_vae_slicing(self):\n r\"\"\"\n Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to\n computing decoding in one step.\n \"\"\"\n self.vae.disable_slicing()\n\n def enable_vae_tiling(self):\n r\"\"\"\n Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to\n compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow\n processing larger images.\n \"\"\"\n self.vae.enable_tiling()\n\n def disable_vae_tiling(self):\n r\"\"\"\n Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to\n computing decoding in one step.\n \"\"\"\n self.vae.disable_tiling()\n\n def _encode_prompt(\n self,\n prompt,\n device,\n num_images_per_prompt,\n do_classifier_free_guidance,\n negative_prompt=None,\n prompt_embeds: Optional[torch.FloatTensor] = None,\n negative_prompt_embeds: Optional[torch.FloatTensor] = None,\n lora_scale: Optional[float] = None,\n **kwargs,\n ):\n deprecation_message = \"`_encode_prompt()` is deprecated and it will be removed in a future version. Use `encode_prompt()` instead. Also, be aware that the output format changed from a concatenated tensor to a tuple.\"\n deprecate(\"_encode_prompt()\", \"1.0.0\", deprecation_message, standard_warn=False)\n\n prompt_embeds_tuple = self.encode_prompt(\n prompt=prompt,\n device=device,\n num_images_per_prompt=num_images_per_prompt,\n do_classifier_free_guidance=do_classifier_free_guidance,\n negative_prompt=negative_prompt,\n prompt_embeds=prompt_embeds,\n negative_prompt_embeds=negative_prompt_embeds,\n lora_scale=lora_scale,\n **kwargs,\n )\n\n # concatenate for backwards comp\n prompt_embeds = torch.cat([prompt_embeds_tuple[1], prompt_embeds_tuple[0]])\n\n return prompt_embeds\n\n def encode_prompt(\n self,\n prompt,\n device,\n num_images_per_prompt,\n do_classifier_free_guidance,\n negative_prompt=None,\n prompt_embeds: Optional[torch.FloatTensor] = None,\n negative_prompt_embeds: Optional[torch.FloatTensor] = None,\n lora_scale: Optional[float] = None,\n clip_skip: Optional[int] = None,\n ):\n r\"\"\"\n Encodes the prompt into text encoder hidden states.\n\n Args:\n prompt (`str` or `List[str]`, *optional*):\n prompt to be encoded\n device: (`torch.device`):\n torch device\n num_images_per_prompt (`int`):\n number of images that should be generated per prompt\n do_classifier_free_guidance (`bool`):\n whether to use classifier free guidance or not\n negative_prompt (`str` or `List[str]`, *optional*):\n The prompt or prompts not to guide the image generation. If not defined, one has to pass\n `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is\n less than `1`).\n prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not\n provided, text embeddings will be generated from `prompt` input argument.\n negative_prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt\n weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input\n argument.\n lora_scale (`float`, *optional*):\n A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.\n clip_skip (`int`, *optional*):\n Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that\n the output of the pre-final layer will be used for computing the prompt embeddings.\n \"\"\"\n # set lora scale so that monkey patched LoRA\n # function of text encoder can correctly access it\n if lora_scale is not None and isinstance(self, LoraLoaderMixin):\n self._lora_scale = lora_scale\n\n # dynamically adjust the LoRA scale\n if not USE_PEFT_BACKEND:\n adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)\n else:\n scale_lora_layers(self.text_encoder, lora_scale)\n\n if prompt is not None and isinstance(prompt, str):\n batch_size = 1\n elif prompt is not None and isinstance(prompt, list):\n batch_size = len(prompt)\n else:\n batch_size = prompt_embeds.shape[0]\n\n if prompt_embeds is None:\n # textual inversion: procecss multi-vector tokens if necessary\n if isinstance(self, TextualInversionLoaderMixin):\n prompt = self.maybe_convert_prompt(prompt, self.tokenizer)\n\n text_inputs = self.tokenizer(\n prompt,\n padding=\"max_length\",\n max_length=self.tokenizer.model_max_length,\n truncation=True,\n return_tensors=\"pt\",\n )\n text_input_ids = text_inputs.input_ids\n untruncated_ids = self.tokenizer(prompt, padding=\"longest\", return_tensors=\"pt\").input_ids\n\n if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(\n text_input_ids, untruncated_ids\n ):\n removed_text = self.tokenizer.batch_decode(\n untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1]\n )\n logger.warning(\n \"The following part of your input was truncated because CLIP can only handle sequences up to\"\n f\" {self.tokenizer.model_max_length} tokens: {removed_text}\"\n )\n\n if hasattr(self.text_encoder.config, \"use_attention_mask\") and self.text_encoder.config.use_attention_mask:\n attention_mask = text_inputs.attention_mask.to(device)\n else:\n attention_mask = None\n\n if clip_skip is None:\n prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)\n prompt_embeds = prompt_embeds[0]\n else:\n prompt_embeds = self.text_encoder(\n text_input_ids.to(device), attention_mask=attention_mask, output_hidden_states=True\n )\n # Access the `hidden_states` first, that contains a tuple of\n # all the hidden states from the encoder layers. Then index into\n # the tuple to access the hidden states from the desired layer.\n prompt_embeds = prompt_embeds[-1][-(clip_skip + 1)]\n # We also need to apply the final LayerNorm here to not mess with the\n # representations. The `last_hidden_states` that we typically use for\n # obtaining the final prompt representations passes through the LayerNorm\n # layer.\n prompt_embeds = self.text_encoder.text_model.final_layer_norm(prompt_embeds)\n\n if self.text_encoder is not None:\n prompt_embeds_dtype = self.text_encoder.dtype\n elif self.unet is not None:\n prompt_embeds_dtype = self.unet.dtype\n else:\n prompt_embeds_dtype = prompt_embeds.dtype\n\n prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)\n\n bs_embed, seq_len, _ = prompt_embeds.shape\n # duplicate text embeddings for each generation per prompt, using mps friendly method\n prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)\n prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)\n\n # get unconditional embeddings for classifier free guidance\n if do_classifier_free_guidance and negative_prompt_embeds is None:\n uncond_tokens: List[str]\n if negative_prompt is None:\n uncond_tokens = [\"\"] * batch_size\n elif prompt is not None and type(prompt) is not type(negative_prompt):\n raise TypeError(\n f\"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !=\"\n f\" {type(prompt)}.\"\n )\n elif isinstance(negative_prompt, str):\n uncond_tokens = [negative_prompt]\n elif batch_size != len(negative_prompt):\n raise ValueError(\n f\"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:\"\n f\" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches\"\n \" the batch size of `prompt`.\"\n )\n else:\n uncond_tokens = negative_prompt\n\n # textual inversion: procecss multi-vector tokens if necessary\n if isinstance(self, TextualInversionLoaderMixin):\n uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.tokenizer)\n\n max_length = prompt_embeds.shape[1]\n uncond_input = self.tokenizer(\n uncond_tokens,\n padding=\"max_length\",\n max_length=max_length,\n truncation=True,\n return_tensors=\"pt\",\n )\n\n if hasattr(self.text_encoder.config, \"use_attention_mask\") and self.text_encoder.config.use_attention_mask:\n attention_mask = uncond_input.attention_mask.to(device)\n else:\n attention_mask = None\n\n negative_prompt_embeds = self.text_encoder(\n uncond_input.input_ids.to(device),\n attention_mask=attention_mask,\n )\n negative_prompt_embeds = negative_prompt_embeds[0]\n\n if do_classifier_free_guidance:\n # duplicate unconditional embeddings for each generation per prompt, using mps friendly method\n seq_len = negative_prompt_embeds.shape[1]\n\n negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)\n\n negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)\n negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)\n\n if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND:\n # Retrieve the original scale by scaling back the LoRA layers\n unscale_lora_layers(self.text_encoder, lora_scale)\n\n return prompt_embeds, negative_prompt_embeds\n\n def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):\n dtype = next(self.image_encoder.parameters()).dtype\n\n if not isinstance(image, torch.Tensor):\n image = self.feature_extractor(image, return_tensors=\"pt\").pixel_values\n\n image = image.to(device=device, dtype=dtype)\n if output_hidden_states:\n image_enc_hidden_states = self.image_encoder(image, output_hidden_states=True).hidden_states[-2]\n image_enc_hidden_states = image_enc_hidden_states.repeat_interleave(num_images_per_prompt, dim=0)\n uncond_image_enc_hidden_states = self.image_encoder(\n torch.zeros_like(image), output_hidden_states=True\n ).hidden_states[-2]\n uncond_image_enc_hidden_states = uncond_image_enc_hidden_states.repeat_interleave(\n num_images_per_prompt, dim=0\n )\n return image_enc_hidden_states, uncond_image_enc_hidden_states\n else:\n image_embeds = self.image_encoder(image).image_embeds\n image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)\n uncond_image_embeds = torch.zeros_like(image_embeds)\n\n return image_embeds, uncond_image_embeds\n\n def prepare_ip_adapter_image_embeds(self, ip_adapter_image, device, num_images_per_prompt):\n if not isinstance(ip_adapter_image, list):\n ip_adapter_image = [ip_adapter_image]\n\n if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):\n raise ValueError(\n f\"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters.\"\n )\n\n image_embeds = []\n for single_ip_adapter_image, image_proj_layer in zip(\n ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers\n ):\n output_hidden_state = not isinstance(image_proj_layer, ImageProjection)\n single_image_embeds, single_negative_image_embeds = self.encode_image(\n single_ip_adapter_image, device, 1, output_hidden_state\n )\n single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)\n single_negative_image_embeds = torch.stack([single_negative_image_embeds] * num_images_per_prompt, dim=0)\n\n if self.do_classifier_free_guidance:\n single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])\n single_image_embeds = single_image_embeds.to(device)\n\n image_embeds.append(single_image_embeds)\n\n return image_embeds\n\n def run_safety_checker(self, image, device, dtype):\n if self.safety_checker is None:\n has_nsfw_concept = None\n else:\n if torch.is_tensor(image):\n feature_extractor_input = self.image_processor.postprocess(image, output_type=\"pil\")\n else:\n feature_extractor_input = self.image_processor.numpy_to_pil(image)\n safety_checker_input = self.feature_extractor(feature_extractor_input, return_tensors=\"pt\").to(device)\n image, has_nsfw_concept = self.safety_checker(\n images=image, clip_input=safety_checker_input.pixel_values.to(dtype)\n )\n return image, has_nsfw_concept\n\n def decode_latents(self, latents):\n deprecation_message = \"The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead\"\n deprecate(\"decode_latents\", \"1.0.0\", deprecation_message, standard_warn=False)\n\n latents = 1 / self.vae.config.scaling_factor * latents\n image = self.vae.decode(latents, return_dict=False)[0]\n image = (image / 2 + 0.5).clamp(0, 1)\n # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16\n image = image.cpu().permute(0, 2, 3, 1).float().numpy()\n return image\n\n def prepare_extra_step_kwargs(self, generator, eta):\n # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature\n # eta (ฮท) is only used with the DDIMScheduler, it will be ignored for other schedulers.\n # eta corresponds to ฮท in DDIM paper: https://arxiv.org/abs/2010.02502\n # and should be between [0, 1]\n\n accepts_eta = \"eta\" in set(inspect.signature(self.scheduler.step).parameters.keys())\n extra_step_kwargs = {}\n if accepts_eta:\n extra_step_kwargs[\"eta\"] = eta\n\n # check if the scheduler accepts generator\n accepts_generator = \"generator\" in set(inspect.signature(self.scheduler.step).parameters.keys())\n if accepts_generator:\n extra_step_kwargs[\"generator\"] = generator\n return extra_step_kwargs\n\n def check_inputs(\n self,\n prompt,\n height,\n width,\n callback_steps,\n negative_prompt=None,\n prompt_embeds=None,\n negative_prompt_embeds=None,\n callback_on_step_end_tensor_inputs=None,\n ):\n if height % 8 != 0 or width % 8 != 0:\n raise ValueError(f\"`height` and `width` have to be divisible by 8 but are {height} and {width}.\")\n\n if callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0):\n raise ValueError(\n f\"`callback_steps` has to be a positive integer but is {callback_steps} of type\"\n f\" {type(callback_steps)}.\"\n )\n if callback_on_step_end_tensor_inputs is not None and not all(\n k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs\n ):\n raise ValueError(\n f\"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}\"\n )\n\n if prompt is not None and prompt_embeds is not None:\n raise ValueError(\n f\"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to\"\n \" only forward one of the two.\"\n )\n elif prompt is None and prompt_embeds is None:\n raise ValueError(\n \"Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined.\"\n )\n elif prompt is not None and (not isinstance(prompt, str) and not isinstance(prompt, list)):\n raise ValueError(f\"`prompt` has to be of type `str` or `list` but is {type(prompt)}\")\n\n if negative_prompt is not None and negative_prompt_embeds is not None:\n raise ValueError(\n f\"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:\"\n f\" {negative_prompt_embeds}. Please make sure to only forward one of the two.\"\n )\n\n if prompt_embeds is not None and negative_prompt_embeds is not None:\n if prompt_embeds.shape != negative_prompt_embeds.shape:\n raise ValueError(\n \"`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but\"\n f\" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`\"\n f\" {negative_prompt_embeds.shape}.\"\n )\n\n def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):\n shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)\n if isinstance(generator, list) and len(generator) != batch_size:\n raise ValueError(\n f\"You have passed a list of generators of length {len(generator)}, but requested an effective batch\"\n f\" size of {batch_size}. Make sure the batch size matches the length of the generators.\"\n )\n\n if latents is None:\n latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)\n else:\n latents = latents.to(device)\n\n # scale the initial noise by the standard deviation required by the scheduler\n latents = latents * self.scheduler.init_noise_sigma\n return latents\n\n def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):\n r\"\"\"Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.\n\n The suffixes after the scaling factors represent the stages where they are being applied.\n\n Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values\n that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.\n\n Args:\n s1 (`float`):\n Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to\n mitigate \"oversmoothing effect\" in the enhanced denoising process.\n s2 (`float`):\n Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to\n mitigate \"oversmoothing effect\" in the enhanced denoising process.\n b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.\n b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.\n \"\"\"\n if not hasattr(self, \"unet\"):\n raise ValueError(\"The pipeline must have `unet` for using FreeU.\")\n self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)\n\n def disable_freeu(self):\n \"\"\"Disables the FreeU mechanism if enabled.\"\"\"\n self.unet.disable_freeu()\n\n # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.fuse_qkv_projections\n def fuse_qkv_projections(self, unet: bool = True, vae: bool = True):\n \"\"\"\n Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query,\n key, value) are fused. For cross-attention modules, key and value projection matrices are fused.\n\n \n\n This API is ๐Ÿงช experimental.\n\n \n\n Args:\n unet (`bool`, defaults to `True`): To apply fusion on the UNet.\n vae (`bool`, defaults to `True`): To apply fusion on the VAE.\n \"\"\"\n self.fusing_unet = False\n self.fusing_vae = False\n\n if unet:\n self.fusing_unet = True\n self.unet.fuse_qkv_projections()\n self.unet.set_attn_processor(FusedAttnProcessor2_0())\n\n if vae:\n if not isinstance(self.vae, AutoencoderKL):\n raise ValueError(\"`fuse_qkv_projections()` is only supported for the VAE of type `AutoencoderKL`.\")\n\n self.fusing_vae = True\n self.vae.fuse_qkv_projections()\n self.vae.set_attn_processor(FusedAttnProcessor2_0())\n\n # Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.unfuse_qkv_projections\n def unfuse_qkv_projections(self, unet: bool = True, vae: bool = True):\n \"\"\"Disable QKV projection fusion if enabled.\n\n \n\n This API is ๐Ÿงช experimental.\n\n \n\n Args:\n unet (`bool`, defaults to `True`): To apply fusion on the UNet.\n vae (`bool`, defaults to `True`): To apply fusion on the VAE.\n\n \"\"\"\n if unet:\n if not self.fusing_unet:\n logger.warning(\"The UNet was not initially fused for QKV projections. Doing nothing.\")\n else:\n self.unet.unfuse_qkv_projections()\n self.fusing_unet = False\n\n if vae:\n if not self.fusing_vae:\n logger.warning(\"The VAE was not initially fused for QKV projections. Doing nothing.\")\n else:\n self.vae.unfuse_qkv_projections()\n self.fusing_vae = False\n\n # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding\n def get_guidance_scale_embedding(self, w, embedding_dim=512, dtype=torch.float32):\n \"\"\"\n See https://github.com/google-research/vdm/blob/dc27b98a554f65cdc654b800da5aa1846545d41b/model_vdm.py#L298\n\n Args:\n timesteps (`torch.Tensor`):\n generate embedding vectors at these timesteps\n embedding_dim (`int`, *optional*, defaults to 512):\n dimension of the embeddings to generate\n dtype:\n data type of the generated embeddings\n\n Returns:\n `torch.FloatTensor`: Embedding vectors with shape `(len(timesteps), embedding_dim)`\n \"\"\"\n assert len(w.shape) == 1\n w = w * 1000.0\n\n half_dim = embedding_dim // 2\n emb = torch.log(torch.tensor(10000.0)) / (half_dim - 1)\n emb = torch.exp(torch.arange(half_dim, dtype=dtype) * -emb)\n emb = w.to(dtype)[:, None] * emb[None, :]\n emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)\n if embedding_dim % 2 == 1: # zero pad\n emb = torch.nn.functional.pad(emb, (0, 1))\n assert emb.shape == (w.shape[0], embedding_dim)\n return emb\n\n @property\n def guidance_scale(self):\n return self._guidance_scale\n\n @property\n def guidance_rescale(self):\n return self._guidance_rescale\n\n @property\n def clip_skip(self):\n return self._clip_skip\n\n # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)\n # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`\n # corresponds to doing no classifier free guidance.\n @property\n def do_classifier_free_guidance(self):\n return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None\n\n @property\n def cross_attention_kwargs(self):\n return self._cross_attention_kwargs\n\n @property\n def num_timesteps(self):\n return self._num_timesteps\n\n @property\n def interrupt(self):\n return self._interrupt\n\n @torch.no_grad()\n @replace_example_docstring(EXAMPLE_DOC_STRING)\n def __call__(\n self,\n prompt: Union[str, List[str]] = None,\n height: Optional[int] = None,\n width: Optional[int] = None,\n num_inference_steps: int = 50,\n timesteps: List[int] = None,\n guidance_scale: float = 7.5,\n negative_prompt: Optional[Union[str, List[str]]] = None,\n num_images_per_prompt: Optional[int] = 1,\n eta: float = 0.0,\n generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,\n latents: Optional[torch.FloatTensor] = None,\n prompt_embeds: Optional[torch.FloatTensor] = None,\n negative_prompt_embeds: Optional[torch.FloatTensor] = None,\n ip_adapter_image: Optional[PipelineImageInput] = None,\n output_type: Optional[str] = \"pil\",\n return_dict: bool = True,\n cross_attention_kwargs: Optional[Dict[str, Any]] = None,\n guidance_rescale: float = 0.0,\n clip_skip: Optional[int] = None,\n callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,\n callback_on_step_end_tensor_inputs: List[str] = [\"latents\"],\n lora_composite: bool = False, \n **kwargs,\n ):\n r\"\"\"\n The call function to the pipeline for generation.\n\n Args:\n prompt (`str` or `List[str]`, *optional*):\n The prompt or prompts to guide image generation. If not defined, you need to pass `prompt_embeds`.\n height (`int`, *optional*, defaults to `self.unet.config.sample_size * self.vae_scale_factor`):\n The height in pixels of the generated image.\n width (`int`, *optional*, defaults to `self.unet.config.sample_size * self.vae_scale_factor`):\n The width in pixels of the generated image.\n num_inference_steps (`int`, *optional*, defaults to 50):\n The number of denoising steps. More denoising steps usually lead to a higher quality image at the\n expense of slower inference.\n timesteps (`List[int]`, *optional*):\n Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument\n in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is\n passed will be used. Must be in descending order.\n guidance_scale (`float`, *optional*, defaults to 7.5):\n A higher guidance scale value encourages the model to generate images closely linked to the text\n `prompt` at the expense of lower image quality. Guidance scale is enabled when `guidance_scale > 1`.\n negative_prompt (`str` or `List[str]`, *optional*):\n The prompt or prompts to guide what to not include in image generation. If not defined, you need to\n pass `negative_prompt_embeds` instead. Ignored when not using guidance (`guidance_scale < 1`).\n num_images_per_prompt (`int`, *optional*, defaults to 1):\n The number of images to generate per prompt.\n eta (`float`, *optional*, defaults to 0.0):\n Corresponds to parameter eta (ฮท) from the [DDIM](https://arxiv.org/abs/2010.02502) paper. Only applies\n to the [`~schedulers.DDIMScheduler`], and is ignored in other schedulers.\n generator (`torch.Generator` or `List[torch.Generator]`, *optional*):\n A [`torch.Generator`](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make\n generation deterministic.\n latents (`torch.FloatTensor`, *optional*):\n Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for image\n generation. Can be used to tweak the same generation with different prompts. If not provided, a latents\n tensor is generated by sampling using the supplied random `generator`.\n prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated text embeddings. Can be used to easily tweak text inputs (prompt weighting). If not\n provided, text embeddings are generated from the `prompt` input argument.\n negative_prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated negative text embeddings. Can be used to easily tweak text inputs (prompt weighting). If\n not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.\n ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.\n output_type (`str`, *optional*, defaults to `\"pil\"`):\n The output format of the generated image. Choose between `PIL.Image` or `np.array`.\n return_dict (`bool`, *optional*, defaults to `True`):\n Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a\n plain tuple.\n cross_attention_kwargs (`dict`, *optional*):\n A kwargs dictionary that if specified is passed along to the [`AttentionProcessor`] as defined in\n [`self.processor`](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).\n guidance_rescale (`float`, *optional*, defaults to 0.0):\n Guidance rescale factor from [Common Diffusion Noise Schedules and Sample Steps are\n Flawed](https://arxiv.org/pdf/2305.08891.pdf). Guidance rescale factor should fix overexposure when\n using zero terminal SNR.\n clip_skip (`int`, *optional*):\n Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that\n the output of the pre-final layer will be used for computing the prompt embeddings.\n callback_on_step_end (`Callable`, *optional*):\n A function that calls at the end of each denoising steps during the inference. The function is called\n with the following arguments: `callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int,\n callback_kwargs: Dict)`. `callback_kwargs` will include a list of all tensors as specified by\n `callback_on_step_end_tensor_inputs`.\n callback_on_step_end_tensor_inputs (`List`, *optional*):\n The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list\n will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the\n `._callback_tensor_inputs` attribute of your pipeline class.\n lora_composite (`bool`, *optional*, defaults to `False`):\n Whether to use the `LoRA Composite` method from the paper\n `Multi-LoRA Composition for Image Generation` to generate the image\n given multiple LoRAs.\n Examples:\n\n Returns:\n [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:\n If `return_dict` is `True`, [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] is returned,\n otherwise a `tuple` is returned where the first element is a list with the generated images and the\n second element is a list of `bool`s indicating whether the corresponding generated image contains\n \"not-safe-for-work\" (nsfw) content.\n \"\"\"\n\n callback = kwargs.pop(\"callback\", None)\n callback_steps = kwargs.pop(\"callback_steps\", None)\n\n if callback is not None:\n deprecate(\n \"callback\",\n \"1.0.0\",\n \"Passing `callback` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`\",\n )\n if callback_steps is not None:\n deprecate(\n \"callback_steps\",\n \"1.0.0\",\n \"Passing `callback_steps` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`\",\n )\n\n # 0. Default height and width to unet\n height = height or self.unet.config.sample_size * self.vae_scale_factor\n width = width or self.unet.config.sample_size * self.vae_scale_factor\n # to deal with lora scaling and other possible forward hooks\n\n # 1. Check inputs. Raise error if not correct\n self.check_inputs(\n prompt,\n height,\n width,\n callback_steps,\n negative_prompt,\n prompt_embeds,\n negative_prompt_embeds,\n callback_on_step_end_tensor_inputs,\n )\n\n self._guidance_scale = guidance_scale\n self._guidance_rescale = guidance_rescale\n self._clip_skip = clip_skip\n self._cross_attention_kwargs = cross_attention_kwargs\n self._interrupt = False\n\n # 2. Define call parameters\n if prompt is not None and isinstance(prompt, str):\n batch_size = 1\n elif prompt is not None and isinstance(prompt, list):\n batch_size = len(prompt)\n else:\n batch_size = prompt_embeds.shape[0]\n\n device = self._execution_device\n\n # 3. Encode input prompt\n lora_scale = (\n self.cross_attention_kwargs.get(\"scale\", None) if self.cross_attention_kwargs is not None else None\n )\n\n prompt_embeds, negative_prompt_embeds = self.encode_prompt(\n prompt,\n device,\n num_images_per_prompt,\n self.do_classifier_free_guidance,\n negative_prompt,\n prompt_embeds=prompt_embeds,\n negative_prompt_embeds=negative_prompt_embeds,\n lora_scale=lora_scale,\n clip_skip=self.clip_skip,\n )\n\n # For classifier free guidance, we need to do two forward passes.\n # Here we concatenate the unconditional and text embeddings into a single batch\n # to avoid doing two forward passes\n if self.do_classifier_free_guidance:\n prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds])\n\n if ip_adapter_image is not None:\n image_embeds = self.prepare_ip_adapter_image_embeds(\n ip_adapter_image, device, batch_size * num_images_per_prompt\n )\n\n # 4. Prepare timesteps\n timesteps, num_inference_steps = retrieve_timesteps(self.scheduler, num_inference_steps, device, timesteps)\n\n # 5. Prepare latent variables\n num_channels_latents = self.unet.config.in_channels\n latents = self.prepare_latents(\n batch_size * num_images_per_prompt,\n num_channels_latents,\n height,\n width,\n prompt_embeds.dtype,\n device,\n generator,\n latents,\n )\n\n # 6. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline\n extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)\n\n # 6.1 Add image embeds for IP-Adapter\n added_cond_kwargs = {\"image_embeds\": image_embeds} if ip_adapter_image is not None else None\n\n # 6.2 Optionally get Guidance Scale Embedding\n timestep_cond = None\n if self.unet.config.time_cond_proj_dim is not None:\n guidance_scale_tensor = torch.tensor(self.guidance_scale - 1).repeat(batch_size * num_images_per_prompt)\n timestep_cond = self.get_guidance_scale_embedding(\n guidance_scale_tensor, embedding_dim=self.unet.config.time_cond_proj_dim\n ).to(device=device, dtype=latents.dtype)\n\n # 7. Denoising loop\n num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order\n\n if lora_composite:\n adapters = self.get_active_adapters()\n\n self._num_timesteps = len(timesteps)\n with self.progress_bar(total=num_inference_steps) as progress_bar:\n for i, t in enumerate(timesteps):\n if self.interrupt:\n continue\n\n # expand the latents if we are doing classifier free guidance\n latent_model_input = torch.cat([latents] * 2) if self.do_classifier_free_guidance else latents\n latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)\n\n # predict the noise residual\n if lora_composite:\n noise_preds = []\n # get noise_pred conditioned on each lora\n self.enable_lora()\n for adapter in adapters:\n self.set_adapters(adapter)\n noise_pred = self.unet(\n latent_model_input,\n t,\n encoder_hidden_states=prompt_embeds,\n timestep_cond=timestep_cond,\n cross_attention_kwargs=self.cross_attention_kwargs,\n added_cond_kwargs=added_cond_kwargs,\n return_dict=False,\n )[0]\n noise_preds.append(noise_pred)\n else:\n noise_pred = self.unet(\n latent_model_input,\n t,\n encoder_hidden_states=prompt_embeds,\n timestep_cond=timestep_cond,\n cross_attention_kwargs=self.cross_attention_kwargs,\n added_cond_kwargs=added_cond_kwargs,\n return_dict=False,\n )[0]\n\n # perform guidance\n if self.do_classifier_free_guidance:\n if lora_composite:\n noise_preds = torch.stack(noise_preds, dim=0)\n noise_pred_uncond, noise_pred_text = noise_preds.chunk(2, dim=1)\n noise_pred_uncond = noise_pred_uncond.mean(dim=0)\n noise_pred_text = noise_pred_text.mean(dim=0)\n noise_pred = noise_pred_uncond + self.guidance_scale * (noise_pred_text - noise_pred_uncond)\n else:\n noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)\n noise_pred = noise_pred_uncond + self.guidance_scale * (noise_pred_text - noise_pred_uncond)\n\n if self.do_classifier_free_guidance and self.guidance_rescale > 0.0:\n # Based on 3.4. in https://arxiv.org/pdf/2305.08891.pdf\n noise_pred = rescale_noise_cfg(noise_pred, noise_pred_text, guidance_rescale=self.guidance_rescale)\n\n # compute the previous noisy sample x_t -> x_t-1\n latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]\n\n if callback_on_step_end is not None:\n callback_kwargs = {}\n for k in callback_on_step_end_tensor_inputs:\n callback_kwargs[k] = locals()[k]\n callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)\n\n latents = callback_outputs.pop(\"latents\", latents)\n prompt_embeds = callback_outputs.pop(\"prompt_embeds\", prompt_embeds)\n negative_prompt_embeds = callback_outputs.pop(\"negative_prompt_embeds\", negative_prompt_embeds)\n\n # call the callback, if provided\n if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):\n progress_bar.update()\n if callback is not None and i % callback_steps == 0:\n step_idx = i // getattr(self.scheduler, \"order\", 1)\n callback(step_idx, t, latents)\n\n if not output_type == \"latent\":\n image = self.vae.decode(latents / self.vae.config.scaling_factor, return_dict=False, generator=generator)[\n 0\n ]\n image, has_nsfw_concept = self.run_safety_checker(image, device, prompt_embeds.dtype)\n else:\n image = latents\n has_nsfw_concept = None\n\n if has_nsfw_concept is None:\n do_denormalize = [True] * image.shape[0]\n else:\n do_denormalize = [not has_nsfw for has_nsfw in has_nsfw_concept]\n\n image = self.image_processor.postprocess(image, output_type=output_type, do_denormalize=do_denormalize)\n\n # Offload all models\n self.maybe_free_model_hooks()\n\n if not return_dict:\n return (image, has_nsfw_concept)\n\n return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)\n" }, { "path": "pipelines/sdxl_0.26.3/pipeline.py", "content": "# Copyright 2023 The HuggingFace Team. All rights reserved.\n#\n# Licensed under the Apache License, Version 2.0 (the \"License\");\n# you may not use this file except in compliance with the License.\n# You may obtain a copy of the License at\n#\n# http://www.apache.org/licenses/LICENSE-2.0\n#\n# Unless required by applicable law or agreed to in writing, software\n# distributed under the License is distributed on an \"AS IS\" BASIS,\n# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.\n\nimport inspect\nfrom typing import Any, Callable, Dict, List, Optional, Tuple, Union\n\nimport torch\nfrom transformers import (\n CLIPImageProcessor,\n CLIPTextModel,\n CLIPTextModelWithProjection,\n CLIPTokenizer,\n CLIPVisionModelWithProjection,\n)\n\nfrom diffusers.image_processor import PipelineImageInput, VaeImageProcessor\nfrom diffusers.loaders import (\n FromSingleFileMixin,\n IPAdapterMixin,\n StableDiffusionXLLoraLoaderMixin,\n TextualInversionLoaderMixin,\n)\nfrom diffusers.models import AutoencoderKL, ImageProjection, UNet2DConditionModel\nfrom diffusers.models.attention_processor import (\n AttnProcessor2_0,\n FusedAttnProcessor2_0,\n LoRAAttnProcessor2_0,\n LoRAXFormersAttnProcessor,\n XFormersAttnProcessor,\n)\nfrom diffusers.models.lora import adjust_lora_scale_text_encoder\nfrom diffusers.schedulers import KarrasDiffusionSchedulers\nfrom diffusers.utils import (\n USE_PEFT_BACKEND,\n deprecate,\n is_invisible_watermark_available,\n is_torch_xla_available,\n logging,\n replace_example_docstring,\n scale_lora_layers,\n unscale_lora_layers,\n)\nfrom diffusers.utils.torch_utils import randn_tensor\nfrom diffusers.pipelines.pipeline_utils import DiffusionPipeline\nfrom diffusers.pipelines.stable_diffusion_xl.pipeline_output import StableDiffusionXLPipelineOutput\n\n\nif is_invisible_watermark_available():\n from .watermark import StableDiffusionXLWatermarker\n\nif is_torch_xla_available():\n import torch_xla.core.xla_model as xm\n\n XLA_AVAILABLE = True\nelse:\n XLA_AVAILABLE = False\n\n\nlogger = logging.get_logger(__name__) # pylint: disable=invalid-name\n\nEXAMPLE_DOC_STRING = \"\"\"\n Examples:\n ```py\n >>> import torch\n >>> from diffusers import StableDiffusionXLPipeline\n\n >>> pipe = StableDiffusionXLPipeline.from_pretrained(\n ... \"stabilityai/stable-diffusion-xl-base-1.0\", torch_dtype=torch.float16\n ... )\n >>> pipe = pipe.to(\"cuda\")\n\n >>> prompt = \"a photo of an astronaut riding a horse on mars\"\n >>> image = pipe(prompt).images[0]\n ```\n\"\"\"\n\n\n# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.rescale_noise_cfg\ndef rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0):\n \"\"\"\n Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and\n Sample Steps are Flawed](https://arxiv.org/pdf/2305.08891.pdf). See Section 3.4\n \"\"\"\n std_text = noise_pred_text.std(dim=list(range(1, noise_pred_text.ndim)), keepdim=True)\n std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True)\n # rescale the results from guidance (fixes overexposure)\n noise_pred_rescaled = noise_cfg * (std_text / std_cfg)\n # mix with the original results from guidance by factor guidance_rescale to avoid \"plain looking\" images\n noise_cfg = guidance_rescale * noise_pred_rescaled + (1 - guidance_rescale) * noise_cfg\n return noise_cfg\n\n\n# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps\ndef retrieve_timesteps(\n scheduler,\n num_inference_steps: Optional[int] = None,\n device: Optional[Union[str, torch.device]] = None,\n timesteps: Optional[List[int]] = None,\n **kwargs,\n):\n \"\"\"\n Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles\n custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.\n\n Args:\n scheduler (`SchedulerMixin`):\n The scheduler to get timesteps from.\n num_inference_steps (`int`):\n The number of diffusion steps used when generating samples with a pre-trained model. If used,\n `timesteps` must be `None`.\n device (`str` or `torch.device`, *optional*):\n The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.\n timesteps (`List[int]`, *optional*):\n Custom timesteps used to support arbitrary spacing between timesteps. If `None`, then the default\n timestep spacing strategy of the scheduler is used. If `timesteps` is passed, `num_inference_steps`\n must be `None`.\n\n Returns:\n `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the\n second element is the number of inference steps.\n \"\"\"\n if timesteps is not None:\n accepts_timesteps = \"timesteps\" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())\n if not accepts_timesteps:\n raise ValueError(\n f\"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom\"\n f\" timestep schedules. Please check whether you are using the correct scheduler.\"\n )\n scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)\n timesteps = scheduler.timesteps\n num_inference_steps = len(timesteps)\n else:\n scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)\n timesteps = scheduler.timesteps\n return timesteps, num_inference_steps\n\n\nclass StableDiffusionXLPipeline(\n DiffusionPipeline,\n FromSingleFileMixin,\n StableDiffusionXLLoraLoaderMixin,\n TextualInversionLoaderMixin,\n IPAdapterMixin,\n):\n r\"\"\"\n Pipeline for text-to-image generation using Stable Diffusion XL.\n\n This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods the\n library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)\n\n The pipeline also inherits the following loading methods:\n - [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings\n - [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files\n - [`~loaders.StableDiffusionXLLoraLoaderMixin.load_lora_weights`] for loading LoRA weights\n - [`~loaders.StableDiffusionXLLoraLoaderMixin.save_lora_weights`] for saving LoRA weights\n - [`~loaders.IPAdapterMixin.load_ip_adapter`] for loading IP Adapters\n\n Args:\n vae ([`AutoencoderKL`]):\n Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.\n text_encoder ([`CLIPTextModel`]):\n Frozen text-encoder. Stable Diffusion XL uses the text portion of\n [CLIP](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), specifically\n the [clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) variant.\n text_encoder_2 ([` CLIPTextModelWithProjection`]):\n Second frozen text-encoder. Stable Diffusion XL uses the text and pool portion of\n [CLIP](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModelWithProjection),\n specifically the\n [laion/CLIP-ViT-bigG-14-laion2B-39B-b160k](https://huggingface.co/laion/CLIP-ViT-bigG-14-laion2B-39B-b160k)\n variant.\n tokenizer (`CLIPTokenizer`):\n Tokenizer of class\n [CLIPTokenizer](https://huggingface.co/docs/transformers/v4.21.0/en/model_doc/clip#transformers.CLIPTokenizer).\n tokenizer_2 (`CLIPTokenizer`):\n Second Tokenizer of class\n [CLIPTokenizer](https://huggingface.co/docs/transformers/v4.21.0/en/model_doc/clip#transformers.CLIPTokenizer).\n unet ([`UNet2DConditionModel`]): Conditional U-Net architecture to denoise the encoded image latents.\n scheduler ([`SchedulerMixin`]):\n A scheduler to be used in combination with `unet` to denoise the encoded image latents. Can be one of\n [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].\n force_zeros_for_empty_prompt (`bool`, *optional*, defaults to `\"True\"`):\n Whether the negative prompt embeddings shall be forced to always be set to 0. Also see the config of\n `stabilityai/stable-diffusion-xl-base-1-0`.\n add_watermarker (`bool`, *optional*):\n Whether to use the [invisible_watermark library](https://github.com/ShieldMnt/invisible-watermark/) to\n watermark output images. If not defined, it will default to True if the package is installed, otherwise no\n watermarker will be used.\n \"\"\"\n\n model_cpu_offload_seq = \"text_encoder->text_encoder_2->image_encoder->unet->vae\"\n _optional_components = [\n \"tokenizer\",\n \"tokenizer_2\",\n \"text_encoder\",\n \"text_encoder_2\",\n \"image_encoder\",\n \"feature_extractor\",\n ]\n _callback_tensor_inputs = [\n \"latents\",\n \"prompt_embeds\",\n \"negative_prompt_embeds\",\n \"add_text_embeds\",\n \"add_time_ids\",\n \"negative_pooled_prompt_embeds\",\n \"negative_add_time_ids\",\n ]\n\n def __init__(\n self,\n vae: AutoencoderKL,\n text_encoder: CLIPTextModel,\n text_encoder_2: CLIPTextModelWithProjection,\n tokenizer: CLIPTokenizer,\n tokenizer_2: CLIPTokenizer,\n unet: UNet2DConditionModel,\n scheduler: KarrasDiffusionSchedulers,\n image_encoder: CLIPVisionModelWithProjection = None,\n feature_extractor: CLIPImageProcessor = None,\n force_zeros_for_empty_prompt: bool = True,\n add_watermarker: Optional[bool] = None,\n ):\n super().__init__()\n\n self.register_modules(\n vae=vae,\n text_encoder=text_encoder,\n text_encoder_2=text_encoder_2,\n tokenizer=tokenizer,\n tokenizer_2=tokenizer_2,\n unet=unet,\n scheduler=scheduler,\n image_encoder=image_encoder,\n feature_extractor=feature_extractor,\n )\n self.register_to_config(force_zeros_for_empty_prompt=force_zeros_for_empty_prompt)\n self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)\n self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)\n\n self.default_sample_size = self.unet.config.sample_size\n\n add_watermarker = add_watermarker if add_watermarker is not None else is_invisible_watermark_available()\n\n if add_watermarker:\n self.watermark = StableDiffusionXLWatermarker()\n else:\n self.watermark = None\n\n # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_slicing\n def enable_vae_slicing(self):\n r\"\"\"\n Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to\n compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.\n \"\"\"\n self.vae.enable_slicing()\n\n # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_slicing\n def disable_vae_slicing(self):\n r\"\"\"\n Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to\n computing decoding in one step.\n \"\"\"\n self.vae.disable_slicing()\n\n # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_vae_tiling\n def enable_vae_tiling(self):\n r\"\"\"\n Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to\n compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow\n processing larger images.\n \"\"\"\n self.vae.enable_tiling()\n\n # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_vae_tiling\n def disable_vae_tiling(self):\n r\"\"\"\n Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to\n computing decoding in one step.\n \"\"\"\n self.vae.disable_tiling()\n\n def encode_prompt(\n self,\n prompt: str,\n prompt_2: Optional[str] = None,\n device: Optional[torch.device] = None,\n num_images_per_prompt: int = 1,\n do_classifier_free_guidance: bool = True,\n negative_prompt: Optional[str] = None,\n negative_prompt_2: Optional[str] = None,\n prompt_embeds: Optional[torch.FloatTensor] = None,\n negative_prompt_embeds: Optional[torch.FloatTensor] = None,\n pooled_prompt_embeds: Optional[torch.FloatTensor] = None,\n negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,\n lora_scale: Optional[float] = None,\n clip_skip: Optional[int] = None,\n ):\n r\"\"\"\n Encodes the prompt into text encoder hidden states.\n\n Args:\n prompt (`str` or `List[str]`, *optional*):\n prompt to be encoded\n prompt_2 (`str` or `List[str]`, *optional*):\n The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is\n used in both text-encoders\n device: (`torch.device`):\n torch device\n num_images_per_prompt (`int`):\n number of images that should be generated per prompt\n do_classifier_free_guidance (`bool`):\n whether to use classifier free guidance or not\n negative_prompt (`str` or `List[str]`, *optional*):\n The prompt or prompts not to guide the image generation. If not defined, one has to pass\n `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is\n less than `1`).\n negative_prompt_2 (`str` or `List[str]`, *optional*):\n The prompt or prompts not to guide the image generation to be sent to `tokenizer_2` and\n `text_encoder_2`. If not defined, `negative_prompt` is used in both text-encoders\n prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not\n provided, text embeddings will be generated from `prompt` input argument.\n negative_prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt\n weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input\n argument.\n pooled_prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.\n If not provided, pooled text embeddings will be generated from `prompt` input argument.\n negative_pooled_prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt\n weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`\n input argument.\n lora_scale (`float`, *optional*):\n A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.\n clip_skip (`int`, *optional*):\n Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that\n the output of the pre-final layer will be used for computing the prompt embeddings.\n \"\"\"\n device = device or self._execution_device\n\n # set lora scale so that monkey patched LoRA\n # function of text encoder can correctly access it\n if lora_scale is not None and isinstance(self, StableDiffusionXLLoraLoaderMixin):\n self._lora_scale = lora_scale\n\n # dynamically adjust the LoRA scale\n if self.text_encoder is not None:\n if not USE_PEFT_BACKEND:\n adjust_lora_scale_text_encoder(self.text_encoder, lora_scale)\n else:\n scale_lora_layers(self.text_encoder, lora_scale)\n\n if self.text_encoder_2 is not None:\n if not USE_PEFT_BACKEND:\n adjust_lora_scale_text_encoder(self.text_encoder_2, lora_scale)\n else:\n scale_lora_layers(self.text_encoder_2, lora_scale)\n\n prompt = [prompt] if isinstance(prompt, str) else prompt\n\n if prompt is not None:\n batch_size = len(prompt)\n else:\n batch_size = prompt_embeds.shape[0]\n\n # Define tokenizers and text encoders\n tokenizers = [self.tokenizer, self.tokenizer_2] if self.tokenizer is not None else [self.tokenizer_2]\n text_encoders = (\n [self.text_encoder, self.text_encoder_2] if self.text_encoder is not None else [self.text_encoder_2]\n )\n\n if prompt_embeds is None:\n prompt_2 = prompt_2 or prompt\n prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2\n\n # textual inversion: procecss multi-vector tokens if necessary\n prompt_embeds_list = []\n prompts = [prompt, prompt_2]\n for prompt, tokenizer, text_encoder in zip(prompts, tokenizers, text_encoders):\n if isinstance(self, TextualInversionLoaderMixin):\n prompt = self.maybe_convert_prompt(prompt, tokenizer)\n\n text_inputs = tokenizer(\n prompt,\n padding=\"max_length\",\n max_length=tokenizer.model_max_length,\n truncation=True,\n return_tensors=\"pt\",\n )\n\n text_input_ids = text_inputs.input_ids\n untruncated_ids = tokenizer(prompt, padding=\"longest\", return_tensors=\"pt\").input_ids\n\n if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(\n text_input_ids, untruncated_ids\n ):\n removed_text = tokenizer.batch_decode(untruncated_ids[:, tokenizer.model_max_length - 1 : -1])\n logger.warning(\n \"The following part of your input was truncated because CLIP can only handle sequences up to\"\n f\" {tokenizer.model_max_length} tokens: {removed_text}\"\n )\n\n prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)\n\n # We are only ALWAYS interested in the pooled output of the final text encoder\n pooled_prompt_embeds = prompt_embeds[0]\n if clip_skip is None:\n prompt_embeds = prompt_embeds.hidden_states[-2]\n else:\n # \"2\" because SDXL always indexes from the penultimate layer.\n prompt_embeds = prompt_embeds.hidden_states[-(clip_skip + 2)]\n\n prompt_embeds_list.append(prompt_embeds)\n\n prompt_embeds = torch.concat(prompt_embeds_list, dim=-1)\n\n # get unconditional embeddings for classifier free guidance\n zero_out_negative_prompt = negative_prompt is None and self.config.force_zeros_for_empty_prompt\n if do_classifier_free_guidance and negative_prompt_embeds is None and zero_out_negative_prompt:\n negative_prompt_embeds = torch.zeros_like(prompt_embeds)\n negative_pooled_prompt_embeds = torch.zeros_like(pooled_prompt_embeds)\n elif do_classifier_free_guidance and negative_prompt_embeds is None:\n negative_prompt = negative_prompt or \"\"\n negative_prompt_2 = negative_prompt_2 or negative_prompt\n\n # normalize str to list\n negative_prompt = batch_size * [negative_prompt] if isinstance(negative_prompt, str) else negative_prompt\n negative_prompt_2 = (\n batch_size * [negative_prompt_2] if isinstance(negative_prompt_2, str) else negative_prompt_2\n )\n\n uncond_tokens: List[str]\n if prompt is not None and type(prompt) is not type(negative_prompt):\n raise TypeError(\n f\"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !=\"\n f\" {type(prompt)}.\"\n )\n elif batch_size != len(negative_prompt):\n raise ValueError(\n f\"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:\"\n f\" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches\"\n \" the batch size of `prompt`.\"\n )\n else:\n uncond_tokens = [negative_prompt, negative_prompt_2]\n\n negative_prompt_embeds_list = []\n for negative_prompt, tokenizer, text_encoder in zip(uncond_tokens, tokenizers, text_encoders):\n if isinstance(self, TextualInversionLoaderMixin):\n negative_prompt = self.maybe_convert_prompt(negative_prompt, tokenizer)\n\n max_length = prompt_embeds.shape[1]\n uncond_input = tokenizer(\n negative_prompt,\n padding=\"max_length\",\n max_length=max_length,\n truncation=True,\n return_tensors=\"pt\",\n )\n\n negative_prompt_embeds = text_encoder(\n uncond_input.input_ids.to(device),\n output_hidden_states=True,\n )\n # We are only ALWAYS interested in the pooled output of the final text encoder\n negative_pooled_prompt_embeds = negative_prompt_embeds[0]\n negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]\n\n negative_prompt_embeds_list.append(negative_prompt_embeds)\n\n negative_prompt_embeds = torch.concat(negative_prompt_embeds_list, dim=-1)\n\n if self.text_encoder_2 is not None:\n prompt_embeds = prompt_embeds.to(dtype=self.text_encoder_2.dtype, device=device)\n else:\n prompt_embeds = prompt_embeds.to(dtype=self.unet.dtype, device=device)\n\n bs_embed, seq_len, _ = prompt_embeds.shape\n # duplicate text embeddings for each generation per prompt, using mps friendly method\n prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)\n prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)\n\n if do_classifier_free_guidance:\n # duplicate unconditional embeddings for each generation per prompt, using mps friendly method\n seq_len = negative_prompt_embeds.shape[1]\n\n if self.text_encoder_2 is not None:\n negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.text_encoder_2.dtype, device=device)\n else:\n negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.unet.dtype, device=device)\n\n negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)\n negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)\n\n pooled_prompt_embeds = pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(\n bs_embed * num_images_per_prompt, -1\n )\n if do_classifier_free_guidance:\n negative_pooled_prompt_embeds = negative_pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(\n bs_embed * num_images_per_prompt, -1\n )\n\n if self.text_encoder is not None:\n if isinstance(self, StableDiffusionXLLoraLoaderMixin) and USE_PEFT_BACKEND:\n # Retrieve the original scale by scaling back the LoRA layers\n unscale_lora_layers(self.text_encoder, lora_scale)\n\n if self.text_encoder_2 is not None:\n if isinstance(self, StableDiffusionXLLoraLoaderMixin) and USE_PEFT_BACKEND:\n # Retrieve the original scale by scaling back the LoRA layers\n unscale_lora_layers(self.text_encoder_2, lora_scale)\n\n return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds\n\n # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_image\n def encode_image(self, image, device, num_images_per_prompt, output_hidden_states=None):\n dtype = next(self.image_encoder.parameters()).dtype\n\n if not isinstance(image, torch.Tensor):\n image = self.feature_extractor(image, return_tensors=\"pt\").pixel_values\n\n image = image.to(device=device, dtype=dtype)\n if output_hidden_states:\n image_enc_hidden_states = self.image_encoder(image, output_hidden_states=True).hidden_states[-2]\n image_enc_hidden_states = image_enc_hidden_states.repeat_interleave(num_images_per_prompt, dim=0)\n uncond_image_enc_hidden_states = self.image_encoder(\n torch.zeros_like(image), output_hidden_states=True\n ).hidden_states[-2]\n uncond_image_enc_hidden_states = uncond_image_enc_hidden_states.repeat_interleave(\n num_images_per_prompt, dim=0\n )\n return image_enc_hidden_states, uncond_image_enc_hidden_states\n else:\n image_embeds = self.image_encoder(image).image_embeds\n image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)\n uncond_image_embeds = torch.zeros_like(image_embeds)\n\n return image_embeds, uncond_image_embeds\n\n # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds\n def prepare_ip_adapter_image_embeds(self, ip_adapter_image, device, num_images_per_prompt):\n if not isinstance(ip_adapter_image, list):\n ip_adapter_image = [ip_adapter_image]\n\n if len(ip_adapter_image) != len(self.unet.encoder_hid_proj.image_projection_layers):\n raise ValueError(\n f\"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.unet.encoder_hid_proj.image_projection_layers)} IP Adapters.\"\n )\n\n image_embeds = []\n for single_ip_adapter_image, image_proj_layer in zip(\n ip_adapter_image, self.unet.encoder_hid_proj.image_projection_layers\n ):\n output_hidden_state = not isinstance(image_proj_layer, ImageProjection)\n single_image_embeds, single_negative_image_embeds = self.encode_image(\n single_ip_adapter_image, device, 1, output_hidden_state\n )\n single_image_embeds = torch.stack([single_image_embeds] * num_images_per_prompt, dim=0)\n single_negative_image_embeds = torch.stack([single_negative_image_embeds] * num_images_per_prompt, dim=0)\n\n if self.do_classifier_free_guidance:\n single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds])\n single_image_embeds = single_image_embeds.to(device)\n\n image_embeds.append(single_image_embeds)\n\n return image_embeds\n\n # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs\n def prepare_extra_step_kwargs(self, generator, eta):\n # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature\n # eta (ฮท) is only used with the DDIMScheduler, it will be ignored for other schedulers.\n # eta corresponds to ฮท in DDIM paper: https://arxiv.org/abs/2010.02502\n # and should be between [0, 1]\n\n accepts_eta = \"eta\" in set(inspect.signature(self.scheduler.step).parameters.keys())\n extra_step_kwargs = {}\n if accepts_eta:\n extra_step_kwargs[\"eta\"] = eta\n\n # check if the scheduler accepts generator\n accepts_generator = \"generator\" in set(inspect.signature(self.scheduler.step).parameters.keys())\n if accepts_generator:\n extra_step_kwargs[\"generator\"] = generator\n return extra_step_kwargs\n\n def check_inputs(\n self,\n prompt,\n prompt_2,\n height,\n width,\n callback_steps,\n negative_prompt=None,\n negative_prompt_2=None,\n prompt_embeds=None,\n negative_prompt_embeds=None,\n pooled_prompt_embeds=None,\n negative_pooled_prompt_embeds=None,\n callback_on_step_end_tensor_inputs=None,\n ):\n if height % 8 != 0 or width % 8 != 0:\n raise ValueError(f\"`height` and `width` have to be divisible by 8 but are {height} and {width}.\")\n\n if callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0):\n raise ValueError(\n f\"`callback_steps` has to be a positive integer but is {callback_steps} of type\"\n f\" {type(callback_steps)}.\"\n )\n\n if callback_on_step_end_tensor_inputs is not None and not all(\n k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs\n ):\n raise ValueError(\n f\"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}\"\n )\n\n if prompt is not None and prompt_embeds is not None:\n raise ValueError(\n f\"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to\"\n \" only forward one of the two.\"\n )\n elif prompt_2 is not None and prompt_embeds is not None:\n raise ValueError(\n f\"Cannot forward both `prompt_2`: {prompt_2} and `prompt_embeds`: {prompt_embeds}. Please make sure to\"\n \" only forward one of the two.\"\n )\n elif prompt is None and prompt_embeds is None:\n raise ValueError(\n \"Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined.\"\n )\n elif prompt is not None and (not isinstance(prompt, str) and not isinstance(prompt, list)):\n raise ValueError(f\"`prompt` has to be of type `str` or `list` but is {type(prompt)}\")\n elif prompt_2 is not None and (not isinstance(prompt_2, str) and not isinstance(prompt_2, list)):\n raise ValueError(f\"`prompt_2` has to be of type `str` or `list` but is {type(prompt_2)}\")\n\n if negative_prompt is not None and negative_prompt_embeds is not None:\n raise ValueError(\n f\"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:\"\n f\" {negative_prompt_embeds}. Please make sure to only forward one of the two.\"\n )\n elif negative_prompt_2 is not None and negative_prompt_embeds is not None:\n raise ValueError(\n f\"Cannot forward both `negative_prompt_2`: {negative_prompt_2} and `negative_prompt_embeds`:\"\n f\" {negative_prompt_embeds}. Please make sure to only forward one of the two.\"\n )\n\n if prompt_embeds is not None and negative_prompt_embeds is not None:\n if prompt_embeds.shape != negative_prompt_embeds.shape:\n raise ValueError(\n \"`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but\"\n f\" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`\"\n f\" {negative_prompt_embeds.shape}.\"\n )\n\n if prompt_embeds is not None and pooled_prompt_embeds is None:\n raise ValueError(\n \"If `prompt_embeds` are provided, `pooled_prompt_embeds` also have to be passed. Make sure to generate `pooled_prompt_embeds` from the same text encoder that was used to generate `prompt_embeds`.\"\n )\n\n if negative_prompt_embeds is not None and negative_pooled_prompt_embeds is None:\n raise ValueError(\n \"If `negative_prompt_embeds` are provided, `negative_pooled_prompt_embeds` also have to be passed. Make sure to generate `negative_pooled_prompt_embeds` from the same text encoder that was used to generate `negative_prompt_embeds`.\"\n )\n\n # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents\n def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):\n shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)\n if isinstance(generator, list) and len(generator) != batch_size:\n raise ValueError(\n f\"You have passed a list of generators of length {len(generator)}, but requested an effective batch\"\n f\" size of {batch_size}. Make sure the batch size matches the length of the generators.\"\n )\n\n if latents is None:\n latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)\n else:\n latents = latents.to(device)\n\n # scale the initial noise by the standard deviation required by the scheduler\n latents = latents * self.scheduler.init_noise_sigma\n return latents\n\n def _get_add_time_ids(\n self, original_size, crops_coords_top_left, target_size, dtype, text_encoder_projection_dim=None\n ):\n add_time_ids = list(original_size + crops_coords_top_left + target_size)\n\n passed_add_embed_dim = (\n self.unet.config.addition_time_embed_dim * len(add_time_ids) + text_encoder_projection_dim\n )\n expected_add_embed_dim = self.unet.add_embedding.linear_1.in_features\n\n if expected_add_embed_dim != passed_add_embed_dim:\n raise ValueError(\n f\"Model expects an added time embedding vector of length {expected_add_embed_dim}, but a vector of {passed_add_embed_dim} was created. The model has an incorrect config. Please check `unet.config.time_embedding_type` and `text_encoder_2.config.projection_dim`.\"\n )\n\n add_time_ids = torch.tensor([add_time_ids], dtype=dtype)\n return add_time_ids\n\n def upcast_vae(self):\n dtype = self.vae.dtype\n self.vae.to(dtype=torch.float32)\n use_torch_2_0_or_xformers = isinstance(\n self.vae.decoder.mid_block.attentions[0].processor,\n (\n AttnProcessor2_0,\n XFormersAttnProcessor,\n LoRAXFormersAttnProcessor,\n LoRAAttnProcessor2_0,\n FusedAttnProcessor2_0,\n ),\n )\n # if xformers or torch_2_0 is used attention block does not need\n # to be in float32 which can save lots of memory\n if use_torch_2_0_or_xformers:\n self.vae.post_quant_conv.to(dtype)\n self.vae.decoder.conv_in.to(dtype)\n self.vae.decoder.mid_block.to(dtype)\n\n # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_freeu\n def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):\n r\"\"\"Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.\n\n The suffixes after the scaling factors represent the stages where they are being applied.\n\n Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values\n that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.\n\n Args:\n s1 (`float`):\n Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to\n mitigate \"oversmoothing effect\" in the enhanced denoising process.\n s2 (`float`):\n Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to\n mitigate \"oversmoothing effect\" in the enhanced denoising process.\n b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.\n b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.\n \"\"\"\n if not hasattr(self, \"unet\"):\n raise ValueError(\"The pipeline must have `unet` for using FreeU.\")\n self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)\n\n # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.disable_freeu\n def disable_freeu(self):\n \"\"\"Disables the FreeU mechanism if enabled.\"\"\"\n self.unet.disable_freeu()\n\n def fuse_qkv_projections(self, unet: bool = True, vae: bool = True):\n \"\"\"\n Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query,\n key, value) are fused. For cross-attention modules, key and value projection matrices are fused.\n\n \n\n This API is ๐Ÿงช experimental.\n\n \n\n Args:\n unet (`bool`, defaults to `True`): To apply fusion on the UNet.\n vae (`bool`, defaults to `True`): To apply fusion on the VAE.\n \"\"\"\n self.fusing_unet = False\n self.fusing_vae = False\n\n if unet:\n self.fusing_unet = True\n self.unet.fuse_qkv_projections()\n self.unet.set_attn_processor(FusedAttnProcessor2_0())\n\n if vae:\n if not isinstance(self.vae, AutoencoderKL):\n raise ValueError(\"`fuse_qkv_projections()` is only supported for the VAE of type `AutoencoderKL`.\")\n\n self.fusing_vae = True\n self.vae.fuse_qkv_projections()\n self.vae.set_attn_processor(FusedAttnProcessor2_0())\n\n def unfuse_qkv_projections(self, unet: bool = True, vae: bool = True):\n \"\"\"Disable QKV projection fusion if enabled.\n\n \n\n This API is ๐Ÿงช experimental.\n\n \n\n Args:\n unet (`bool`, defaults to `True`): To apply fusion on the UNet.\n vae (`bool`, defaults to `True`): To apply fusion on the VAE.\n\n \"\"\"\n if unet:\n if not self.fusing_unet:\n logger.warning(\"The UNet was not initially fused for QKV projections. Doing nothing.\")\n else:\n self.unet.unfuse_qkv_projections()\n self.fusing_unet = False\n\n if vae:\n if not self.fusing_vae:\n logger.warning(\"The VAE was not initially fused for QKV projections. Doing nothing.\")\n else:\n self.vae.unfuse_qkv_projections()\n self.fusing_vae = False\n\n # Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding\n def get_guidance_scale_embedding(self, w, embedding_dim=512, dtype=torch.float32):\n \"\"\"\n See https://github.com/google-research/vdm/blob/dc27b98a554f65cdc654b800da5aa1846545d41b/model_vdm.py#L298\n\n Args:\n timesteps (`torch.Tensor`):\n generate embedding vectors at these timesteps\n embedding_dim (`int`, *optional*, defaults to 512):\n dimension of the embeddings to generate\n dtype:\n data type of the generated embeddings\n\n Returns:\n `torch.FloatTensor`: Embedding vectors with shape `(len(timesteps), embedding_dim)`\n \"\"\"\n assert len(w.shape) == 1\n w = w * 1000.0\n\n half_dim = embedding_dim // 2\n emb = torch.log(torch.tensor(10000.0)) / (half_dim - 1)\n emb = torch.exp(torch.arange(half_dim, dtype=dtype) * -emb)\n emb = w.to(dtype)[:, None] * emb[None, :]\n emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)\n if embedding_dim % 2 == 1: # zero pad\n emb = torch.nn.functional.pad(emb, (0, 1))\n assert emb.shape == (w.shape[0], embedding_dim)\n return emb\n\n @property\n def guidance_scale(self):\n return self._guidance_scale\n\n @property\n def guidance_rescale(self):\n return self._guidance_rescale\n\n @property\n def clip_skip(self):\n return self._clip_skip\n\n # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)\n # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`\n # corresponds to doing no classifier free guidance.\n @property\n def do_classifier_free_guidance(self):\n return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None\n\n @property\n def cross_attention_kwargs(self):\n return self._cross_attention_kwargs\n\n @property\n def denoising_end(self):\n return self._denoising_end\n\n @property\n def num_timesteps(self):\n return self._num_timesteps\n\n @property\n def interrupt(self):\n return self._interrupt\n\n @torch.no_grad()\n @replace_example_docstring(EXAMPLE_DOC_STRING)\n def __call__(\n self,\n prompt: Union[str, List[str]] = None,\n prompt_2: Optional[Union[str, List[str]]] = None,\n height: Optional[int] = None,\n width: Optional[int] = None,\n num_inference_steps: int = 50,\n timesteps: List[int] = None,\n denoising_end: Optional[float] = None,\n guidance_scale: float = 5.0,\n negative_prompt: Optional[Union[str, List[str]]] = None,\n negative_prompt_2: Optional[Union[str, List[str]]] = None,\n num_images_per_prompt: Optional[int] = 1,\n eta: float = 0.0,\n generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,\n latents: Optional[torch.FloatTensor] = None,\n prompt_embeds: Optional[torch.FloatTensor] = None,\n negative_prompt_embeds: Optional[torch.FloatTensor] = None,\n pooled_prompt_embeds: Optional[torch.FloatTensor] = None,\n negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,\n ip_adapter_image: Optional[PipelineImageInput] = None,\n output_type: Optional[str] = \"pil\",\n return_dict: bool = True,\n cross_attention_kwargs: Optional[Dict[str, Any]] = None,\n guidance_rescale: float = 0.0,\n original_size: Optional[Tuple[int, int]] = None,\n crops_coords_top_left: Tuple[int, int] = (0, 0),\n target_size: Optional[Tuple[int, int]] = None,\n negative_original_size: Optional[Tuple[int, int]] = None,\n negative_crops_coords_top_left: Tuple[int, int] = (0, 0),\n negative_target_size: Optional[Tuple[int, int]] = None,\n clip_skip: Optional[int] = None,\n callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,\n callback_on_step_end_tensor_inputs: List[str] = [\"latents\"],\n lora_composite: bool = False,\n **kwargs,\n ):\n r\"\"\"\n Function invoked when calling the pipeline for generation.\n\n Args:\n prompt (`str` or `List[str]`, *optional*):\n The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.\n instead.\n prompt_2 (`str` or `List[str]`, *optional*):\n The prompt or prompts to be sent to the `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is\n used in both text-encoders\n height (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):\n The height in pixels of the generated image. This is set to 1024 by default for the best results.\n Anything below 512 pixels won't work well for\n [stabilityai/stable-diffusion-xl-base-1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0)\n and checkpoints that are not specifically fine-tuned on low resolutions.\n width (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):\n The width in pixels of the generated image. This is set to 1024 by default for the best results.\n Anything below 512 pixels won't work well for\n [stabilityai/stable-diffusion-xl-base-1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0)\n and checkpoints that are not specifically fine-tuned on low resolutions.\n num_inference_steps (`int`, *optional*, defaults to 50):\n The number of denoising steps. More denoising steps usually lead to a higher quality image at the\n expense of slower inference.\n timesteps (`List[int]`, *optional*):\n Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument\n in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is\n passed will be used. Must be in descending order.\n denoising_end (`float`, *optional*):\n When specified, determines the fraction (between 0.0 and 1.0) of the total denoising process to be\n completed before it is intentionally prematurely terminated. As a result, the returned sample will\n still retain a substantial amount of noise as determined by the discrete timesteps selected by the\n scheduler. The denoising_end parameter should ideally be utilized when this pipeline forms a part of a\n \"Mixture of Denoisers\" multi-pipeline setup, as elaborated in [**Refining the Image\n Output**](https://huggingface.co/docs/diffusers/api/pipelines/stable_diffusion/stable_diffusion_xl#refining-the-image-output)\n guidance_scale (`float`, *optional*, defaults to 5.0):\n Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).\n `guidance_scale` is defined as `w` of equation 2. of [Imagen\n Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >\n 1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,\n usually at the expense of lower image quality.\n negative_prompt (`str` or `List[str]`, *optional*):\n The prompt or prompts not to guide the image generation. If not defined, one has to pass\n `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is\n less than `1`).\n negative_prompt_2 (`str` or `List[str]`, *optional*):\n The prompt or prompts not to guide the image generation to be sent to `tokenizer_2` and\n `text_encoder_2`. If not defined, `negative_prompt` is used in both text-encoders\n num_images_per_prompt (`int`, *optional*, defaults to 1):\n The number of images to generate per prompt.\n eta (`float`, *optional*, defaults to 0.0):\n Corresponds to parameter eta (ฮท) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to\n [`schedulers.DDIMScheduler`], will be ignored for others.\n generator (`torch.Generator` or `List[torch.Generator]`, *optional*):\n One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html)\n to make generation deterministic.\n latents (`torch.FloatTensor`, *optional*):\n Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image\n generation. Can be used to tweak the same generation with different prompts. If not provided, a latents\n tensor will ge generated by sampling using the supplied random `generator`.\n prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not\n provided, text embeddings will be generated from `prompt` input argument.\n negative_prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt\n weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input\n argument.\n pooled_prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.\n If not provided, pooled text embeddings will be generated from `prompt` input argument.\n negative_pooled_prompt_embeds (`torch.FloatTensor`, *optional*):\n Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt\n weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`\n input argument.\n ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.\n output_type (`str`, *optional*, defaults to `\"pil\"`):\n The output format of the generate image. Choose between\n [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.\n return_dict (`bool`, *optional*, defaults to `True`):\n Whether or not to return a [`~pipelines.stable_diffusion_xl.StableDiffusionXLPipelineOutput`] instead\n of a plain tuple.\n cross_attention_kwargs (`dict`, *optional*):\n A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under\n `self.processor` in\n [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).\n guidance_rescale (`float`, *optional*, defaults to 0.0):\n Guidance rescale factor proposed by [Common Diffusion Noise Schedules and Sample Steps are\n Flawed](https://arxiv.org/pdf/2305.08891.pdf) `guidance_scale` is defined as `ฯ†` in equation 16. of\n [Common Diffusion Noise Schedules and Sample Steps are Flawed](https://arxiv.org/pdf/2305.08891.pdf).\n Guidance rescale factor should fix overexposure when using zero terminal SNR.\n original_size (`Tuple[int]`, *optional*, defaults to (1024, 1024)):\n If `original_size` is not the same as `target_size` the image will appear to be down- or upsampled.\n `original_size` defaults to `(height, width)` if not specified. Part of SDXL's micro-conditioning as\n explained in section 2.2 of\n [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952).\n crops_coords_top_left (`Tuple[int]`, *optional*, defaults to (0, 0)):\n `crops_coords_top_left` can be used to generate an image that appears to be \"cropped\" from the position\n `crops_coords_top_left` downwards. Favorable, well-centered images are usually achieved by setting\n `crops_coords_top_left` to (0, 0). Part of SDXL's micro-conditioning as explained in section 2.2 of\n [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952).\n target_size (`Tuple[int]`, *optional*, defaults to (1024, 1024)):\n For most cases, `target_size` should be set to the desired height and width of the generated image. If\n not specified it will default to `(height, width)`. Part of SDXL's micro-conditioning as explained in\n section 2.2 of [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952).\n negative_original_size (`Tuple[int]`, *optional*, defaults to (1024, 1024)):\n To negatively condition the generation process based on a specific image resolution. Part of SDXL's\n micro-conditioning as explained in section 2.2 of\n [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952). For more\n information, refer to this issue thread: https://github.com/huggingface/diffusers/issues/4208.\n negative_crops_coords_top_left (`Tuple[int]`, *optional*, defaults to (0, 0)):\n To negatively condition the generation process based on a specific crop coordinates. Part of SDXL's\n micro-conditioning as explained in section 2.2 of\n [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952). For more\n information, refer to this issue thread: https://github.com/huggingface/diffusers/issues/4208.\n negative_target_size (`Tuple[int]`, *optional*, defaults to (1024, 1024)):\n To negatively condition the generation process based on a target image resolution. It should be as same\n as the `target_size` for most cases. Part of SDXL's micro-conditioning as explained in section 2.2 of\n [https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952). For more\n information, refer to this issue thread: https://github.com/huggingface/diffusers/issues/4208.\n callback_on_step_end (`Callable`, *optional*):\n A function that calls at the end of each denoising steps during the inference. The function is called\n with the following arguments: `callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int,\n callback_kwargs: Dict)`. `callback_kwargs` will include a list of all tensors as specified by\n `callback_on_step_end_tensor_inputs`.\n callback_on_step_end_tensor_inputs (`List`, *optional*):\n The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list\n will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the\n `._callback_tensor_inputs` attribute of your pipeline class.\n lora_composite (`bool`, *optional*, defaults to `False`):\n Whether to use the `LoRA Composite` method from the paper\n `Multi-LoRA Composition for Image Generation` to generate the image\n given multiple LoRAs.\n\n Examples:\n\n Returns:\n [`~pipelines.stable_diffusion_xl.StableDiffusionXLPipelineOutput`] or `tuple`:\n [`~pipelines.stable_diffusion_xl.StableDiffusionXLPipelineOutput`] if `return_dict` is True, otherwise a\n `tuple`. When returning a tuple, the first element is a list with the generated images.\n \"\"\"\n\n callback = kwargs.pop(\"callback\", None)\n callback_steps = kwargs.pop(\"callback_steps\", None)\n\n if callback is not None:\n deprecate(\n \"callback\",\n \"1.0.0\",\n \"Passing `callback` as an input argument to `__call__` is deprecated, consider use `callback_on_step_end`\",\n )\n if callback_steps is not None:\n deprecate(\n \"callback_steps\",\n \"1.0.0\",\n \"Passing `callback_steps` as an input argument to `__call__` is deprecated, consider use `callback_on_step_end`\",\n )\n\n # 0. Default height and width to unet\n height = height or self.default_sample_size * self.vae_scale_factor\n width = width or self.default_sample_size * self.vae_scale_factor\n\n original_size = original_size or (height, width)\n target_size = target_size or (height, width)\n\n # 1. Check inputs. Raise error if not correct\n self.check_inputs(\n prompt,\n prompt_2,\n height,\n width,\n callback_steps,\n negative_prompt,\n negative_prompt_2,\n prompt_embeds,\n negative_prompt_embeds,\n pooled_prompt_embeds,\n negative_pooled_prompt_embeds,\n callback_on_step_end_tensor_inputs,\n )\n\n self._guidance_scale = guidance_scale\n self._guidance_rescale = guidance_rescale\n self._clip_skip = clip_skip\n self._cross_attention_kwargs = cross_attention_kwargs\n self._denoising_end = denoising_end\n self._interrupt = False\n\n # 2. Define call parameters\n if prompt is not None and isinstance(prompt, str):\n batch_size = 1\n elif prompt is not None and isinstance(prompt, list):\n batch_size = len(prompt)\n else:\n batch_size = prompt_embeds.shape[0]\n\n device = self._execution_device\n\n # 3. Encode input prompt\n lora_scale = (\n self.cross_attention_kwargs.get(\"scale\", None) if self.cross_attention_kwargs is not None else None\n )\n\n (\n prompt_embeds,\n negative_prompt_embeds,\n pooled_prompt_embeds,\n negative_pooled_prompt_embeds,\n ) = self.encode_prompt(\n prompt=prompt,\n prompt_2=prompt_2,\n device=device,\n num_images_per_prompt=num_images_per_prompt,\n do_classifier_free_guidance=self.do_classifier_free_guidance,\n negative_prompt=negative_prompt,\n negative_prompt_2=negative_prompt_2,\n prompt_embeds=prompt_embeds,\n negative_prompt_embeds=negative_prompt_embeds,\n pooled_prompt_embeds=pooled_prompt_embeds,\n negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,\n lora_scale=lora_scale,\n clip_skip=self.clip_skip,\n )\n\n # 4. Prepare timesteps\n timesteps, num_inference_steps = retrieve_timesteps(self.scheduler, num_inference_steps, device, timesteps)\n\n # 5. Prepare latent variables\n num_channels_latents = self.unet.config.in_channels\n latents = self.prepare_latents(\n batch_size * num_images_per_prompt,\n num_channels_latents,\n height,\n width,\n prompt_embeds.dtype,\n device,\n generator,\n latents,\n )\n\n # 6. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline\n extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)\n\n # 7. Prepare added time ids & embeddings\n add_text_embeds = pooled_prompt_embeds\n if self.text_encoder_2 is None:\n text_encoder_projection_dim = int(pooled_prompt_embeds.shape[-1])\n else:\n text_encoder_projection_dim = self.text_encoder_2.config.projection_dim\n\n add_time_ids = self._get_add_time_ids(\n original_size,\n crops_coords_top_left,\n target_size,\n dtype=prompt_embeds.dtype,\n text_encoder_projection_dim=text_encoder_projection_dim,\n )\n if negative_original_size is not None and negative_target_size is not None:\n negative_add_time_ids = self._get_add_time_ids(\n negative_original_size,\n negative_crops_coords_top_left,\n negative_target_size,\n dtype=prompt_embeds.dtype,\n text_encoder_projection_dim=text_encoder_projection_dim,\n )\n else:\n negative_add_time_ids = add_time_ids\n\n if self.do_classifier_free_guidance:\n prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds], dim=0)\n add_text_embeds = torch.cat([negative_pooled_prompt_embeds, add_text_embeds], dim=0)\n add_time_ids = torch.cat([negative_add_time_ids, add_time_ids], dim=0)\n\n prompt_embeds = prompt_embeds.to(device)\n add_text_embeds = add_text_embeds.to(device)\n add_time_ids = add_time_ids.to(device).repeat(batch_size * num_images_per_prompt, 1)\n\n if ip_adapter_image is not None:\n image_embeds = self.prepare_ip_adapter_image_embeds(\n ip_adapter_image, device, batch_size * num_images_per_prompt\n )\n\n # 8. Denoising loop\n num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)\n\n if lora_composite:\n adapters = self.get_active_adapters()\n\n # 8.1 Apply denoising_end\n if (\n self.denoising_end is not None\n and isinstance(self.denoising_end, float)\n and self.denoising_end > 0\n and self.denoising_end < 1\n ):\n discrete_timestep_cutoff = int(\n round(\n self.scheduler.config.num_train_timesteps\n - (self.denoising_end * self.scheduler.config.num_train_timesteps)\n )\n )\n num_inference_steps = len(list(filter(lambda ts: ts >= discrete_timestep_cutoff, timesteps)))\n timesteps = timesteps[:num_inference_steps]\n\n # 9. Optionally get Guidance Scale Embedding\n timestep_cond = None\n if self.unet.config.time_cond_proj_dim is not None:\n guidance_scale_tensor = torch.tensor(self.guidance_scale - 1).repeat(batch_size * num_images_per_prompt)\n timestep_cond = self.get_guidance_scale_embedding(\n guidance_scale_tensor, embedding_dim=self.unet.config.time_cond_proj_dim\n ).to(device=device, dtype=latents.dtype)\n\n self._num_timesteps = len(timesteps)\n with self.progress_bar(total=num_inference_steps) as progress_bar:\n for i, t in enumerate(timesteps):\n if self.interrupt:\n continue\n\n # expand the latents if we are doing classifier free guidance\n latent_model_input = torch.cat([latents] * 2) if self.do_classifier_free_guidance else latents\n\n latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)\n\n # predict the noise residual\n added_cond_kwargs = {\"text_embeds\": add_text_embeds, \"time_ids\": add_time_ids}\n if ip_adapter_image is not None:\n added_cond_kwargs[\"image_embeds\"] = image_embeds\n\n # predict the noise residual\n if lora_composite:\n noise_preds = []\n # get noise_pred conditioned on each lora\n self.enable_lora()\n for adapter in adapters:\n self.set_adapters(adapter)\n noise_pred = self.unet(\n latent_model_input,\n t,\n encoder_hidden_states=prompt_embeds,\n timestep_cond=timestep_cond,\n cross_attention_kwargs=self.cross_attention_kwargs,\n added_cond_kwargs=added_cond_kwargs,\n return_dict=False,\n )[0]\n noise_preds.append(noise_pred)\n else:\n noise_pred = self.unet(\n latent_model_input,\n t,\n encoder_hidden_states=prompt_embeds,\n timestep_cond=timestep_cond,\n cross_attention_kwargs=self.cross_attention_kwargs,\n added_cond_kwargs=added_cond_kwargs,\n return_dict=False,\n )[0]\n\n # perform guidance\n if self.do_classifier_free_guidance:\n if lora_composite:\n noise_preds = torch.stack(noise_preds, dim=0)\n noise_pred_uncond, noise_pred_text = noise_preds.chunk(2, dim=1)\n noise_pred_uncond = noise_pred_uncond.mean(dim=0)\n noise_pred_text = noise_pred_text.mean(dim=0)\n noise_pred = noise_pred_uncond + self.guidance_scale * (noise_pred_text - noise_pred_uncond)\n else:\n noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)\n noise_pred = noise_pred_uncond + self.guidance_scale * (noise_pred_text - noise_pred_uncond)\n\n if self.do_classifier_free_guidance and self.guidance_rescale > 0.0:\n # Based on 3.4. in https://arxiv.org/pdf/2305.08891.pdf\n noise_pred = rescale_noise_cfg(noise_pred, noise_pred_text, guidance_rescale=self.guidance_rescale)\n\n # compute the previous noisy sample x_t -> x_t-1\n latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]\n\n if callback_on_step_end is not None:\n callback_kwargs = {}\n for k in callback_on_step_end_tensor_inputs:\n callback_kwargs[k] = locals()[k]\n callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)\n\n latents = callback_outputs.pop(\"latents\", latents)\n prompt_embeds = callback_outputs.pop(\"prompt_embeds\", prompt_embeds)\n negative_prompt_embeds = callback_outputs.pop(\"negative_prompt_embeds\", negative_prompt_embeds)\n add_text_embeds = callback_outputs.pop(\"add_text_embeds\", add_text_embeds)\n negative_pooled_prompt_embeds = callback_outputs.pop(\n \"negative_pooled_prompt_embeds\", negative_pooled_prompt_embeds\n )\n add_time_ids = callback_outputs.pop(\"add_time_ids\", add_time_ids)\n negative_add_time_ids = callback_outputs.pop(\"negative_add_time_ids\", negative_add_time_ids)\n\n # call the callback, if provided\n if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):\n progress_bar.update()\n if callback is not None and i % callback_steps == 0:\n step_idx = i // getattr(self.scheduler, \"order\", 1)\n callback(step_idx, t, latents)\n\n if XLA_AVAILABLE:\n xm.mark_step()\n\n if not output_type == \"latent\":\n # make sure the VAE is in float32 mode, as it overflows in float16\n needs_upcasting = self.vae.dtype == torch.float16 and self.vae.config.force_upcast\n\n if needs_upcasting:\n self.upcast_vae()\n latents = latents.to(next(iter(self.vae.post_quant_conv.parameters())).dtype)\n\n image = self.vae.decode(latents / self.vae.config.scaling_factor, return_dict=False)[0]\n\n # cast back to fp16 if needed\n if needs_upcasting:\n self.vae.to(dtype=torch.float16)\n else:\n image = latents\n\n if not output_type == \"latent\":\n # apply watermark if available\n if self.watermark is not None:\n image = self.watermark.apply_watermark(image)\n\n image = self.image_processor.postprocess(image, output_type=output_type)\n\n # Offload all models\n self.maybe_free_model_hooks()\n\n if not return_dict:\n return (image,)\n\n return StableDiffusionXLPipelineOutput(images=image)\n" }, { "path": "pipelines/sdxl_0.26.3/watermark.py", "content": "import numpy as np\nimport torch\n\nfrom diffusers.utils import is_invisible_watermark_available\n\n\nif is_invisible_watermark_available():\n from imwatermark import WatermarkEncoder\n\n\n# Copied from https://github.com/Stability-AI/generative-models/blob/613af104c6b85184091d42d374fef420eddb356d/scripts/demo/streamlit_helpers.py#L66\nWATERMARK_MESSAGE = 0b101100111110110010010000011110111011000110011110\n# bin(x)[2:] gives bits of x as str, use int to convert them to 0/1\nWATERMARK_BITS = [int(bit) for bit in bin(WATERMARK_MESSAGE)[2:]]\n\n\nclass StableDiffusionXLWatermarker:\n def __init__(self):\n self.watermark = WATERMARK_BITS\n self.encoder = WatermarkEncoder()\n\n self.encoder.set_watermark(\"bits\", self.watermark)\n\n def apply_watermark(self, images: torch.FloatTensor):\n # can't encode images that are smaller than 256\n if images.shape[-1] < 256:\n return images\n\n images = (255 * (images / 2 + 0.5)).cpu().permute(0, 2, 3, 1).float().numpy()\n\n images = [self.encoder.encode(image, \"dwtDct\") for image in images]\n\n images = torch.from_numpy(np.array(images)).permute(0, 3, 1, 2)\n\n images = torch.clamp(2 * (images / 255 - 0.5), min=-1.0, max=1.0)\n return images\n" }, { "path": "reality_lora_info.json", "content": "{\n \"character\": [\n {\n \"id\": \"character_1\",\n \"name\": \"IU (Lee Ji Eun, Korean singer)\",\n \"trigger\": [\n \"iu1\",\n \"long straight black hair\",\n \"hazel eyes\",\n \"diamond stud earrings\"\n ],\n \"url\": \"https://civitai.com/models/11722/iu?modelVersionId=18576\"\n },\n {\n \"id\": \"character_2\",\n \"name\": \"Scarlett Johansson\",\n \"trigger\": [\n \"scarlett\",\n \"short red hair\",\n \"blue eyes\"\n ],\n \"url\": \"https://civitai.com/models/7468/scarlett-johanssonlora\"\n },\n {\n \"id\": \"character_3\",\n \"name\": \"The Rock (Dwayne Johnson)\",\n \"trigger\": [\n \"th3r0ck with no hair\",\n \"muscular male\",\n \"serious look on his face\"\n ],\n \"url\": \"https://civitai.com/models/22345/dwayne-the-rock-johnsonlora?modelVersionId=26680\"\n }\n ],\n \"clothing\": [\n {\n \"id\": \"clothing_1\",\n \"name\": \"Thai University Uniform\",\n \"trigger\": [\n \"mahalaiuniform\",\n \"white shirt short sleeves\",\n \"black pencil skirt\"\n ],\n \"url\": \"https://civitai.com/models/12657/thai-university-uniform?modelVersionId=58690\"\n },\n {\n \"id\": \"clothing_2\",\n \"name\": \"School Dress\",\n \"trigger\": [\n \"school uniform\",\n \"white shirt\",\n \"red tie\",\n \"blue pleated microskirt\"\n ],\n \"url\": \"https://civitai.com/models/201305?modelVersionId=291486\"\n }\n ],\n \"style\": [\n {\n \"id\": \"style_1\",\n \"name\": \"Japanese Film Color Style\",\n \"trigger\": [\n \"film overlay\",\n \"film grain\"\n ],\n \"url\": \"https://civitai.com/models/90393/japan-vibes-film-color?modelVersionId=107032\"\n },\n {\n \"id\": \"style_2\",\n \"name\": \"Bright Style\",\n \"trigger\": [\n \"bright lighting\"\n ],\n \"url\": \"https://civitai.com/models/70034/brightness-tweaker-lora-lora\"\n }\n ],\n \"background\": [\n {\n \"id\": \"background_1\",\n \"name\": \"Library Bookshelf Background\",\n \"trigger\": [\n \"lib_bg\",\n \"library bookshelf\"\n ],\n \"url\": \"https://civitai.com/models/113488/library-bookshelf?modelVersionId=125699\"\n },\n {\n \"id\": \"background_2\",\n \"name\": \"Forest Background\",\n \"trigger\": [\n \"slg\",\n \"river\",\n \"forest\"\n ],\n \"url\": \"https://civitai.com/models/104292/the-forest-light?modelVersionId=127699\"\n }\n ],\n \"object\": [\n {\n \"id\": \"object_1\",\n \"name\": \"Umbrella\",\n \"trigger\": [\n \"transparent umbrella\"\n ],\n \"url\": \"https://civitai.com/models/54218/umbrellalora?modelVersionId=58578\"\n },\n {\n \"id\": \"object_2\",\n \"name\": \"Bubble Gum\",\n \"trigger\": [\n \"blow bubble gum\"\n ],\n \"url\": \"https://civitai.com/models/97550/bubble-gum-kaugummi-v20?modelVersionId=117038\"\n }\n ]\n}" }, { "path": "requirements.txt", "content": "diffusers[torch]==0.26.3\nopenai\npeft==0.8.2\ntransformers\n" }, { "path": "sdxl_example.py", "content": "import torch\nimport argparse\nfrom diffusers import DiffusionPipeline\nfrom callbacks import make_callback\n\ndef get_example_prompt():\n prompt = \"fluffy pikachu wearing a VR headset, faces visible, cinematic, screencap, high quality\"\n negative_prompt = \"out of frame, lowres, text, error, cropped, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, out of frame, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck, username, watermark, signature\"\n return prompt, negative_prompt\n\ndef main(args):\n\n # set the prompts for image generation\n prompt, negative_prompt = get_example_prompt()\n\n # base model for the realistic style example\n model_name = 'stabilityai/stable-diffusion-xl-base-1.0'\n\n # set base model\n pipeline = DiffusionPipeline.from_pretrained(\n model_name,\n custom_pipeline=\"./pipelines/sdxl_0.26.3\",\n torch_dtype=torch.float16,\n use_safetensors=True,\n variant=\"fp16\"\n ).to(\"cuda\")\n\n # initialize LoRAs\n # This example shows the composition of a character LoRA and a clothing LoRA\n pipeline.load_lora_weights(\"TheLastBen/Pikachu_SDXL\", weight_name=\"pikachu.safetensors\", adapter_name=\"character\")\n pipeline.load_lora_weights(\"fofr/sdxl-vision-pro\", weight_name=\"lora.safetensors\", adapter_name=\"object\")\n cur_loras = [\"character\", \"object\"]\n\n # select the method for the composition\n if args.method == \"merge\":\n pipeline.set_adapters(cur_loras)\n switch_callback = None\n elif args.method == \"switch\":\n pipeline.set_adapters([cur_loras[0]])\n switch_callback = make_callback(switch_step=args.switch_step, loras=cur_loras)\n else:\n pipeline.set_adapters(cur_loras)\n switch_callback = None\n\n image = pipeline(\n prompt=prompt, \n negative_prompt=negative_prompt,\n height=args.height,\n width=args.width,\n num_inference_steps=args.denoise_steps,\n guidance_scale=args.cfg_scale,\n generator=args.generator,\n cross_attention_kwargs={\"scale\": args.lora_scale},\n callback_on_step_end=switch_callback,\n lora_composite=True if args.method == \"composite\" else False\n ).images[0]\n\n image.save(args.save_path)\n\nif __name__ == \"__main__\":\n parser = argparse.ArgumentParser(\n description='Example code for multi-LoRA composition'\n )\n\n # Arguments for composing LoRAs\n parser.add_argument('--method', default='switch',\n choices=['merge', 'switch', 'composite'],\n help='methods for combining LoRAs', type=str)\n parser.add_argument('--save_path', default='example.png',\n help='path to save the generated image', type=str)\n parser.add_argument('--lora_path', default='models/lora/reality',\n help='path to store all LoRAs', type=str)\n parser.add_argument('--lora_scale', default=0.8,\n help='scale of each LoRA when generating images', type=float)\n parser.add_argument('--switch_step', default=5,\n help='number of steps to switch LoRA during denoising, applicable only in the switch method', type=int)\n\n # Arguments for generating images\n parser.add_argument('--height', default=1024,\n help='height of the generated images', type=int)\n parser.add_argument('--width', default=1024,\n help='width of the generated images', type=int)\n parser.add_argument('--denoise_steps', default=100,\n help='number of the denoising steps', type=int)\n parser.add_argument('--cfg_scale', default=7,\n help='scale for classifier-free guidance', type=float)\n parser.add_argument('--seed', default=11,\n help='seed for generating images', type=int)\n\n args = parser.parse_args()\n args.generator = torch.manual_seed(args.seed)\n\n main(args)" }, { "path": "utils.py", "content": "import json\nimport itertools\n\ndef load_lora_info(image_style, path='lora_info.json'):\n path = f\"{image_style}_{path}\"\n with open(path) as f:\n lora_info = json.loads(f.read())\n return lora_info\n\ndef get_prompt(image_style):\n if image_style == 'anime':\n prompt = \"masterpiece, best quality\"\n negative_prompt = \"EasyNegative, extra fingers, extra limbs, fewer fingers, fewer limbs, multiple girls, multiple views, worst quality, low quality, depth of field, blurry, greyscale, 3D face, cropped, lowres, text, jpeg artifacts, signature, watermark, username, blurry, artist name, trademark, watermark, title, reference sheet, curvy, plump, fat, muscular female, strabismus, clothing cutout, side slit, tattoo, nsfw\"\n else:\n prompt = \"RAW photo, subject, 8k uhd, dslr, high quality, Fujifilm XT3, half-length portrait from knees up\"\n negative_prompt = \"extra heads, nsfw, deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, text, cropped, out of frame, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck\"\n return prompt, negative_prompt\n\ndef get_eval_prompt(combo):\n prompt = \"I need assistance in comparatively evaluating two text-to-image models based on their ability to compose different elements into a single image. The elements and their key features are as follows:\\n\\n\"\n cur_index = 1\n element_prompt = \"\"\n for lora in combo:\n ele_type = lora['id'].split('_')[0]\n name = lora['name']\n features = ', '.join(lora['trigger'])\n element_prompt += f\"{cur_index}. {ele_type} ({name}): {features}\\n\"\n cur_index += 1\n prompt += element_prompt + '\\n'\n prompt += \"Please help me rate both given images on the following evaluation dimensions and criteria:\\n\\nComposition quality:\\n- Score on a scale of 0 to 10, in 0.5 increments, where 10 is the best and 0 is the worst.\\n- Deduct 3 points if any element is missing or incorrectly depicted.\\n- Deduct 1 point for each missing or incorrect feature within an element.\\n- Deduct 1 point for minor inconsistencies or lack of harmony between elements.\\n- Additional deductions can be made for compositions that lack coherence, creativity, or realism.\\n\\nImage quality:\\n- Score on a scale of 0 to 10, in 0.5 increments, where 10 is the best and 0 is the worst.\\n- Deduct 3 points for each deformity in the image (e.g., extra limbs or fingers, distorted face, incorrect proportions).\\n- Deduct 2 points for noticeable issues with texture, lighting, or color.\\n- Deduct 1 point for each minor flaw or imperfection.\\n- Additional deductions can be made for any issues affecting the overall aesthetic or clarity of the image.\\n\\nPlease format the evaluation as follows:\\n\\nFor Image 1:\\n[Explanation of evaluation]\\n\\nFor Image 2:\\n[Explanation of evaluation]\\n\\nScores:\\nImage 1: Composition Quality: [score]/10, Image Quality: [score]/10\\nImage 2: Composition Quality: [score]/10, Image Quality: [score]/10\\n\\nBased on the above guidelines, help me to conduct a step-by-step comparative evaluation of the given images. The scoring should follow two principles:\\n1. Please evaluate critically.\\n2. Try not to let the two models end in a tie on both dimensions.\"\n return prompt\n\ndef generate_combinations(lora_info, compos_num):\n \"\"\"\n Generate all combinations of LoRA elements ensuring that each combination includes at least one 'character'.\n\n Args:\n lora_info (dict): A dictionary containing LoRA elements and their instances.\n compos_num (int): The number of elements to be included in each combination.\n\n Returns:\n list: A list of all possible combinations, each combination is a list of element instances.\n \"\"\"\n elements = list(lora_info.keys())\n\n # Check if the composition number is greater than the number of element types\n if compos_num > len(elements):\n raise ValueError(\"The composition number cannot be greater than the number of elements.\")\n\n all_combinations = []\n\n # Ensure that 'character' is always included in the combinations\n if 'character' in elements:\n # Remove 'character' from the list to avoid duplicating\n elements.remove('character')\n\n # Generate all possible combinations of the remaining element types\n selected_types = list(itertools.combinations(elements, compos_num - 1))\n\n # For each combination of types, generate all possible combinations of instances\n for types in selected_types:\n # Add 'character' to the current combination of types\n current_types = ['character', *types]\n\n # Gather instances for each type in the current combination\n instances = [lora_info[t] for t in current_types]\n\n # Create combinations of instances across the selected types\n for combination in itertools.product(*instances):\n all_combinations.append(combination)\n\n return all_combinations" } ]