Showing preview only (280K chars total). Download the full file or copy to clipboard to get everything.
Repository: ant-research/DepthLab
Branch: main
Commit: 33500d7caae2
Files: 19
Total size: 270.6 KB
Directory structure:
gitextract_lf1bmt9g/
├── LICENSE
├── README.md
├── checkpoints/
│ └── place_checkpoints_here.txt
├── environment.yaml
├── infer.py
├── inference/
│ └── depthlab_pipeline.py
├── output/
│ └── in-the-wild_example/
│ └── depth_npy/
│ └── RGB_pred.npy
├── scripts/
│ └── infer.sh
├── src/
│ └── models/
│ ├── attention.py
│ ├── mutual_self_attention.py
│ ├── projection.py
│ ├── transformer_2d.py
│ ├── unet_2d_blocks.py
│ ├── unet_2d_condition.py
│ └── unet_2d_condition_main.py
├── test_cases/
│ ├── know_depth.npy
│ └── mask.npy
└── utils/
├── image_util.py
└── seed_all.py
================================================
FILE CONTENTS
================================================
================================================
FILE: LICENSE
================================================
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
================================================
FILE: README.md
================================================
# DepthLab: From Partial to Complete
This repository represents the official implementation of the paper titled "DepthLab: From Partial to Complete".
[](https://johanan528.github.io/depthlab_web/)
[](https://arxiv.org/abs/2412.18153)
[](https://www.apache.org/licenses/LICENSE-2.0)
[](https://huggingface.co/Johanan0528/DepthLab/tree/main)
<p align="center">
<a href="https://johanan528.github.io/"><strong>Zhiheng Liu*</strong></a>
·
<a href="https://felixcheng97.github.io/"><strong>Ka Leong Cheng*</strong></a>
·
<a href="https://github.com/qiuyu96"><strong>Qiuyu Wang</strong></a>
·
<a href="https://ffrivera0.github.io/"><strong>Shuzhe Wang</strong></a>
·
<a href="https://ken-ouyang.github.io/"><strong>Hao Ouyang</strong></a>
·
<a href="https://icetttb.github.io/"><strong>Bin Tan</strong></a>
·
<a href="https://scholar.google.com/citations?user=Mo_2YsgAAAAJ&hl=zh-CN"><strong>Kai Zhu</strong></a>
·
<a href="https://shenyujun.github.io/"><strong>Yujun Shen</strong></a>
·
<a href="https://cqf.io/"><strong>Qifeng Chen</strong></a>
·
<a href="http://luoping.me/"><strong>Ping Luo</strong></a>
<br>
</p>
We present **DepthLab**, a robust depth inpainting foundation model that can be applied to various downstream tasks to enhance performance. Many tasks naturally contain partial depth information, such as (1) *3D Gaussian inpainting*, (2) *LiDAR depth completion*, (3) *sparse-view reconstruction with DUSt3R*, and (4) *text-to-scene generation*. Our model leverages this known information to achieve improved depth estimation, enhancing performance in downstream tasks. We hope to motivate more related tasks to adopt DepthLab.
## 📢 News
* 2024-12-25: Inference code and paper is released.
* [To-do]: Release the training code to facilitate fine-tuning, allowing adaptation to different mask types in your downstream tasks.
## 🛠️ Setup
### 📦 Repository
Clone the repository (requires git):
```bash
git clone https://github.com/Johanan528/DepthLab.git
cd DepthLab
```
### 💻 Dependencies
Install with `conda`:
```bash
conda env create -f environment.yaml
conda activate DepthLab
```
### 📦 Checkpoints
Download the Marigold checkpoint [here](https://huggingface.co/prs-eth/marigold-depth-v1-0), the image encoder checkpoint [here](https://huggingface.co/laion/CLIP-ViT-H-14-laion2B-s32B-b79K), and our checkpoints at [Hugging Face](https://huggingface.co/Johanan0528/DepthLab/tree/main). The downloaded checkpoint directory has the following structure:
```
.
`-- checkpoints
|-- marigold-depth-v1-0
|-- CLIP-ViT-H-14-laion2B-s32B-b79K
`-- DepthLab
|-- denoising_unet.pth
|-- reference_unet.pth
`-- mapping_layer.pth
```
## 🏃 Testing on your cases
### 📷 Prepare images, masks, known depths
Masks: PNG/JPG or Numpy, where black (0) represents the known regions, and white (1) indicates the predicted areas.
Known depths: Numpy
Images: PNG/JPG
#### We provide a case in 'test_cases' folder.
### 🎮 Run inference
```bash
cd scripts
bash infer.sh
```
You can find all results in `output/in-the-wild_example`. Enjoy!
### ⚙️ Inference settings
The default settings are optimized for the best result. However, the behavior of the code can be customized:
- `--denoise_steps`: Number of denoising steps of each inference pass. For the original (DDIM) version, it's recommended to use 20-50 steps.
- `--processing_res`: The processing resolution. **For cases where the mask is sparse, such as in depth completion scenarios, it is advisable to set the 'processing_res' and the mask size to be the same in order to avoid accuracy loss in the mask due to resizing. For non-sparse completion tasks, we recommend using a resolution of 640 or 768 for inference.**
- `--normalize_scale`: When the known depth scale cannot encompass the global scale, it is possible to reduce the normalization scale, allowing the model to better predict the depth of distant objects.
- `--strength`: When set to 1, the prediction is entirely based on the model itself. When set to a value less than 1, the model is partially assisted by interpolated masked depth to some extent.
- `--blend`: Whether to use Blend Diffusion, a commonly used technique in image inpainting.
- `--refine`: If you want to refine depthmap of DUSt3R, or you have a full initial depthmap, turn this option on.
## 🌺 Acknowledgements
This project is developped on the codebase of [Marigold](https://github.com/prs-eth/Marigold) and [MagicAnimate](https://github.com/magic-research/magic-animate). We appreciate their great works!
## 🎓 Citation
Please cite our paper:
```bibtex
@article{liu2024depthlab,
title={DepthLab: From Partial to Complete},
author={Liu, Zhiheng and Cheng, Ka Leong and Wang, Qiuyu and Wang, Shuzhe and Ouyang, Hao and Tan, Bin and Zhu, Kai and Shen, Yujun and Chen, Qifeng and Luo, Ping},
journal={arXiv preprint arXiv:2412.18153},
year={2024}
}
```
================================================
FILE: checkpoints/place_checkpoints_here.txt
================================================
================================================
FILE: environment.yaml
================================================
name: DepthLab
channels:
- defaults
dependencies:
- pip=24.2=py39h06a4308_0
- python=3.9.19=h955ad1f_1
- pip:
- accelerate==0.34.2
- diffusers==0.27.2
- dill==0.3.8
- einops==0.8.0
- huggingface-hub==0.24.7
- imageio==2.35.1
- importlib-metadata==8.5.0
- importlib-resources==6.4.5
- ipython==8.18.1
- ipywidgets==8.1.5
- matplotlib==3.5.2
- numpy==1.26.4
- open3d==0.18.0
- opencv-python==4.10.0.84
- packaging==24.1
- pandas==2.2.2
- pillow==10.4.0
- prompt-toolkit==3.0.48
- pyyaml==6.0.2
- requests==2.32.3
- safetensors==0.4.5
- scikit-image==0.24.0
- scikit-learn==1.5.2
- scipy==1.13.1
- torch==2.1.0
- torchaudio==2.1.0
- torchvision==0.16.0
- tqdm==4.66.5
- transformers==4.44.2
- trimesh==4.4.9
- zipp==3.20.2
================================================
FILE: infer.py
================================================
import argparse
import logging
import os
import random
import numpy as np
import torch
from PIL import Image
from tqdm.auto import tqdm
from diffusers import (
DDIMScheduler,
AutoencoderKL,
)
import cv2
import torch.nn as nn
from transformers import CLIPTextModel, CLIPTokenizer
from src.models.unet_2d_condition import UNet2DConditionModel
from src.models.unet_2d_condition_main import UNet2DConditionModel_main
from src.models.projection import My_proj
from transformers import CLIPVisionModelWithProjection
from inference.depthlab_pipeline import DepthLabPipeline
from utils.seed_all import seed_all
from utils.image_util import get_filled_for_latents
def load_and_process_mask(mask_path):
image = Image.open(mask_path).convert('L')
mask = np.array(image)
mask = mask / 255.0
mask[mask > 0.5] = 1
mask[mask <= 0.5] = 0
return mask
if "__main__" == __name__:
logging.basicConfig(level=logging.INFO)
# -------------------- Arguments --------------------
parser = argparse.ArgumentParser(
description="Run single-image depth estimation using Marigold."
)
parser.add_argument(
"--output_dir", type=str, required=True, help="Output directory."
)
# inference setting
parser.add_argument(
"--denoise_steps",
type=int,
default=50, # quantitative evaluation uses 50 steps
help="Diffusion denoising steps, more steps results in higher accuracy but slower inference speed.",
)
# resolution setting
parser.add_argument(
"--processing_res",
type=int,
default=0,
help="Maximum resolution of processing. 0 for using input image resolution. Default: 0.",
)
parser.add_argument(
"--normalize_scale",
type=float,
default=1,
help="Maximum resolution of processing. 0 for using input image resolution. Default: 0.",
)
parser.add_argument(
"--strength",
type=float,
default=0.8,
help="Maximum resolution of processing. 0 for using input image resolution. Default: 0.",
)
parser.add_argument("--seed", type=int, default=None, help="Random seed.")
parser.add_argument(
"--pretrained_model_name_or_path",
type=str,
default=None,
required=True,
help="Path to pretrained model or model identifier from huggingface.co/models.",
)
parser.add_argument(
"--image_encoder_path",
type=str,
default=None,
required=True,
help="Path to pretrained model or model identifier from huggingface.co/models.",
)
parser.add_argument(
"--denoising_unet_path", type=str, required=True, help="Path to depth inpainting model."
)
parser.add_argument(
"--mapping_path", type=str, required=True, help="Path to depth inpainting model."
)
parser.add_argument(
"--reference_unet_path", type=str, required=True, help="Path to depth inpainting model."
)
parser.add_argument(
"--input_image_paths",
nargs='+',
default=None,
help="input_image_paths",
)
parser.add_argument(
"--known_depth_paths",
nargs='+',
default=None,
help="known_depth_paths",
)
parser.add_argument(
"--blend",
action="store_true",
help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
)
parser.add_argument(
"--refine",
action="store_true",
help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
)
parser.add_argument(
"--masks_paths",
nargs='+',
default=None,
help="masks_paths",
)
args = parser.parse_args()
output_dir = args.output_dir
denoise_steps = args.denoise_steps
processing_res = args.processing_res
seed = args.seed
output_dir_color = os.path.join(output_dir, "depth_colored")
output_dir_npy = os.path.join(output_dir, "depth_npy")
os.makedirs(output_dir, exist_ok=True)
os.makedirs(output_dir_color, exist_ok=True)
os.makedirs(output_dir_npy, exist_ok=True)
logging.info(f"output dir = {output_dir}")
if args.input_image_paths is not None:
assert len(args.input_image_paths) == len(args.known_depth_paths)
assert len(args.input_image_paths) == len(args.masks_paths)
input_image_paths = args.input_image_paths
known_depth_paths = args.known_depth_paths
masks_paths = args.masks_paths
print(f"arguments: {args}")
if seed is None:
import time
seed = int(time.time())
seed_all(seed)
# -------------------- Device --------------------
if torch.cuda.is_available():
device = torch.device("cuda")
else:
device = torch.device("cpu")
logging.warning("CUDA is not available. Running on CPU will be slow.")
logging.info(f"device = {device}")
# -------------------- Model --------------------
vae = AutoencoderKL.from_pretrained(args.pretrained_model_name_or_path,
subfolder='vae')
text_encoder = CLIPTextModel.from_pretrained(args.pretrained_model_name_or_path,
subfolder='text_encoder')
denoising_unet = UNet2DConditionModel_main.from_pretrained(args.pretrained_model_name_or_path,subfolder="unet",
in_channels=12, sample_size=96,
low_cpu_mem_usage=False,
ignore_mismatched_sizes=True)
reference_unet = UNet2DConditionModel.from_pretrained(args.pretrained_model_name_or_path,subfolder="unet",
in_channels=4, sample_size=96,
low_cpu_mem_usage=False,
ignore_mismatched_sizes=True)
image_enc = CLIPVisionModelWithProjection.from_pretrained(args.image_encoder_path)
mapping_layer=My_proj()
mapping_layer.load_state_dict(
torch.load(args.mapping_path, map_location="cpu"),
strict=False,
)
mapping_device = torch.device("cuda")
mapping_layer.to(mapping_device )
reference_unet.load_state_dict(
torch.load(args.reference_unet_path, map_location="cpu"),
)
denoising_unet.load_state_dict(
torch.load(args.denoising_unet_path, map_location="cpu"),
strict=False,
)
tokenizer = CLIPTokenizer.from_pretrained(args.pretrained_model_name_or_path,subfolder='tokenizer')
scheduler = DDIMScheduler.from_pretrained(args.pretrained_model_name_or_path,subfolder='scheduler')
pipe = DepthLabPipeline(reference_unet=reference_unet,
denoising_unet = denoising_unet,
mapping_layer=mapping_layer,
vae=vae,
text_encoder=text_encoder,
tokenizer=tokenizer,
image_enc=image_enc,
scheduler=scheduler,
).to('cuda')
try:
pipe.enable_xformers_memory_efficient_attention()
except ImportError:
logging.debug("run without xformers")
# -------------------- Inference and saving --------------------
with torch.no_grad():
for i in range(len(input_image_paths)):
input_image_path = input_image_paths[i]
mask_path = masks_paths[i]
known_depth_path = known_depth_paths[i]
# save path
rgb_name_base = os.path.splitext(os.path.basename(input_image_path))[0]
pred_name_base = rgb_name_base + "_pred"
npy_save_path = os.path.join(output_dir_npy, f"{pred_name_base}.npy")
colored_save_path = os.path.join(
output_dir_color, f"{pred_name_base}_colored.png"
)
input_image = Image.open(input_image_path)
try:
mask = np.load(mask_path)
mask[mask>0.5]=1
mask[mask<0.5]=0
except:
mask = load_and_process_mask(mask_path)
depth_numpy=np.load(known_depth_path)
if args.refine is not True:
depth_numpy=get_filled_for_latents(mask,depth_numpy)
pipe_out = pipe(
input_image,
denosing_steps = denoise_steps,
processing_res = processing_res,
match_input_res = True,
batch_size =1,
color_map = "Spectral",
show_progress_bar = True,
depth_numpy_origin = depth_numpy,
mask_origin = mask,
guidance_scale = 1,
normalize_scale = args.normalize_scale,
strength = args.strength,
blend = args.blend)
depth_pred: np.ndarray = pipe_out.depth_np
if os.path.exists(colored_save_path):
logging.warning(f"Existing file: '{colored_save_path}' will be overwritten")
np.save(npy_save_path,depth_pred)
pipe_out.depth_colored.save(colored_save_path)
================================================
FILE: inference/depthlab_pipeline.py
================================================
import os, sys
parent_dir = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
if parent_dir not in sys.path:
sys.path.insert(0, parent_dir)
from typing import Any, Dict, Union
import torch.nn.functional as F
import torch
from torch.utils.data import DataLoader, TensorDataset
import numpy as np
from tqdm.auto import tqdm
from PIL import Image
import torch.nn as nn
from diffusers import (
DiffusionPipeline,
DDIMScheduler,
AutoencoderKL,
)
from transformers import CLIPImageProcessor
from transformers import CLIPVisionModelWithProjection
from src.models.mutual_self_attention import ReferenceAttentionControl
from src.models.unet_2d_condition import UNet2DConditionModel
from src.models.unet_2d_condition_main import UNet2DConditionModel_main
from src.models.projection import My_proj
import cv2
from diffusers.utils import BaseOutput
from transformers import CLIPTextModel, CLIPTokenizer
from utils.image_util import resize_max_res, resize_max_res_cv2, colorize_depth_maps, chw2hwc, Disparity_Normalization_mask_scale,get_filled_depth
from scipy.interpolate import griddata
sys.path.pop(0)
class DepthPipelineOutput(BaseOutput):
"""
Output class for Marigold monocular depth prediction pipeline.
Args:
depth_np (`np.ndarray`):
Predicted depth map, with depth values in the range of [0, 1].
depth_colored (`PIL.Image.Image`):
Colorized depth map, with the shape of [3, H, W] and values in [0, 1].
uncertainty (`None` or `np.ndarray`):
Uncalibrated uncertainty(MAD, median absolute deviation) coming from ensembling.
"""
depth_np: np.ndarray
depth_norm: Image.Image
depth_colored: Image.Image
depth_pred_numpy_origin:np.ndarray
class DepthLabPipeline(DiffusionPipeline):
# two hyper-parameters
rgb_latent_scale_factor = 0.18215
depth_latent_scale_factor = 0.18215
def __init__(self,
reference_unet:UNet2DConditionModel,
denoising_unet:UNet2DConditionModel_main,
mapping_layer:My_proj,
vae:AutoencoderKL,
text_encoder:CLIPTextModel,
tokenizer:CLIPTokenizer,
image_enc:CLIPVisionModelWithProjection,
scheduler:DDIMScheduler,
):
super().__init__()
self.register_modules(
reference_unet=reference_unet,
denoising_unet=denoising_unet,
mapping_layer=mapping_layer,
vae=vae,
image_enc=image_enc,
scheduler=scheduler,
tokenizer=tokenizer,
text_encoder=text_encoder
)
self.empty_text_embed = None
self.clip_image_processor = CLIPImageProcessor()
@torch.no_grad()
def __call__(self,
input_image: str,
denosing_steps: int = 20,
processing_res: int = 768,
match_input_res: bool = True,
batch_size: int = 0,
color_map: str="Spectral",
show_progress_bar: bool = True,
ensemble_kwargs: Dict = None,
depth_numpy_origin = None,
mask_origin = None,
guidance_scale = 1,
normalize_scale = 1,
strength=0.8,
blend=True
) -> DepthPipelineOutput:
# inherit from thea Diffusion Pipeline
device = self.image_enc.device
clip_image_unresize = input_image
try:
depth_origin = depth_numpy_origin
except:
raise NotImplementedError
assert depth_origin.min() >= 0.
mask_origin = np.array(mask_origin)
mask_origin [mask_origin <0.5] = 0.
mask_origin [mask_origin >0.5] = 1.
clip_image = self.clip_image_processor.preprocess(
clip_image_unresize, return_tensors="pt"
).pixel_values
clip_image_embeds = self.image_enc(
clip_image.to(device, dtype=self.image_enc.dtype)
).image_embeds
encoder_hidden_states = clip_image_embeds.unsqueeze(1)
encoder_hidden_states =self.mapping_layer(encoder_hidden_states )
prompt = ""
text_inputs =self.tokenizer(
prompt,
padding="do_not_pad",
max_length=self.tokenizer.model_max_length,
truncation=True,
return_tensors="pt",
)
text_input_ids = text_inputs.input_ids.to(device)
empty_text_embed = self.text_encoder(text_input_ids)[0].to(self.dtype)
uncond_encoder_hidden_states = empty_text_embed.repeat((1, 1, 1))[:,0,:].unsqueeze(0)
do_classifier_free_guidance = True
# adjust the input resolution.
if not match_input_res:
assert (
processing_res is not None
)," Value Error: `resize_output_back` is only valid with "
assert processing_res >= 0
assert denosing_steps >= 1
# --------------- Image Processing ------------------------
# Resize image
original_H, original_W = depth_origin.shape
if processing_res > 0:
image = resize_max_res(
input_image, max_edge_resolution=processing_res, resample=Image.BICUBIC
)
depth = resize_max_res_cv2(
depth_origin, max_edge_resolution=processing_res, interpolation=cv2.INTER_LINEAR
)
mask = resize_max_res_cv2(
mask_origin , max_edge_resolution=processing_res, interpolation=cv2.INTER_NEAREST
)
# Convert the image to RGB, to 1. reomve the alpha channel.
image = np.array(image)
# Normalize RGB Values.
rgb = np.transpose(image,(2,0,1))
rgb_norm = rgb / 255.0 * 2.0 - 1.0
rgb_norm = torch.from_numpy(rgb_norm).to(self.dtype).to(device)
assert rgb_norm.min() >= -1.0 and rgb_norm.max() <= 1.0
# depth
depth = torch.from_numpy(depth).to(self.dtype).to(device)[None]
assert depth.min() >= 0.
# mask
mask = torch.from_numpy(mask).to(self.dtype).to(device)[None]
assert mask.min() >= 0. and mask.max() <= 1.
# ----------------- predicting depth -----------------
single_rgb_dataset = TensorDataset(rgb_norm[None], depth[None], mask[None])
# find the batch size
if batch_size>0:
_bs = batch_size
else:
_bs = 1
single_rgb_loader = DataLoader(single_rgb_dataset,batch_size=_bs,shuffle=False)
# classifier guidance
if do_classifier_free_guidance:
encoder_hidden_states = torch.cat(
[uncond_encoder_hidden_states, encoder_hidden_states], dim=0
)
reference_control_writer = ReferenceAttentionControl(
self.reference_unet,
do_classifier_free_guidance=do_classifier_free_guidance,
mode="write",
batch_size=batch_size,
fusion_blocks="full",
)
reference_control_reader = ReferenceAttentionControl(
self.denoising_unet,
do_classifier_free_guidance=do_classifier_free_guidance,
mode="read",
batch_size=batch_size,
fusion_blocks="full",
)
if show_progress_bar:
iterable_bar = tqdm(
single_rgb_loader, desc=" " * 2 + "Inference batches", leave=False
)
else:
iterable_bar = single_rgb_loader
for batch in iterable_bar:
(image, depth, mask)= batch
depth_pred_raw, max_value, min_value = self.single_infer(
image = image,
depth = depth,
mask = mask,
num_inference_steps = denosing_steps,
show_pbar = show_progress_bar,
guidance_scale = guidance_scale,
encoder_hidden_states = encoder_hidden_states,
reference_control_writer = reference_control_writer,
reference_control_reader = reference_control_reader,
strength = strength,
blend = blend,
normalize_scale = normalize_scale,
generator = None,
)
depth_pred = depth_pred_raw
torch.cuda.empty_cache()
# ----------------- Post processing -----------------
depth_pred = (depth_pred * (max_value - min_value) + min_value)
depth_pred_numpy = depth_pred.detach().cpu().numpy().squeeze()
depth_pred_numpy = cv2.resize(depth_pred_numpy.astype(float), (original_W, original_H))
depth_pred_numpy = depth_pred_numpy.clip(min=0.)
depth_pred_numpy_origin=depth_pred_numpy.copy()
depth_pred_norm = (depth_pred_numpy - depth_pred_numpy.min()) / (depth_pred_numpy.max() - depth_pred_numpy.min())
depth_pred_colored = colorize_depth_maps(
depth_pred_norm, 0, 1, cmap="Spectral"
).squeeze() # [3, H, W], value in (0, 1)
depth_pred_colored = (depth_pred_colored * 255).astype(np.uint8)
depth_pred_colored = Image.fromarray(chw2hwc(depth_pred_colored))
return DepthPipelineOutput(
depth_np = depth_pred_numpy,
depth_norm = depth_pred_norm,
depth_colored = depth_pred_colored,
depth_pred_numpy_origin=depth_pred_numpy_origin
)
def get_timesteps(self, num_inference_steps, strength, device):
# get the original timestep using init_timestep
init_timestep = min(int(num_inference_steps * strength), num_inference_steps)
t_start = max(num_inference_steps - init_timestep, 0)
timesteps = self.scheduler.timesteps[t_start * self.scheduler.order :]
return timesteps, num_inference_steps - t_start
@torch.no_grad()
def single_infer(self,
image: torch.Tensor,
depth: torch.Tensor,
mask: torch.Tensor,
num_inference_steps: int,
show_pbar: bool,
guidance_scale: float,
encoder_hidden_states: torch.Tensor,
reference_control_writer: ReferenceAttentionControl,
reference_control_reader: ReferenceAttentionControl,
strength: float,
blend: bool,
normalize_scale: float,
generator=None,
):
do_classifier_free_guidance = True
try:
device = self.image_enc.device
except:
import pdb; pdb.set_trace()
h, w = image.shape[-2:]
# Set timesteps: inherit from the diffuison pipeline
self.scheduler.set_timesteps(num_inference_steps, device=device)
timesteps, _ = self.get_timesteps(num_inference_steps, strength, device)
# Encode image
rgb_latent = self.encode_RGB(image) #
min_value = depth[mask == 0].min()
max_value = depth[mask == 0].max()
depth = Disparity_Normalization_mask_scale(depth, min_value, max_value, scale=normalize_scale)
masked_depth = depth.clone()
masked_depth[mask == 1] = 0
masked_depth = get_filled_depth(masked_depth.squeeze().cpu().numpy(), mask.squeeze().cpu().numpy(),method='linear')
masked_depth = torch.from_numpy(masked_depth).unsqueeze(0).unsqueeze(0).float().to(device)
masked_depth = masked_depth.repeat(1,3,1,1)
masked_depth_latent = self.encode_depth(masked_depth)
depth = depth.repeat(1,3,1,1)
depth_latent = self.encode_depth(depth)
# process mask
mask_latents=self.encode_RGB(mask.repeat(1,3,1,1).to(rgb_latent.dtype) * 2 - 1)
mask_down = torch.nn.functional.interpolate(mask, size=(h//8, w//8),mode='nearest')
noise = torch.randn_like(depth_latent)
if strength < 1.:
noisy_depth_latent = self.scheduler.add_noise(depth_latent, noise, timesteps[:1])
else:
noisy_depth_latent = torch.randn(
depth_latent.shape,
device=device,
dtype=depth_latent.dtype,
generator=generator,
)
# Denoising loop
if show_pbar:
iterable = tqdm(
enumerate(timesteps),
total=len(timesteps),
leave=False,
desc=" " * 4 + "Diffusion denoising",
)
else:
iterable = enumerate(timesteps)
for i, t in iterable:
if i == 0:
self.reference_unet(
rgb_latent.repeat(
(2 if do_classifier_free_guidance else 1), 1, 1, 1
),
torch.zeros_like(t),
encoder_hidden_states=encoder_hidden_states,
return_dict=False,
)
reference_control_reader.update(reference_control_writer)
unet_input = torch.cat([noisy_depth_latent, mask_latents, masked_depth_latent], dim=1)
noise_pred = self.denoising_unet(
unet_input.repeat(
(2 if do_classifier_free_guidance else 1), 1, 1, 1
).to(dtype=self.denoising_unet.dtype), t, encoder_hidden_states=encoder_hidden_states
).sample # [B, 4, h, w]
if do_classifier_free_guidance:
noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
noise_pred = noise_pred_uncond + guidance_scale * (
noise_pred_text - noise_pred_uncond
)
# compute the previous noisy sample x_t -> x_t-1
noisy_depth_latent = self.scheduler.step(noise_pred, t, noisy_depth_latent).prev_sample.to(self.dtype)
if blend:
# Blend diffusion https://arxiv.org/abs/2111.14818
if i < len(timesteps) - 1:
noise_timestep = timesteps[i + 1]
depth_latent_step = self.scheduler.add_noise(
depth_latent, noise, torch.tensor([noise_timestep])
)
mask_blend = mask_down.repeat(1,4,1,1).float()
noisy_depth_latent = (1 - mask_blend) * depth_latent_step + mask_blend * noisy_depth_latent
reference_control_reader.clear()
reference_control_writer.clear()
torch.cuda.empty_cache()
depth = self.decode_depth(noisy_depth_latent)
# depth = torch.clip(depth, -1.0, 1.0)
# shift
depth = (depth + normalize_scale) / (normalize_scale*2)
return depth, max_value, min_value
def encode_RGB(self, rgb_in: torch.Tensor) -> torch.Tensor:
"""
Encode RGB image into latent.
Args:
rgb_in (`torch.Tensor`):
Input RGB image to be encoded.
Returns:
`torch.Tensor`: Image latent.
"""
# encode
h = self.vae.encoder(rgb_in)
moments = self.vae.quant_conv(h)
mean, logvar = torch.chunk(moments, 2, dim=1)
rgb_latent = mean * self.rgb_latent_scale_factor
return rgb_latent
def decode_depth(self, depth_latent: torch.Tensor) -> torch.Tensor:
"""
Decode depth latent into depth map.
Args:
depth_latent (`torch.Tensor`):
Depth latent to be decoded.
Returns:
`torch.Tensor`: Decoded depth map.
"""
depth_latent = depth_latent / self.depth_latent_scale_factor
depth_latent = depth_latent
try:
z = self.vae.post_quant_conv(depth_latent)
stacked = self.vae.decoder(z)
except:
stacked = self.vae.decode(depth_latent)
# mean of output channels
depth_mean = stacked.mean(dim=1, keepdim=True)
return depth_mean
def encode_depth(self, depth_latent: torch.Tensor) -> torch.Tensor:
"""
Decode depth latent into depth map.
Args:
depth_latent (`torch.Tensor`):
Depth latent to be decoded.
Returns:
`torch.Tensor`: Decoded depth map.
"""
# scale latent
h_disp = self.vae.encoder(depth_latent)
moments_disp = self.vae.quant_conv(h_disp)
mean_disp, logvar_disp = torch.chunk(moments_disp, 2, dim=1)
disp_latents = mean_disp *self.depth_latent_scale_factor
return disp_latents
================================================
FILE: scripts/infer.sh
================================================
#!/usr/bin/env bash
set -e
set -x
pretrained_model_name_or_path='./checkpoints/marigold-depth-v1-0'
image_encoder_path='./checkpoints/CLIP-ViT-H-14-laion2B-s32B-b79K'
denoising_unet_path='./checkpoints/DepthLab/denoising_unet.pth'
reference_unet_path='./checkpoints/DepthLab/reference_unet.pth'
mapping_path='./checkpoints/DepthLab/mapping_layer.pth'
export CUDA_VISIBLE_DEVICES=0
cd ..
python infer.py \
--seed 1234 \
--denoise_steps 50 \
--processing_res 768 \
--normalize_scale 1 \
--strength 0.8 \
--pretrained_model_name_or_path $pretrained_model_name_or_path --image_encoder_path $image_encoder_path \
--denoising_unet_path $denoising_unet_path \
--reference_unet_path $reference_unet_path \
--mapping_path $mapping_path \
--output_dir 'output/in-the-wild_example' \
--input_image_paths test_cases/RGB.JPG \
--known_depth_paths test_cases/know_depth.npy \
--masks_paths test_cases/mask.npy \
--blend
================================================
FILE: src/models/attention.py
================================================
# Adapted from https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention.py
from typing import Any, Dict, Optional
import torch
from diffusers.models.attention import AdaLayerNorm, Attention, FeedForward
from diffusers.models.embeddings import SinusoidalPositionalEmbedding
from einops import rearrange
from torch import nn
class BasicTransformerBlock(nn.Module):
r"""
A basic Transformer block.
Parameters:
dim (`int`): The number of channels in the input and output.
num_attention_heads (`int`): The number of heads to use for multi-head attention.
attention_head_dim (`int`): The number of channels in each head.
dropout (`float`, *optional*, defaults to 0.0): The dropout probability to use.
cross_attention_dim (`int`, *optional*): The size of the encoder_hidden_states vector for cross attention.
activation_fn (`str`, *optional*, defaults to `"geglu"`): Activation function to be used in feed-forward.
num_embeds_ada_norm (:
obj: `int`, *optional*): The number of diffusion steps used during training. See `Transformer2DModel`.
attention_bias (:
obj: `bool`, *optional*, defaults to `False`): Configure if the attentions should contain a bias parameter.
only_cross_attention (`bool`, *optional*):
Whether to use only cross-attention layers. In this case two cross attention layers are used.
double_self_attention (`bool`, *optional*):
Whether to use two self-attention layers. In this case no cross attention layers are used.
upcast_attention (`bool`, *optional*):
Whether to upcast the attention computation to float32. This is useful for mixed precision training.
norm_elementwise_affine (`bool`, *optional*, defaults to `True`):
Whether to use learnable elementwise affine parameters for normalization.
norm_type (`str`, *optional*, defaults to `"layer_norm"`):
The normalization layer to use. Can be `"layer_norm"`, `"ada_norm"` or `"ada_norm_zero"`.
final_dropout (`bool` *optional*, defaults to False):
Whether to apply a final dropout after the last feed-forward layer.
attention_type (`str`, *optional*, defaults to `"default"`):
The type of attention to use. Can be `"default"` or `"gated"` or `"gated-text-image"`.
positional_embeddings (`str`, *optional*, defaults to `None`):
The type of positional embeddings to apply to.
num_positional_embeddings (`int`, *optional*, defaults to `None`):
The maximum number of positional embeddings to apply.
"""
def __init__(
self,
dim: int,
num_attention_heads: int,
attention_head_dim: int,
dropout=0.0,
cross_attention_dim: Optional[int] = None,
activation_fn: str = "geglu",
num_embeds_ada_norm: Optional[int] = None,
attention_bias: bool = False,
only_cross_attention: bool = False,
double_self_attention: bool = False,
upcast_attention: bool = False,
norm_elementwise_affine: bool = True,
norm_type: str = "layer_norm", # 'layer_norm', 'ada_norm', 'ada_norm_zero', 'ada_norm_single'
norm_eps: float = 1e-5,
final_dropout: bool = False,
attention_type: str = "default",
positional_embeddings: Optional[str] = None,
num_positional_embeddings: Optional[int] = None,
):
super().__init__()
self.only_cross_attention = only_cross_attention
self.use_ada_layer_norm_zero = (
num_embeds_ada_norm is not None
) and norm_type == "ada_norm_zero"
self.use_ada_layer_norm = (
num_embeds_ada_norm is not None
) and norm_type == "ada_norm"
self.use_ada_layer_norm_single = norm_type == "ada_norm_single"
self.use_layer_norm = norm_type == "layer_norm"
if norm_type in ("ada_norm", "ada_norm_zero") and num_embeds_ada_norm is None:
raise ValueError(
f"`norm_type` is set to {norm_type}, but `num_embeds_ada_norm` is not defined. Please make sure to"
f" define `num_embeds_ada_norm` if setting `norm_type` to {norm_type}."
)
if positional_embeddings and (num_positional_embeddings is None):
raise ValueError(
"If `positional_embedding` type is defined, `num_positition_embeddings` must also be defined."
)
if positional_embeddings == "sinusoidal":
self.pos_embed = SinusoidalPositionalEmbedding(
dim, max_seq_length=num_positional_embeddings
)
else:
self.pos_embed = None
# Define 3 blocks. Each block has its own normalization layer.
# 1. Self-Attn
if self.use_ada_layer_norm:
self.norm1 = AdaLayerNorm(dim, num_embeds_ada_norm)
elif self.use_ada_layer_norm_zero:
self.norm1 = AdaLayerNormZero(dim, num_embeds_ada_norm)
else:
self.norm1 = nn.LayerNorm(
dim, elementwise_affine=norm_elementwise_affine, eps=norm_eps
)
self.attn1 = Attention(
query_dim=dim,
heads=num_attention_heads,
dim_head=attention_head_dim,
dropout=dropout,
bias=attention_bias,
cross_attention_dim=cross_attention_dim if only_cross_attention else None,
upcast_attention=upcast_attention,
)
# 2. Cross-Attn
if cross_attention_dim is not None or double_self_attention:
# We currently only use AdaLayerNormZero for self attention where there will only be one attention block.
# I.e. the number of returned modulation chunks from AdaLayerZero would not make sense if returned during
# the second cross attention block.
self.norm2 = (
AdaLayerNorm(dim, num_embeds_ada_norm)
if self.use_ada_layer_norm
else nn.LayerNorm(
dim, elementwise_affine=norm_elementwise_affine, eps=norm_eps
)
)
self.attn2 = Attention(
query_dim=dim,
cross_attention_dim=cross_attention_dim
if not double_self_attention
else None,
heads=num_attention_heads,
dim_head=attention_head_dim,
dropout=dropout,
bias=attention_bias,
upcast_attention=upcast_attention,
) # is self-attn if encoder_hidden_states is none
else:
self.norm2 = None
self.attn2 = None
# 3. Feed-forward
if not self.use_ada_layer_norm_single:
self.norm3 = nn.LayerNorm(
dim, elementwise_affine=norm_elementwise_affine, eps=norm_eps
)
self.ff = FeedForward(
dim,
dropout=dropout,
activation_fn=activation_fn,
final_dropout=final_dropout,
)
# 4. Fuser
if attention_type == "gated" or attention_type == "gated-text-image":
self.fuser = GatedSelfAttentionDense(
dim, cross_attention_dim, num_attention_heads, attention_head_dim
)
# 5. Scale-shift for PixArt-Alpha.
if self.use_ada_layer_norm_single:
self.scale_shift_table = nn.Parameter(torch.randn(6, dim) / dim**0.5)
# let chunk size default to None
self._chunk_size = None
self._chunk_dim = 0
def set_chunk_feed_forward(self, chunk_size: Optional[int], dim: int = 0):
# Sets chunk feed-forward
self._chunk_size = chunk_size
self._chunk_dim = dim
def forward(
self,
hidden_states: torch.FloatTensor,
attention_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
timestep: Optional[torch.LongTensor] = None,
cross_attention_kwargs: Dict[str, Any] = None,
class_labels: Optional[torch.LongTensor] = None,
) -> torch.FloatTensor:
# Notice that normalization is always applied before the real computation in the following blocks.
# 0. Self-Attention
batch_size = hidden_states.shape[0]
if self.use_ada_layer_norm:
norm_hidden_states = self.norm1(hidden_states, timestep)
elif self.use_ada_layer_norm_zero:
norm_hidden_states, gate_msa, shift_mlp, scale_mlp, gate_mlp = self.norm1(
hidden_states, timestep, class_labels, hidden_dtype=hidden_states.dtype
)
elif self.use_layer_norm:
norm_hidden_states = self.norm1(hidden_states)
elif self.use_ada_layer_norm_single:
shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = (
self.scale_shift_table[None] + timestep.reshape(batch_size, 6, -1)
).chunk(6, dim=1)
norm_hidden_states = self.norm1(hidden_states)
norm_hidden_states = norm_hidden_states * (1 + scale_msa) + shift_msa
norm_hidden_states = norm_hidden_states.squeeze(1)
else:
raise ValueError("Incorrect norm used")
if self.pos_embed is not None:
norm_hidden_states = self.pos_embed(norm_hidden_states)
# 1. Retrieve lora scale.
lora_scale = (
cross_attention_kwargs.get("scale", 1.0)
if cross_attention_kwargs is not None
else 1.0
)
# 2. Prepare GLIGEN inputs
cross_attention_kwargs = (
cross_attention_kwargs.copy() if cross_attention_kwargs is not None else {}
)
gligen_kwargs = cross_attention_kwargs.pop("gligen", None)
attn_output = self.attn1(
norm_hidden_states,
encoder_hidden_states=encoder_hidden_states
if self.only_cross_attention
else None,
attention_mask=attention_mask,
**cross_attention_kwargs,
)
if self.use_ada_layer_norm_zero:
attn_output = gate_msa.unsqueeze(1) * attn_output
elif self.use_ada_layer_norm_single:
attn_output = gate_msa * attn_output
hidden_states = attn_output + hidden_states
if hidden_states.ndim == 4:
hidden_states = hidden_states.squeeze(1)
# 2.5 GLIGEN Control
if gligen_kwargs is not None:
hidden_states = self.fuser(hidden_states, gligen_kwargs["objs"])
# 3. Cross-Attention
if self.attn2 is not None:
if self.use_ada_layer_norm:
norm_hidden_states = self.norm2(hidden_states, timestep)
elif self.use_ada_layer_norm_zero or self.use_layer_norm:
norm_hidden_states = self.norm2(hidden_states)
elif self.use_ada_layer_norm_single:
# For PixArt norm2 isn't applied here:
# https://github.com/PixArt-alpha/PixArt-alpha/blob/0f55e922376d8b797edd44d25d0e7464b260dcab/diffusion/model/nets/PixArtMS.py#L70C1-L76C103
norm_hidden_states = hidden_states
else:
raise ValueError("Incorrect norm")
if self.pos_embed is not None and self.use_ada_layer_norm_single is False:
norm_hidden_states = self.pos_embed(norm_hidden_states)
attn_output = self.attn2(
norm_hidden_states,
encoder_hidden_states=encoder_hidden_states,
attention_mask=encoder_attention_mask,
**cross_attention_kwargs,
)
hidden_states = attn_output + hidden_states
# 4. Feed-forward
if not self.use_ada_layer_norm_single:
norm_hidden_states = self.norm3(hidden_states)
if self.use_ada_layer_norm_zero:
norm_hidden_states = (
norm_hidden_states * (1 + scale_mlp[:, None]) + shift_mlp[:, None]
)
if self.use_ada_layer_norm_single:
norm_hidden_states = self.norm2(hidden_states)
norm_hidden_states = norm_hidden_states * (1 + scale_mlp) + shift_mlp
ff_output = self.ff(norm_hidden_states, scale=lora_scale)
if self.use_ada_layer_norm_zero:
ff_output = gate_mlp.unsqueeze(1) * ff_output
elif self.use_ada_layer_norm_single:
ff_output = gate_mlp * ff_output
hidden_states = ff_output + hidden_states
if hidden_states.ndim == 4:
hidden_states = hidden_states.squeeze(1)
return hidden_states
class TemporalBasicTransformerBlock(nn.Module):
def __init__(
self,
dim: int,
num_attention_heads: int,
attention_head_dim: int,
dropout=0.0,
cross_attention_dim: Optional[int] = None,
activation_fn: str = "geglu",
num_embeds_ada_norm: Optional[int] = None,
attention_bias: bool = False,
only_cross_attention: bool = False,
upcast_attention: bool = False,
unet_use_cross_frame_attention=None,
unet_use_temporal_attention=None,
):
super().__init__()
self.only_cross_attention = only_cross_attention
self.use_ada_layer_norm = num_embeds_ada_norm is not None
self.unet_use_cross_frame_attention = unet_use_cross_frame_attention
self.unet_use_temporal_attention = unet_use_temporal_attention
# SC-Attn
self.attn1 = Attention(
query_dim=dim,
heads=num_attention_heads,
dim_head=attention_head_dim,
dropout=dropout,
bias=attention_bias,
upcast_attention=upcast_attention,
)
self.norm1 = (
AdaLayerNorm(dim, num_embeds_ada_norm)
if self.use_ada_layer_norm
else nn.LayerNorm(dim)
)
# Cross-Attn
if cross_attention_dim is not None:
self.attn2 = Attention(
query_dim=dim,
cross_attention_dim=cross_attention_dim,
heads=num_attention_heads,
dim_head=attention_head_dim,
dropout=dropout,
bias=attention_bias,
upcast_attention=upcast_attention,
)
else:
self.attn2 = None
if cross_attention_dim is not None:
self.norm2 = (
AdaLayerNorm(dim, num_embeds_ada_norm)
if self.use_ada_layer_norm
else nn.LayerNorm(dim)
)
else:
self.norm2 = None
# Feed-forward
self.ff = FeedForward(dim, dropout=dropout, activation_fn=activation_fn)
self.norm3 = nn.LayerNorm(dim)
self.use_ada_layer_norm_zero = False
# Temp-Attn
assert unet_use_temporal_attention is not None
if unet_use_temporal_attention:
self.attn_temp = Attention(
query_dim=dim,
heads=num_attention_heads,
dim_head=attention_head_dim,
dropout=dropout,
bias=attention_bias,
upcast_attention=upcast_attention,
)
nn.init.zeros_(self.attn_temp.to_out[0].weight.data)
self.norm_temp = (
AdaLayerNorm(dim, num_embeds_ada_norm)
if self.use_ada_layer_norm
else nn.LayerNorm(dim)
)
def forward(
self,
hidden_states,
encoder_hidden_states=None,
timestep=None,
attention_mask=None,
video_length=None,
):
norm_hidden_states = (
self.norm1(hidden_states, timestep)
if self.use_ada_layer_norm
else self.norm1(hidden_states)
)
if self.unet_use_cross_frame_attention:
hidden_states = (
self.attn1(
norm_hidden_states,
attention_mask=attention_mask,
video_length=video_length,
)
+ hidden_states
)
else:
hidden_states = (
self.attn1(norm_hidden_states, attention_mask=attention_mask)
+ hidden_states
)
if self.attn2 is not None:
# Cross-Attention
norm_hidden_states = (
self.norm2(hidden_states, timestep)
if self.use_ada_layer_norm
else self.norm2(hidden_states)
)
hidden_states = (
self.attn2(
norm_hidden_states,
encoder_hidden_states=encoder_hidden_states,
attention_mask=attention_mask,
)
+ hidden_states
)
# Feed-forward
hidden_states = self.ff(self.norm3(hidden_states)) + hidden_states
# Temporal-Attention
if self.unet_use_temporal_attention:
d = hidden_states.shape[1]
hidden_states = rearrange(
hidden_states, "(b f) d c -> (b d) f c", f=video_length
)
norm_hidden_states = (
self.norm_temp(hidden_states, timestep)
if self.use_ada_layer_norm
else self.norm_temp(hidden_states)
)
hidden_states = self.attn_temp(norm_hidden_states) + hidden_states
hidden_states = rearrange(hidden_states, "(b d) f c -> (b f) d c", d=d)
return hidden_states
================================================
FILE: src/models/mutual_self_attention.py
================================================
# Adapted from https://github.com/magic-research/magic-animate/blob/main/magicanimate/models/mutual_self_attention.py
from typing import Any, Dict, Optional
import torch
from einops import rearrange
from .attention import TemporalBasicTransformerBlock
from .attention import BasicTransformerBlock
def torch_dfs(model: torch.nn.Module):
result = [model]
for child in model.children():
result += torch_dfs(child)
return result
class ReferenceAttentionControl:
def __init__(
self,
unet,
mode="write",
do_classifier_free_guidance=False,
attention_auto_machine_weight=float("inf"),
gn_auto_machine_weight=1.0,
style_fidelity=1.0,
reference_attn=True,
reference_adain=False,
fusion_blocks="midup",
batch_size=1,
) -> None:
# 10. Modify self attention and group norm
self.unet = unet
assert mode in ["read", "write"]
assert fusion_blocks in ["midup", "full"]
self.reference_attn = reference_attn
self.reference_adain = reference_adain
self.fusion_blocks = fusion_blocks
self.register_reference_hooks(
mode,
do_classifier_free_guidance,
attention_auto_machine_weight,
gn_auto_machine_weight,
style_fidelity,
reference_attn,
reference_adain,
fusion_blocks,
batch_size=batch_size,
)
def register_reference_hooks(
self,
mode,
do_classifier_free_guidance,
attention_auto_machine_weight,
gn_auto_machine_weight,
style_fidelity,
reference_attn,
reference_adain,
dtype=torch.float16,
batch_size=1,
num_images_per_prompt=1,
device=torch.device("cpu"),
fusion_blocks="midup",
):
MODE = mode
do_classifier_free_guidance = do_classifier_free_guidance
attention_auto_machine_weight = attention_auto_machine_weight
gn_auto_machine_weight = gn_auto_machine_weight
style_fidelity = style_fidelity
reference_attn = reference_attn
reference_adain = reference_adain
fusion_blocks = fusion_blocks
num_images_per_prompt = num_images_per_prompt
dtype = dtype
if do_classifier_free_guidance:
uc_mask = (
torch.Tensor(
[1] * batch_size * num_images_per_prompt * 16
+ [0] * batch_size * num_images_per_prompt * 16
)
.to(device)
.bool()
)
else:
uc_mask = (
torch.Tensor([0] * batch_size * num_images_per_prompt * 2)
.to(device)
.bool()
)
def hacked_basic_transformer_inner_forward(
self,
hidden_states: torch.FloatTensor,
attention_mask: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
timestep: Optional[torch.LongTensor] = None,
cross_attention_kwargs: Dict[str, Any] = None,
class_labels: Optional[torch.LongTensor] = None,
video_length=None,
):
if self.use_ada_layer_norm: # False
norm_hidden_states = self.norm1(hidden_states, timestep)
elif self.use_ada_layer_norm_zero:
(
norm_hidden_states,
gate_msa,
shift_mlp,
scale_mlp,
gate_mlp,
) = self.norm1(
hidden_states,
timestep,
class_labels,
hidden_dtype=hidden_states.dtype,
)
else:
norm_hidden_states = self.norm1(hidden_states)
# 1. Self-Attention
# self.only_cross_attention = False
cross_attention_kwargs = (
cross_attention_kwargs if cross_attention_kwargs is not None else {}
)
if self.only_cross_attention:
attn_output = self.attn1(
norm_hidden_states,
encoder_hidden_states=encoder_hidden_states
if self.only_cross_attention
else None,
attention_mask=attention_mask,
**cross_attention_kwargs,
)
else:
if MODE == "write":
self.bank.append(norm_hidden_states.clone())
attn_output = self.attn1(
norm_hidden_states,
encoder_hidden_states=encoder_hidden_states
if self.only_cross_attention
else None,
attention_mask=attention_mask,
**cross_attention_kwargs,
)
if MODE == "read":
# if len(self.bank)>0:
# import ipdb;ipdb.set_trace()
# bank_fea = [ d for d in self.bank]
bank_fea = [
rearrange(
d.unsqueeze(1).repeat(1, 1, 1, 1),
"b t l c -> (b t) l c",
)
for d in self.bank
]
modify_norm_hidden_states = torch.cat(
[norm_hidden_states] + bank_fea, dim=1
)
hidden_states_uc = (
self.attn1(
norm_hidden_states,
encoder_hidden_states=modify_norm_hidden_states,
attention_mask=attention_mask,
)
+ hidden_states
)
if do_classifier_free_guidance:
hidden_states_c = hidden_states_uc.clone()
_uc_mask = uc_mask.clone()
if hidden_states.shape[0] != _uc_mask.shape[0]:
_uc_mask = (
torch.Tensor(
[1] * (hidden_states.shape[0] // 2)
+ [0] * (hidden_states.shape[0] // 2)
)
.to(device)
.bool()
)
hidden_states_c[_uc_mask] = (
self.attn1(
norm_hidden_states[_uc_mask],
encoder_hidden_states=norm_hidden_states[_uc_mask],
attention_mask=attention_mask,
)
+ hidden_states[_uc_mask]
)
hidden_states = hidden_states_c.clone()
else:
hidden_states = hidden_states_uc
if self.attn2 is not None:
# Cross-Attention
norm_hidden_states = (
self.norm2(hidden_states, timestep)
if self.use_ada_layer_norm
else self.norm2(hidden_states)
)
hidden_states = (
self.attn2(
norm_hidden_states,
encoder_hidden_states=encoder_hidden_states,
attention_mask=attention_mask,
)
+ hidden_states
)
# Feed-forward
hidden_states = self.ff(self.norm3(hidden_states)) + hidden_states
return hidden_states
# import ipdb;ipdb.set_trace()
if self.use_ada_layer_norm_zero:
attn_output = gate_msa.unsqueeze(1) * attn_output
try:
hidden_states = attn_output + hidden_states
except:
import ipdb;ipdb.set_trace()
if self.attn2 is not None:
norm_hidden_states = (
self.norm2(hidden_states, timestep)
if self.use_ada_layer_norm
else self.norm2(hidden_states)
)
# 2. Cross-Attention
attn_output = self.attn2(
norm_hidden_states,
encoder_hidden_states=encoder_hidden_states,
attention_mask=encoder_attention_mask,
**cross_attention_kwargs,
)
hidden_states = attn_output + hidden_states
# 3. Feed-forward
norm_hidden_states = self.norm3(hidden_states)
if self.use_ada_layer_norm_zero:
norm_hidden_states = (
norm_hidden_states * (1 + scale_mlp[:, None]) + shift_mlp[:, None]
)
ff_output = self.ff(norm_hidden_states)
if self.use_ada_layer_norm_zero:
ff_output = gate_mlp.unsqueeze(1) * ff_output
hidden_states = ff_output + hidden_states
return hidden_states
if self.reference_attn:
if self.fusion_blocks == "midup":
attn_modules = [
module
for module in (
torch_dfs(self.unet.mid_block) + torch_dfs(self.unet.up_blocks)
)
if isinstance(module, BasicTransformerBlock)
or isinstance(module, TemporalBasicTransformerBlock)
]
elif self.fusion_blocks == "full":
attn_modules = [
module
for module in torch_dfs(self.unet)
if isinstance(module, BasicTransformerBlock)
or isinstance(module, TemporalBasicTransformerBlock)
]
attn_modules = sorted(
attn_modules, key=lambda x: -x.norm1.normalized_shape[0]
)
for i, module in enumerate(attn_modules):
module._original_inner_forward = module.forward
if isinstance(module, BasicTransformerBlock):
module.forward = hacked_basic_transformer_inner_forward.__get__(
module, BasicTransformerBlock
)
if isinstance(module, TemporalBasicTransformerBlock):
module.forward = hacked_basic_transformer_inner_forward.__get__(
module, TemporalBasicTransformerBlock
)
module.bank = []
module.attn_weight = float(i) / float(len(attn_modules))
def update(self, writer, dtype=torch.float16):
if self.reference_attn:
if self.fusion_blocks == "midup":
reader_attn_modules = [
module
for module in (
torch_dfs(self.unet.mid_block) + torch_dfs(self.unet.up_blocks)
)
if isinstance(module, TemporalBasicTransformerBlock)
]
writer_attn_modules = [
module
for module in (
torch_dfs(writer.unet.mid_block)
+ torch_dfs(writer.unet.up_blocks)
)
if isinstance(module, BasicTransformerBlock)
]
elif self.fusion_blocks == "full":
reader_attn_modules = [
module
for module in torch_dfs(self.unet)
if isinstance(module, BasicTransformerBlock)
]
writer_attn_modules = [
module
for module in torch_dfs(writer.unet)
if isinstance(module, BasicTransformerBlock)
]
reader_attn_modules = sorted(
reader_attn_modules, key=lambda x: -x.norm1.normalized_shape[0]
)
writer_attn_modules = sorted(
writer_attn_modules, key=lambda x: -x.norm1.normalized_shape[0]
)
for r, w in zip(reader_attn_modules, writer_attn_modules):
r.bank = [v.clone().to(dtype) for v in w.bank]
# w.bank.clear()
def clear(self):
if self.reference_attn:
if self.fusion_blocks == "midup":
reader_attn_modules = [
module
for module in (
torch_dfs(self.unet.mid_block) + torch_dfs(self.unet.up_blocks)
)
if isinstance(module, BasicTransformerBlock)
or isinstance(module, TemporalBasicTransformerBlock)
]
elif self.fusion_blocks == "full":
reader_attn_modules = [
module
for module in torch_dfs(self.unet)
if isinstance(module, BasicTransformerBlock)
or isinstance(module, TemporalBasicTransformerBlock)
]
reader_attn_modules = sorted(
reader_attn_modules, key=lambda x: -x.norm1.normalized_shape[0]
)
for r in reader_attn_modules:
r.bank.clear()
================================================
FILE: src/models/projection.py
================================================
import torch
import torch.nn as nn
class My_proj(nn.Module):
def __init__(self, input=1024, dtype=torch.float32):
super(My_proj, self).__init__()
self.mapping_layer = nn.Linear(input, 1024)
self.dtype = dtype
def forward(self, x):
x = self.mapping_layer(x)
return x
================================================
FILE: src/models/transformer_2d.py
================================================
# Adapted from https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/transformer_2d.py
from dataclasses import dataclass
from typing import Any, Dict, Optional
import torch
from diffusers.configuration_utils import ConfigMixin, register_to_config
# from diffusers.models.embeddings import CaptionProjection
from diffusers.models.lora import LoRACompatibleConv, LoRACompatibleLinear
from diffusers.models.modeling_utils import ModelMixin
from diffusers.models.normalization import AdaLayerNormSingle
from diffusers.utils import USE_PEFT_BACKEND, BaseOutput, deprecate, is_torch_version
from torch import nn
from .attention import BasicTransformerBlock
@dataclass
class Transformer2DModelOutput(BaseOutput):
"""
The output of [`Transformer2DModel`].
Args:
sample (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)` or `(batch size, num_vector_embeds - 1, num_latent_pixels)` if [`Transformer2DModel`] is discrete):
The hidden states output conditioned on the `encoder_hidden_states` input. If discrete, returns probability
distributions for the unnoised latent pixels.
"""
sample: torch.FloatTensor
ref_feature: torch.FloatTensor
class Transformer2DModel(ModelMixin, ConfigMixin):
"""
A 2D Transformer model for image-like data.
Parameters:
num_attention_heads (`int`, *optional*, defaults to 16): The number of heads to use for multi-head attention.
attention_head_dim (`int`, *optional*, defaults to 88): The number of channels in each head.
in_channels (`int`, *optional*):
The number of channels in the input and output (specify if the input is **continuous**).
num_layers (`int`, *optional*, defaults to 1): The number of layers of Transformer blocks to use.
dropout (`float`, *optional*, defaults to 0.0): The dropout probability to use.
cross_attention_dim (`int`, *optional*): The number of `encoder_hidden_states` dimensions to use.
sample_size (`int`, *optional*): The width of the latent images (specify if the input is **discrete**).
This is fixed during training since it is used to learn a number of position embeddings.
num_vector_embeds (`int`, *optional*):
The number of classes of the vector embeddings of the latent pixels (specify if the input is **discrete**).
Includes the class for the masked latent pixel.
activation_fn (`str`, *optional*, defaults to `"geglu"`): Activation function to use in feed-forward.
num_embeds_ada_norm ( `int`, *optional*):
The number of diffusion steps used during training. Pass if at least one of the norm_layers is
`AdaLayerNorm`. This is fixed during training since it is used to learn a number of embeddings that are
added to the hidden states.
During inference, you can denoise for up to but not more steps than `num_embeds_ada_norm`.
attention_bias (`bool`, *optional*):
Configure if the `TransformerBlocks` attention should contain a bias parameter.
"""
_supports_gradient_checkpointing = True
@register_to_config
def __init__(
self,
num_attention_heads: int = 16,
attention_head_dim: int = 88,
in_channels: Optional[int] = None,
out_channels: Optional[int] = None,
num_layers: int = 1,
dropout: float = 0.0,
norm_num_groups: int = 32,
cross_attention_dim: Optional[int] = None,
attention_bias: bool = False,
sample_size: Optional[int] = None,
num_vector_embeds: Optional[int] = None,
patch_size: Optional[int] = None,
activation_fn: str = "geglu",
num_embeds_ada_norm: Optional[int] = None,
use_linear_projection: bool = False,
only_cross_attention: bool = False,
double_self_attention: bool = False,
upcast_attention: bool = False,
norm_type: str = "layer_norm",
norm_elementwise_affine: bool = True,
norm_eps: float = 1e-5,
attention_type: str = "default",
caption_channels: int = None,
):
super().__init__()
self.use_linear_projection = use_linear_projection
self.num_attention_heads = num_attention_heads
self.attention_head_dim = attention_head_dim
inner_dim = num_attention_heads * attention_head_dim
conv_cls = nn.Conv2d if USE_PEFT_BACKEND else LoRACompatibleConv
linear_cls = nn.Linear if USE_PEFT_BACKEND else LoRACompatibleLinear
# 1. Transformer2DModel can process both standard continuous images of shape `(batch_size, num_channels, width, height)` as well as quantized image embeddings of shape `(batch_size, num_image_vectors)`
# Define whether input is continuous or discrete depending on configuration
self.is_input_continuous = (in_channels is not None) and (patch_size is None)
self.is_input_vectorized = num_vector_embeds is not None
self.is_input_patches = in_channels is not None and patch_size is not None
if norm_type == "layer_norm" and num_embeds_ada_norm is not None:
deprecation_message = (
f"The configuration file of this model: {self.__class__} is outdated. `norm_type` is either not set or"
" incorrectly set to `'layer_norm'`.Make sure to set `norm_type` to `'ada_norm'` in the config."
" Please make sure to update the config accordingly as leaving `norm_type` might led to incorrect"
" results in future versions. If you have downloaded this checkpoint from the Hugging Face Hub, it"
" would be very nice if you could open a Pull request for the `transformer/config.json` file"
)
deprecate(
"norm_type!=num_embeds_ada_norm",
"1.0.0",
deprecation_message,
standard_warn=False,
)
norm_type = "ada_norm"
if self.is_input_continuous and self.is_input_vectorized:
raise ValueError(
f"Cannot define both `in_channels`: {in_channels} and `num_vector_embeds`: {num_vector_embeds}. Make"
" sure that either `in_channels` or `num_vector_embeds` is None."
)
elif self.is_input_vectorized and self.is_input_patches:
raise ValueError(
f"Cannot define both `num_vector_embeds`: {num_vector_embeds} and `patch_size`: {patch_size}. Make"
" sure that either `num_vector_embeds` or `num_patches` is None."
)
elif (
not self.is_input_continuous
and not self.is_input_vectorized
and not self.is_input_patches
):
raise ValueError(
f"Has to define `in_channels`: {in_channels}, `num_vector_embeds`: {num_vector_embeds}, or patch_size:"
f" {patch_size}. Make sure that `in_channels`, `num_vector_embeds` or `num_patches` is not None."
)
# 2. Define input layers
self.in_channels = in_channels
self.norm = torch.nn.GroupNorm(
num_groups=norm_num_groups,
num_channels=in_channels,
eps=1e-6,
affine=True,
)
if use_linear_projection:
self.proj_in = linear_cls(in_channels, inner_dim)
else:
self.proj_in = conv_cls(
in_channels, inner_dim, kernel_size=1, stride=1, padding=0
)
# 3. Define transformers blocks
self.transformer_blocks = nn.ModuleList(
[
BasicTransformerBlock(
inner_dim,
num_attention_heads,
attention_head_dim,
dropout=dropout,
cross_attention_dim=cross_attention_dim,
activation_fn=activation_fn,
num_embeds_ada_norm=num_embeds_ada_norm,
attention_bias=attention_bias,
only_cross_attention=only_cross_attention,
double_self_attention=double_self_attention,
upcast_attention=upcast_attention,
norm_type=norm_type,
norm_elementwise_affine=norm_elementwise_affine,
norm_eps=norm_eps,
attention_type=attention_type,
)
for d in range(num_layers)
]
)
# 4. Define output layers
self.out_channels = in_channels if out_channels is None else out_channels
# TODO: should use out_channels for continuous projections
if use_linear_projection:
self.proj_out = linear_cls(inner_dim, in_channels)
else:
self.proj_out = conv_cls(
inner_dim, in_channels, kernel_size=1, stride=1, padding=0
)
# 5. PixArt-Alpha blocks.
self.adaln_single = None
self.use_additional_conditions = False
if norm_type == "ada_norm_single":
self.use_additional_conditions = self.config.sample_size == 128
# TODO(Sayak, PVP) clean this, for now we use sample size to determine whether to use
# additional conditions until we find better name
self.adaln_single = AdaLayerNormSingle(
inner_dim, use_additional_conditions=self.use_additional_conditions
)
self.caption_projection = None
# if caption_channels is not None:
# self.caption_projection = CaptionProjection(
# in_features=caption_channels, hidden_size=inner_dim
# )
self.gradient_checkpointing = False
def _set_gradient_checkpointing(self, module, value=False):
if hasattr(module, "gradient_checkpointing"):
module.gradient_checkpointing = value
def forward(
self,
hidden_states: torch.Tensor,
encoder_hidden_states: Optional[torch.Tensor] = None,
timestep: Optional[torch.LongTensor] = None,
added_cond_kwargs: Dict[str, torch.Tensor] = None,
class_labels: Optional[torch.LongTensor] = None,
cross_attention_kwargs: Dict[str, Any] = None,
attention_mask: Optional[torch.Tensor] = None,
encoder_attention_mask: Optional[torch.Tensor] = None,
return_dict: bool = True,
):
"""
The [`Transformer2DModel`] forward method.
Args:
hidden_states (`torch.LongTensor` of shape `(batch size, num latent pixels)` if discrete, `torch.FloatTensor` of shape `(batch size, channel, height, width)` if continuous):
Input `hidden_states`.
encoder_hidden_states ( `torch.FloatTensor` of shape `(batch size, sequence len, embed dims)`, *optional*):
Conditional embeddings for cross attention layer. If not given, cross-attention defaults to
self-attention.
timestep ( `torch.LongTensor`, *optional*):
Used to indicate denoising step. Optional timestep to be applied as an embedding in `AdaLayerNorm`.
class_labels ( `torch.LongTensor` of shape `(batch size, num classes)`, *optional*):
Used to indicate class labels conditioning. Optional class labels to be applied as an embedding in
`AdaLayerZeroNorm`.
cross_attention_kwargs ( `Dict[str, Any]`, *optional*):
A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
`self.processor` in
[diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
attention_mask ( `torch.Tensor`, *optional*):
An attention mask of shape `(batch, key_tokens)` is applied to `encoder_hidden_states`. If `1` the mask
is kept, otherwise if `0` it is discarded. Mask will be converted into a bias, which adds large
negative values to the attention scores corresponding to "discard" tokens.
encoder_attention_mask ( `torch.Tensor`, *optional*):
Cross-attention mask applied to `encoder_hidden_states`. Two formats supported:
* Mask `(batch, sequence_length)` True = keep, False = discard.
* Bias `(batch, 1, sequence_length)` 0 = keep, -10000 = discard.
If `ndim == 2`: will be interpreted as a mask, then converted into a bias consistent with the format
above. This bias will be added to the cross-attention scores.
return_dict (`bool`, *optional*, defaults to `True`):
Whether or not to return a [`~models.unet_2d_condition.UNet2DConditionOutput`] instead of a plain
tuple.
Returns:
If `return_dict` is True, an [`~models.transformer_2d.Transformer2DModelOutput`] is returned, otherwise a
`tuple` where the first element is the sample tensor.
"""
# ensure attention_mask is a bias, and give it a singleton query_tokens dimension.
# we may have done this conversion already, e.g. if we came here via UNet2DConditionModel#forward.
# we can tell by counting dims; if ndim == 2: it's a mask rather than a bias.
# expects mask of shape:
# [batch, key_tokens]
# adds singleton query_tokens dimension:
# [batch, 1, key_tokens]
# this helps to broadcast it as a bias over attention scores, which will be in one of the following shapes:
# [batch, heads, query_tokens, key_tokens] (e.g. torch sdp attn)
# [batch * heads, query_tokens, key_tokens] (e.g. xformers or classic attn)
if attention_mask is not None and attention_mask.ndim == 2:
# assume that mask is expressed as:
# (1 = keep, 0 = discard)
# convert mask into a bias that can be added to attention scores:
# (keep = +0, discard = -10000.0)
attention_mask = (1 - attention_mask.to(hidden_states.dtype)) * -10000.0
attention_mask = attention_mask.unsqueeze(1)
# convert encoder_attention_mask to a bias the same way we do for attention_mask
if encoder_attention_mask is not None and encoder_attention_mask.ndim == 2:
encoder_attention_mask = (
1 - encoder_attention_mask.to(hidden_states.dtype)
) * -10000.0
encoder_attention_mask = encoder_attention_mask.unsqueeze(1)
# Retrieve lora scale.
lora_scale = (
cross_attention_kwargs.get("scale", 1.0)
if cross_attention_kwargs is not None
else 1.0
)
# 1. Input
batch, _, height, width = hidden_states.shape
residual = hidden_states
hidden_states = self.norm(hidden_states)
if not self.use_linear_projection:
hidden_states = (
self.proj_in(hidden_states, scale=lora_scale)
if not USE_PEFT_BACKEND
else self.proj_in(hidden_states)
)
inner_dim = hidden_states.shape[1]
hidden_states = hidden_states.permute(0, 2, 3, 1).reshape(
batch, height * width, inner_dim
)
else:
inner_dim = hidden_states.shape[1]
hidden_states = hidden_states.permute(0, 2, 3, 1).reshape(
batch, height * width, inner_dim
)
hidden_states = (
self.proj_in(hidden_states, scale=lora_scale)
if not USE_PEFT_BACKEND
else self.proj_in(hidden_states)
)
# 2. Blocks
if self.caption_projection is not None:
batch_size = hidden_states.shape[0]
encoder_hidden_states = self.caption_projection(encoder_hidden_states)
encoder_hidden_states = encoder_hidden_states.view(
batch_size, -1, hidden_states.shape[-1]
)
ref_feature = hidden_states.reshape(batch, height, width, inner_dim)
for block in self.transformer_blocks:
if self.training and self.gradient_checkpointing:
def create_custom_forward(module, return_dict=None):
def custom_forward(*inputs):
if return_dict is not None:
return module(*inputs, return_dict=return_dict)
else:
return module(*inputs)
return custom_forward
ckpt_kwargs: Dict[str, Any] = (
{"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
)
hidden_states = torch.utils.checkpoint.checkpoint(
create_custom_forward(block),
hidden_states,
attention_mask,
encoder_hidden_states,
encoder_attention_mask,
timestep,
cross_attention_kwargs,
class_labels,
**ckpt_kwargs,
)
else:
hidden_states = block(
hidden_states,
attention_mask=attention_mask,
encoder_hidden_states=encoder_hidden_states,
encoder_attention_mask=encoder_attention_mask,
timestep=timestep,
cross_attention_kwargs=cross_attention_kwargs,
class_labels=class_labels,
)
# 3. Output
if self.is_input_continuous:
if not self.use_linear_projection:
hidden_states = (
hidden_states.reshape(batch, height, width, inner_dim)
.permute(0, 3, 1, 2)
.contiguous()
)
hidden_states = (
self.proj_out(hidden_states, scale=lora_scale)
if not USE_PEFT_BACKEND
else self.proj_out(hidden_states)
)
else:
hidden_states = (
self.proj_out(hidden_states, scale=lora_scale)
if not USE_PEFT_BACKEND
else self.proj_out(hidden_states)
)
hidden_states = (
hidden_states.reshape(batch, height, width, inner_dim)
.permute(0, 3, 1, 2)
.contiguous()
)
output = hidden_states + residual
if not return_dict:
return (output, ref_feature)
return Transformer2DModelOutput(sample=output, ref_feature=ref_feature)
================================================
FILE: src/models/unet_2d_blocks.py
================================================
# Adapted from https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/unet_2d_blocks.py
from typing import Any, Dict, Optional, Tuple, Union
import numpy as np
import torch
import torch.nn.functional as F
from diffusers.models.activations import get_activation
from diffusers.models.attention_processor import Attention
from diffusers.models.dual_transformer_2d import DualTransformer2DModel
from diffusers.models.resnet import Downsample2D, ResnetBlock2D, Upsample2D
from diffusers.utils import is_torch_version, logging
from diffusers.utils.torch_utils import apply_freeu
from torch import nn
from .transformer_2d import Transformer2DModel
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
def get_down_block(
down_block_type: str,
num_layers: int,
in_channels: int,
out_channels: int,
temb_channels: int,
add_downsample: bool,
resnet_eps: float,
resnet_act_fn: str,
transformer_layers_per_block: int = 1,
num_attention_heads: Optional[int] = None,
resnet_groups: Optional[int] = None,
cross_attention_dim: Optional[int] = None,
downsample_padding: Optional[int] = None,
dual_cross_attention: bool = False,
use_linear_projection: bool = False,
only_cross_attention: bool = False,
upcast_attention: bool = False,
resnet_time_scale_shift: str = "default",
attention_type: str = "default",
resnet_skip_time_act: bool = False,
resnet_out_scale_factor: float = 1.0,
cross_attention_norm: Optional[str] = None,
attention_head_dim: Optional[int] = None,
downsample_type: Optional[str] = None,
dropout: float = 0.0,
):
# If attn head dim is not defined, we default it to the number of heads
if attention_head_dim is None:
logger.warn(
f"It is recommended to provide `attention_head_dim` when calling `get_down_block`. Defaulting `attention_head_dim` to {num_attention_heads}."
)
attention_head_dim = num_attention_heads
down_block_type = (
down_block_type[7:]
if down_block_type.startswith("UNetRes")
else down_block_type
)
if down_block_type == "DownBlock2D":
return DownBlock2D(
num_layers=num_layers,
in_channels=in_channels,
out_channels=out_channels,
temb_channels=temb_channels,
dropout=dropout,
add_downsample=add_downsample,
resnet_eps=resnet_eps,
resnet_act_fn=resnet_act_fn,
resnet_groups=resnet_groups,
downsample_padding=downsample_padding,
resnet_time_scale_shift=resnet_time_scale_shift,
)
elif down_block_type == "CrossAttnDownBlock2D":
if cross_attention_dim is None:
raise ValueError(
"cross_attention_dim must be specified for CrossAttnDownBlock2D"
)
return CrossAttnDownBlock2D(
num_layers=num_layers,
transformer_layers_per_block=transformer_layers_per_block,
in_channels=in_channels,
out_channels=out_channels,
temb_channels=temb_channels,
dropout=dropout,
add_downsample=add_downsample,
resnet_eps=resnet_eps,
resnet_act_fn=resnet_act_fn,
resnet_groups=resnet_groups,
downsample_padding=downsample_padding,
cross_attention_dim=cross_attention_dim,
num_attention_heads=num_attention_heads,
dual_cross_attention=dual_cross_attention,
use_linear_projection=use_linear_projection,
only_cross_attention=only_cross_attention,
upcast_attention=upcast_attention,
resnet_time_scale_shift=resnet_time_scale_shift,
attention_type=attention_type,
)
raise ValueError(f"{down_block_type} does not exist.")
def get_up_block(
up_block_type: str,
num_layers: int,
in_channels: int,
out_channels: int,
prev_output_channel: int,
temb_channels: int,
add_upsample: bool,
resnet_eps: float,
resnet_act_fn: str,
resolution_idx: Optional[int] = None,
transformer_layers_per_block: int = 1,
num_attention_heads: Optional[int] = None,
resnet_groups: Optional[int] = None,
cross_attention_dim: Optional[int] = None,
dual_cross_attention: bool = False,
use_linear_projection: bool = False,
only_cross_attention: bool = False,
upcast_attention: bool = False,
resnet_time_scale_shift: str = "default",
attention_type: str = "default",
resnet_skip_time_act: bool = False,
resnet_out_scale_factor: float = 1.0,
cross_attention_norm: Optional[str] = None,
attention_head_dim: Optional[int] = None,
upsample_type: Optional[str] = None,
dropout: float = 0.0,
) -> nn.Module:
# If attn head dim is not defined, we default it to the number of heads
if attention_head_dim is None:
logger.warn(
f"It is recommended to provide `attention_head_dim` when calling `get_up_block`. Defaulting `attention_head_dim` to {num_attention_heads}."
)
attention_head_dim = num_attention_heads
up_block_type = (
up_block_type[7:] if up_block_type.startswith("UNetRes") else up_block_type
)
if up_block_type == "UpBlock2D":
return UpBlock2D(
num_layers=num_layers,
in_channels=in_channels,
out_channels=out_channels,
prev_output_channel=prev_output_channel,
temb_channels=temb_channels,
resolution_idx=resolution_idx,
dropout=dropout,
add_upsample=add_upsample,
resnet_eps=resnet_eps,
resnet_act_fn=resnet_act_fn,
resnet_groups=resnet_groups,
resnet_time_scale_shift=resnet_time_scale_shift,
)
elif up_block_type == "CrossAttnUpBlock2D":
if cross_attention_dim is None:
raise ValueError(
"cross_attention_dim must be specified for CrossAttnUpBlock2D"
)
return CrossAttnUpBlock2D(
num_layers=num_layers,
transformer_layers_per_block=transformer_layers_per_block,
in_channels=in_channels,
out_channels=out_channels,
prev_output_channel=prev_output_channel,
temb_channels=temb_channels,
resolution_idx=resolution_idx,
dropout=dropout,
add_upsample=add_upsample,
resnet_eps=resnet_eps,
resnet_act_fn=resnet_act_fn,
resnet_groups=resnet_groups,
cross_attention_dim=cross_attention_dim,
num_attention_heads=num_attention_heads,
dual_cross_attention=dual_cross_attention,
use_linear_projection=use_linear_projection,
only_cross_attention=only_cross_attention,
upcast_attention=upcast_attention,
resnet_time_scale_shift=resnet_time_scale_shift,
attention_type=attention_type,
)
raise ValueError(f"{up_block_type} does not exist.")
def get_mid_block(
mid_block_type: str,
temb_channels: int,
in_channels: int,
resnet_eps: float,
resnet_act_fn: str,
resnet_groups: int,
output_scale_factor: float = 1.0,
transformer_layers_per_block: int = 1,
num_attention_heads: Optional[int] = None,
cross_attention_dim: Optional[int] = None,
dual_cross_attention: bool = False,
use_linear_projection: bool = False,
mid_block_only_cross_attention: bool = False,
upcast_attention: bool = False,
resnet_time_scale_shift: str = "default",
attention_type: str = "default",
resnet_skip_time_act: bool = False,
cross_attention_norm: Optional[str] = None,
attention_head_dim: Optional[int] = 1,
dropout: float = 0.0,
):
if mid_block_type == "UNetMidBlock2DCrossAttn":
return UNetMidBlock2DCrossAttn(
transformer_layers_per_block=transformer_layers_per_block,
in_channels=in_channels,
temb_channels=temb_channels,
dropout=dropout,
resnet_eps=resnet_eps,
resnet_act_fn=resnet_act_fn,
output_scale_factor=output_scale_factor,
resnet_time_scale_shift=resnet_time_scale_shift,
cross_attention_dim=cross_attention_dim,
num_attention_heads=num_attention_heads,
resnet_groups=resnet_groups,
dual_cross_attention=dual_cross_attention,
use_linear_projection=use_linear_projection,
upcast_attention=upcast_attention,
attention_type=attention_type,
)
elif mid_block_type == "UNetMidBlock2D":
return UNetMidBlock2D(
in_channels=in_channels,
temb_channels=temb_channels,
dropout=dropout,
num_layers=0,
resnet_eps=resnet_eps,
resnet_act_fn=resnet_act_fn,
output_scale_factor=output_scale_factor,
resnet_groups=resnet_groups,
resnet_time_scale_shift=resnet_time_scale_shift,
add_attention=False,
)
elif mid_block_type is None:
return None
else:
raise ValueError(f"unknown mid_block_type : {mid_block_type}")
class AutoencoderTinyBlock(nn.Module):
"""
Tiny Autoencoder block used in [`AutoencoderTiny`]. It is a mini residual module consisting of plain conv + ReLU
blocks.
Args:
in_channels (`int`): The number of input channels.
out_channels (`int`): The number of output channels.
act_fn (`str`):
` The activation function to use. Supported values are `"swish"`, `"mish"`, `"gelu"`, and `"relu"`.
Returns:
`torch.FloatTensor`: A tensor with the same shape as the input tensor, but with the number of channels equal to
`out_channels`.
"""
def __init__(self, in_channels: int, out_channels: int, act_fn: str):
super().__init__()
act_fn = get_activation(act_fn)
self.conv = nn.Sequential(
nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1),
act_fn,
nn.Conv2d(out_channels, out_channels, kernel_size=3, padding=1),
act_fn,
nn.Conv2d(out_channels, out_channels, kernel_size=3, padding=1),
)
self.skip = (
nn.Conv2d(in_channels, out_channels, kernel_size=1, bias=False)
if in_channels != out_channels
else nn.Identity()
)
self.fuse = nn.ReLU()
def forward(self, x: torch.FloatTensor) -> torch.FloatTensor:
return self.fuse(self.conv(x) + self.skip(x))
class UNetMidBlock2D(nn.Module):
"""
A 2D UNet mid-block [`UNetMidBlock2D`] with multiple residual blocks and optional attention blocks.
Args:
in_channels (`int`): The number of input channels.
temb_channels (`int`): The number of temporal embedding channels.
dropout (`float`, *optional*, defaults to 0.0): The dropout rate.
num_layers (`int`, *optional*, defaults to 1): The number of residual blocks.
resnet_eps (`float`, *optional*, 1e-6 ): The epsilon value for the resnet blocks.
resnet_time_scale_shift (`str`, *optional*, defaults to `default`):
The type of normalization to apply to the time embeddings. This can help to improve the performance of the
model on tasks with long-range temporal dependencies.
resnet_act_fn (`str`, *optional*, defaults to `swish`): The activation function for the resnet blocks.
resnet_groups (`int`, *optional*, defaults to 32):
The number of groups to use in the group normalization layers of the resnet blocks.
attn_groups (`Optional[int]`, *optional*, defaults to None): The number of groups for the attention blocks.
resnet_pre_norm (`bool`, *optional*, defaults to `True`):
Whether to use pre-normalization for the resnet blocks.
add_attention (`bool`, *optional*, defaults to `True`): Whether to add attention blocks.
attention_head_dim (`int`, *optional*, defaults to 1):
Dimension of a single attention head. The number of attention heads is determined based on this value and
the number of input channels.
output_scale_factor (`float`, *optional*, defaults to 1.0): The output scale factor.
Returns:
`torch.FloatTensor`: The output of the last residual block, which is a tensor of shape `(batch_size,
in_channels, height, width)`.
"""
def __init__(
self,
in_channels: int,
temb_channels: int,
dropout: float = 0.0,
num_layers: int = 1,
resnet_eps: float = 1e-6,
resnet_time_scale_shift: str = "default", # default, spatial
resnet_act_fn: str = "swish",
resnet_groups: int = 32,
attn_groups: Optional[int] = None,
resnet_pre_norm: bool = True,
add_attention: bool = True,
attention_head_dim: int = 1,
output_scale_factor: float = 1.0,
):
super().__init__()
resnet_groups = (
resnet_groups if resnet_groups is not None else min(in_channels // 4, 32)
)
self.add_attention = add_attention
if attn_groups is None:
attn_groups = (
resnet_groups if resnet_time_scale_shift == "default" else None
)
# there is always at least one resnet
resnets = [
ResnetBlock2D(
in_channels=in_channels,
out_channels=in_channels,
temb_channels=temb_channels,
eps=resnet_eps,
groups=resnet_groups,
dropout=dropout,
time_embedding_norm=resnet_time_scale_shift,
non_linearity=resnet_act_fn,
output_scale_factor=output_scale_factor,
pre_norm=resnet_pre_norm,
)
]
attentions = []
if attention_head_dim is None:
logger.warn(
f"It is not recommend to pass `attention_head_dim=None`. Defaulting `attention_head_dim` to `in_channels`: {in_channels}."
)
attention_head_dim = in_channels
for _ in range(num_layers):
if self.add_attention:
attentions.append(
Attention(
in_channels,
heads=in_channels // attention_head_dim,
dim_head=attention_head_dim,
rescale_output_factor=output_scale_factor,
eps=resnet_eps,
norm_num_groups=attn_groups,
spatial_norm_dim=temb_channels
if resnet_time_scale_shift == "spatial"
else None,
residual_connection=True,
bias=True,
upcast_softmax=True,
_from_deprecated_attn_block=True,
)
)
else:
attentions.append(None)
resnets.append(
ResnetBlock2D(
in_channels=in_channels,
out_channels=in_channels,
temb_channels=temb_channels,
eps=resnet_eps,
groups=resnet_groups,
dropout=dropout,
time_embedding_norm=resnet_time_scale_shift,
non_linearity=resnet_act_fn,
output_scale_factor=output_scale_factor,
pre_norm=resnet_pre_norm,
)
)
self.attentions = nn.ModuleList(attentions)
self.resnets = nn.ModuleList(resnets)
def forward(
self, hidden_states: torch.FloatTensor, temb: Optional[torch.FloatTensor] = None
) -> torch.FloatTensor:
hidden_states = self.resnets[0](hidden_states, temb)
for attn, resnet in zip(self.attentions, self.resnets[1:]):
if attn is not None:
hidden_states = attn(hidden_states, temb=temb)
hidden_states = resnet(hidden_states, temb)
return hidden_states
class UNetMidBlock2DCrossAttn(nn.Module):
def __init__(
self,
in_channels: int,
temb_channels: int,
dropout: float = 0.0,
num_layers: int = 1,
transformer_layers_per_block: Union[int, Tuple[int]] = 1,
resnet_eps: float = 1e-6,
resnet_time_scale_shift: str = "default",
resnet_act_fn: str = "swish",
resnet_groups: int = 32,
resnet_pre_norm: bool = True,
num_attention_heads: int = 1,
output_scale_factor: float = 1.0,
cross_attention_dim: int = 1280,
dual_cross_attention: bool = False,
use_linear_projection: bool = False,
upcast_attention: bool = False,
attention_type: str = "default",
):
super().__init__()
self.has_cross_attention = True
self.num_attention_heads = num_attention_heads
resnet_groups = (
resnet_groups if resnet_groups is not None else min(in_channels // 4, 32)
)
# support for variable transformer layers per block
if isinstance(transformer_layers_per_block, int):
transformer_layers_per_block = [transformer_layers_per_block] * num_layers
# there is always at least one resnet
resnets = [
ResnetBlock2D(
in_channels=in_channels,
out_channels=in_channels,
temb_channels=temb_channels,
eps=resnet_eps,
groups=resnet_groups,
dropout=dropout,
time_embedding_norm=resnet_time_scale_shift,
non_linearity=resnet_act_fn,
output_scale_factor=output_scale_factor,
pre_norm=resnet_pre_norm,
)
]
attentions = []
for i in range(num_layers):
if not dual_cross_attention:
attentions.append(
Transformer2DModel(
num_attention_heads,
in_channels // num_attention_heads,
in_channels=in_channels,
num_layers=transformer_layers_per_block[i],
cross_attention_dim=cross_attention_dim,
norm_num_groups=resnet_groups,
use_linear_projection=use_linear_projection,
upcast_attention=upcast_attention,
attention_type=attention_type,
)
)
else:
attentions.append(
DualTransformer2DModel(
num_attention_heads,
in_channels // num_attention_heads,
in_channels=in_channels,
num_layers=1,
cross_attention_dim=cross_attention_dim,
norm_num_groups=resnet_groups,
)
)
resnets.append(
ResnetBlock2D(
in_channels=in_channels,
out_channels=in_channels,
temb_channels=temb_channels,
eps=resnet_eps,
groups=resnet_groups,
dropout=dropout,
time_embedding_norm=resnet_time_scale_shift,
non_linearity=resnet_act_fn,
output_scale_factor=output_scale_factor,
pre_norm=resnet_pre_norm,
)
)
self.attentions = nn.ModuleList(attentions)
self.resnets = nn.ModuleList(resnets)
self.gradient_checkpointing = False
def forward(
self,
hidden_states: torch.FloatTensor,
temb: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
attention_mask: Optional[torch.FloatTensor] = None,
cross_attention_kwargs: Optional[Dict[str, Any]] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
) -> torch.FloatTensor:
lora_scale = (
cross_attention_kwargs.get("scale", 1.0)
if cross_attention_kwargs is not None
else 1.0
)
hidden_states = self.resnets[0](hidden_states, temb, scale=lora_scale)
for attn, resnet in zip(self.attentions, self.resnets[1:]):
if self.training and self.gradient_checkpointing:
def create_custom_forward(module, return_dict=None):
def custom_forward(*inputs):
if return_dict is not None:
return module(*inputs, return_dict=return_dict)
else:
return module(*inputs)
return custom_forward
ckpt_kwargs: Dict[str, Any] = (
{"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
)
hidden_states, ref_feature = attn(
hidden_states,
encoder_hidden_states=encoder_hidden_states,
cross_attention_kwargs=cross_attention_kwargs,
attention_mask=attention_mask,
encoder_attention_mask=encoder_attention_mask,
return_dict=False,
)
hidden_states = torch.utils.checkpoint.checkpoint(
create_custom_forward(resnet),
hidden_states,
temb,
**ckpt_kwargs,
)
else:
hidden_states, ref_feature = attn(
hidden_states,
encoder_hidden_states=encoder_hidden_states,
cross_attention_kwargs=cross_attention_kwargs,
attention_mask=attention_mask,
encoder_attention_mask=encoder_attention_mask,
return_dict=False,
)
hidden_states = resnet(hidden_states, temb, scale=lora_scale)
return hidden_states
class CrossAttnDownBlock2D(nn.Module):
def __init__(
self,
in_channels: int,
out_channels: int,
temb_channels: int,
dropout: float = 0.0,
num_layers: int = 1,
transformer_layers_per_block: Union[int, Tuple[int]] = 1,
resnet_eps: float = 1e-6,
resnet_time_scale_shift: str = "default",
resnet_act_fn: str = "swish",
resnet_groups: int = 32,
resnet_pre_norm: bool = True,
num_attention_heads: int = 1,
cross_attention_dim: int = 1280,
output_scale_factor: float = 1.0,
downsample_padding: int = 1,
add_downsample: bool = True,
dual_cross_attention: bool = False,
use_linear_projection: bool = False,
only_cross_attention: bool = False,
upcast_attention: bool = False,
attention_type: str = "default",
):
super().__init__()
resnets = []
attentions = []
self.has_cross_attention = True
self.num_attention_heads = num_attention_heads
if isinstance(transformer_layers_per_block, int):
transformer_layers_per_block = [transformer_layers_per_block] * num_layers
for i in range(num_layers):
in_channels = in_channels if i == 0 else out_channels
resnets.append(
ResnetBlock2D(
in_channels=in_channels,
out_channels=out_channels,
temb_channels=temb_channels,
eps=resnet_eps,
groups=resnet_groups,
dropout=dropout,
time_embedding_norm=resnet_time_scale_shift,
non_linearity=resnet_act_fn,
output_scale_factor=output_scale_factor,
pre_norm=resnet_pre_norm,
)
)
if not dual_cross_attention:
attentions.append(
Transformer2DModel(
num_attention_heads,
out_channels // num_attention_heads,
in_channels=out_channels,
num_layers=transformer_layers_per_block[i],
cross_attention_dim=cross_attention_dim,
norm_num_groups=resnet_groups,
use_linear_projection=use_linear_projection,
only_cross_attention=only_cross_attention,
upcast_attention=upcast_attention,
attention_type=attention_type,
)
)
else:
attentions.append(
DualTransformer2DModel(
num_attention_heads,
out_channels // num_attention_heads,
in_channels=out_channels,
num_layers=1,
cross_attention_dim=cross_attention_dim,
norm_num_groups=resnet_groups,
)
)
self.attentions = nn.ModuleList(attentions)
self.resnets = nn.ModuleList(resnets)
if add_downsample:
self.downsamplers = nn.ModuleList(
[
Downsample2D(
out_channels,
use_conv=True,
out_channels=out_channels,
padding=downsample_padding,
name="op",
)
]
)
else:
self.downsamplers = None
self.gradient_checkpointing = False
def forward(
self,
hidden_states: torch.FloatTensor,
temb: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
attention_mask: Optional[torch.FloatTensor] = None,
cross_attention_kwargs: Optional[Dict[str, Any]] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
additional_residuals: Optional[torch.FloatTensor] = None,
) -> Tuple[torch.FloatTensor, Tuple[torch.FloatTensor, ...]]:
output_states = ()
lora_scale = (
cross_attention_kwargs.get("scale", 1.0)
if cross_attention_kwargs is not None
else 1.0
)
blocks = list(zip(self.resnets, self.attentions))
for i, (resnet, attn) in enumerate(blocks):
if self.training and self.gradient_checkpointing:
def create_custom_forward(module, return_dict=None):
def custom_forward(*inputs):
if return_dict is not None:
return module(*inputs, return_dict=return_dict)
else:
return module(*inputs)
return custom_forward
ckpt_kwargs: Dict[str, Any] = (
{"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
)
hidden_states = torch.utils.checkpoint.checkpoint(
create_custom_forward(resnet),
hidden_states,
temb,
**ckpt_kwargs,
)
hidden_states, ref_feature = attn(
hidden_states,
encoder_hidden_states=encoder_hidden_states,
cross_attention_kwargs=cross_attention_kwargs,
attention_mask=attention_mask,
encoder_attention_mask=encoder_attention_mask,
return_dict=False,
)
else:
hidden_states = resnet(hidden_states, temb, scale=lora_scale)
hidden_states, ref_feature = attn(
hidden_states,
encoder_hidden_states=encoder_hidden_states,
cross_attention_kwargs=cross_attention_kwargs,
attention_mask=attention_mask,
encoder_attention_mask=encoder_attention_mask,
return_dict=False,
)
# apply additional residuals to the output of the last pair of resnet and attention blocks
if i == len(blocks) - 1 and additional_residuals is not None:
hidden_states = hidden_states + additional_residuals
output_states = output_states + (hidden_states,)
if self.downsamplers is not None:
for downsampler in self.downsamplers:
hidden_states = downsampler(hidden_states, scale=lora_scale)
output_states = output_states + (hidden_states,)
return hidden_states, output_states
class DownBlock2D(nn.Module):
def __init__(
self,
in_channels: int,
out_channels: int,
temb_channels: int,
dropout: float = 0.0,
num_layers: int = 1,
resnet_eps: float = 1e-6,
resnet_time_scale_shift: str = "default",
resnet_act_fn: str = "swish",
resnet_groups: int = 32,
resnet_pre_norm: bool = True,
output_scale_factor: float = 1.0,
add_downsample: bool = True,
downsample_padding: int = 1,
):
super().__init__()
resnets = []
for i in range(num_layers):
in_channels = in_channels if i == 0 else out_channels
resnets.append(
ResnetBlock2D(
in_channels=in_channels,
out_channels=out_channels,
temb_channels=temb_channels,
eps=resnet_eps,
groups=resnet_groups,
dropout=dropout,
time_embedding_norm=resnet_time_scale_shift,
non_linearity=resnet_act_fn,
output_scale_factor=output_scale_factor,
pre_norm=resnet_pre_norm,
)
)
self.resnets = nn.ModuleList(resnets)
if add_downsample:
self.downsamplers = nn.ModuleList(
[
Downsample2D(
out_channels,
use_conv=True,
out_channels=out_channels,
padding=downsample_padding,
name="op",
)
]
)
else:
self.downsamplers = None
self.gradient_checkpointing = False
def forward(
self,
hidden_states: torch.FloatTensor,
temb: Optional[torch.FloatTensor] = None,
scale: float = 1.0,
) -> Tuple[torch.FloatTensor, Tuple[torch.FloatTensor, ...]]:
output_states = ()
for resnet in self.resnets:
if self.training and self.gradient_checkpointing:
def create_custom_forward(module):
def custom_forward(*inputs):
return module(*inputs)
return custom_forward
if is_torch_version(">=", "1.11.0"):
hidden_states = torch.utils.checkpoint.checkpoint(
create_custom_forward(resnet),
hidden_states,
temb,
use_reentrant=False,
)
else:
hidden_states = torch.utils.checkpoint.checkpoint(
create_custom_forward(resnet), hidden_states, temb
)
else:
hidden_states = resnet(hidden_states, temb, scale=scale)
output_states = output_states + (hidden_states,)
if self.downsamplers is not None:
for downsampler in self.downsamplers:
hidden_states = downsampler(hidden_states, scale=scale)
output_states = output_states + (hidden_states,)
return hidden_states, output_states
class CrossAttnUpBlock2D(nn.Module):
def __init__(
self,
in_channels: int,
out_channels: int,
prev_output_channel: int,
temb_channels: int,
resolution_idx: Optional[int] = None,
dropout: float = 0.0,
num_layers: int = 1,
transformer_layers_per_block: Union[int, Tuple[int]] = 1,
resnet_eps: float = 1e-6,
resnet_time_scale_shift: str = "default",
resnet_act_fn: str = "swish",
resnet_groups: int = 32,
resnet_pre_norm: bool = True,
num_attention_heads: int = 1,
cross_attention_dim: int = 1280,
output_scale_factor: float = 1.0,
add_upsample: bool = True,
dual_cross_attention: bool = False,
use_linear_projection: bool = False,
only_cross_attention: bool = False,
upcast_attention: bool = False,
attention_type: str = "default",
):
super().__init__()
resnets = []
attentions = []
self.has_cross_attention = True
self.num_attention_heads = num_attention_heads
if isinstance(transformer_layers_per_block, int):
transformer_layers_per_block = [transformer_layers_per_block] * num_layers
for i in range(num_layers):
res_skip_channels = in_channels if (i == num_layers - 1) else out_channels
resnet_in_channels = prev_output_channel if i == 0 else out_channels
resnets.append(
ResnetBlock2D(
in_channels=resnet_in_channels + res_skip_channels,
out_channels=out_channels,
temb_channels=temb_channels,
eps=resnet_eps,
groups=resnet_groups,
dropout=dropout,
time_embedding_norm=resnet_time_scale_shift,
non_linearity=resnet_act_fn,
output_scale_factor=output_scale_factor,
pre_norm=resnet_pre_norm,
)
)
if not dual_cross_attention:
attentions.append(
Transformer2DModel(
num_attention_heads,
out_channels // num_attention_heads,
in_channels=out_channels,
num_layers=transformer_layers_per_block[i],
cross_attention_dim=cross_attention_dim,
norm_num_groups=resnet_groups,
use_linear_projection=use_linear_projection,
only_cross_attention=only_cross_attention,
upcast_attention=upcast_attention,
attention_type=attention_type,
)
)
else:
attentions.append(
DualTransformer2DModel(
num_attention_heads,
out_channels // num_attention_heads,
in_channels=out_channels,
num_layers=1,
cross_attention_dim=cross_attention_dim,
norm_num_groups=resnet_groups,
)
)
self.attentions = nn.ModuleList(attentions)
self.resnets = nn.ModuleList(resnets)
if add_upsample:
self.upsamplers = nn.ModuleList(
[Upsample2D(out_channels, use_conv=True, out_channels=out_channels)]
)
else:
self.upsamplers = None
self.gradient_checkpointing = False
self.resolution_idx = resolution_idx
def forward(
self,
hidden_states: torch.FloatTensor,
res_hidden_states_tuple: Tuple[torch.FloatTensor, ...],
temb: Optional[torch.FloatTensor] = None,
encoder_hidden_states: Optional[torch.FloatTensor] = None,
cross_attention_kwargs: Optional[Dict[str, Any]] = None,
upsample_size: Optional[int] = None,
attention_mask: Optional[torch.FloatTensor] = None,
encoder_attention_mask: Optional[torch.FloatTensor] = None,
) -> torch.FloatTensor:
lora_scale = (
cross_attention_kwargs.get("scale", 1.0)
if cross_attention_kwargs is not None
else 1.0
)
is_freeu_enabled = (
getattr(self, "s1", None)
and getattr(self, "s2", None)
and getattr(self, "b1", None)
and getattr(self, "b2", None)
)
for resnet, attn in zip(self.resnets, self.attentions):
# pop res hidden states
res_hidden_states = res_hidden_states_tuple[-1]
res_hidden_states_tuple = res_hidden_states_tuple[:-1]
# FreeU: Only operate on the first two stages
if is_freeu_enabled:
hidden_states, res_hidden_states = apply_freeu(
self.resolution_idx,
hidden_states,
res_hidden_states,
s1=self.s1,
s2=self.s2,
b1=self.b1,
b2=self.b2,
)
hidden_states = torch.cat([hidden_states, res_hidden_states], dim=1)
if self.training and self.gradient_checkpointing:
def create_custom_forward(module, return_dict=None):
def custom_forward(*inputs):
if return_dict is not None:
return module(*inputs, return_dict=return_dict)
else:
return module(*inputs)
return custom_forward
ckpt_kwargs: Dict[str, Any] = (
{"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
)
hidden_states = torch.utils.checkpoint.checkpoint(
create_custom_forward(resnet),
hidden_states,
temb,
**ckpt_kwargs,
)
hidden_states, ref_feature = attn(
hidden_states,
encoder_hidden_states=encoder_hidden_states,
cross_attention_kwargs=cross_attention_kwargs,
attention_mask=attention_mask,
encoder_attention_mask=encoder_attention_mask,
return_dict=False,
)
else:
hidden_states = resnet(hidden_states, temb, scale=lora_scale)
hidden_states, ref_feature = attn(
hidden_states,
encoder_hidden_states=encoder_hidden_states,
cross_attention_kwargs=cross_attention_kwargs,
attention_mask=attention_mask,
encoder_attention_mask=encoder_attention_mask,
return_dict=False,
)
if self.upsamplers is not None:
for upsampler in self.upsamplers:
hidden_states = upsampler(
hidden_states, upsample_size, scale=lora_scale
)
return hidden_states
class UpBlock2D(nn.Module):
def __init__(
self,
in_channels: int,
prev_output_channel: int,
out_channels: int,
temb_channels: int,
resolution_idx: Optional[int] = None,
dropout: float = 0.0,
num_layers: int = 1,
resnet_eps: float = 1e-6,
resnet_time_scale_shift: str = "default",
resnet_act_fn: str = "swish",
resnet_groups: int = 32,
resnet_pre_norm: bool = True,
output_scale_factor: float = 1.0,
add_upsample: bool = True,
):
super().__init__()
resnets = []
for i in range(num_layers):
res_skip_channels = in_channels if (i == num_layers - 1) else out_channels
resnet_in_channels = prev_output_channel if i == 0 else out_channels
resnets.append(
ResnetBlock2D(
in_channels=resnet_in_channels + res_skip_channels,
out_channels=out_channels,
temb_channels=temb_channels,
eps=resnet_eps,
groups=resnet_groups,
dropout=dropout,
time_embedding_norm=resnet_time_scale_shift,
non_linearity=resnet_act_fn,
output_scale_factor=output_scale_factor,
pre_norm=resnet_pre_norm,
)
)
self.resnets = nn.ModuleList(resnets)
if add_upsample:
self.upsamplers = nn.ModuleList(
[Upsample2D(out_channels, use_conv=True, out_channels=out_channels)]
)
else:
self.upsamplers = None
self.gradient_checkpointing = False
self.resolution_idx = resolution_idx
def forward(
self,
hidden_states: torch.FloatTensor,
res_hidden_states_tuple: Tuple[torch.FloatTensor, ...],
temb: Optional[torch.FloatTensor] = None,
upsample_size: Optional[int] = None,
scale: float = 1.0,
) -> torch.FloatTensor:
is_freeu_enabled = (
getattr(self, "s1", None)
and getattr(self, "s2", None)
and getattr(self, "b1", None)
and getattr(self, "b2", None)
)
for resnet in self.resnets:
# pop res hidden states
res_hidden_states = res_hidden_states_tuple[-1]
res_hidden_states_tuple = res_hidden_states_tuple[:-1]
# FreeU: Only operate on the first two stages
if is_freeu_enabled:
hidden_states, res_hidden_states = apply_freeu(
self.resolution_idx,
hidden_states,
res_hidden_states,
s1=self.s1,
s2=self.s2,
b1=self.b1,
b2=self.b2,
)
hidden_states = torch.cat([hidden_states, res_hidden_states], dim=1)
if self.training and self.gradient_checkpointing:
def create_custom_forward(module):
def custom_forward(*inputs):
return module(*inputs)
return custom_forward
if is_torch_version(">=", "1.11.0"):
hidden_states = torch.utils.checkpoint.checkpoint(
create_custom_forward(resnet),
hidden_states,
temb,
use_reentrant=False,
)
else:
hidden_states = torch.utils.checkpoint.checkpoint(
create_custom_forward(resnet), hidden_states, temb
)
else:
hidden_states = resnet(hidden_states, temb, scale=scale)
if self.upsamplers is not None:
for upsampler in self.upsamplers:
hidden_states = upsampler(hidden_states, upsample_size, scale=scale)
return hidden_states
================================================
FILE: src/models/unet_2d_condition.py
================================================
# Adapted from https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/unet_2d_condition.py
from dataclasses import dataclass
from typing import Any, Dict, List, Optional, Tuple, Union
import torch
import torch.nn as nn
import torch.utils.checkpoint
from diffusers.configuration_utils import ConfigMixin, register_to_config
from diffusers.loaders import UNet2DConditionLoadersMixin
from diffusers.models.activations import get_activation
from diffusers.models.attention_processor import (
ADDED_KV_ATTENTION_PROCESSORS,
CROSS_ATTENTION_PROCESSORS,
AttentionProcessor,
AttnAddedKVProcessor,
AttnProcessor,
)
from diffusers.models.embeddings import (
GaussianFourierProjection,
ImageHintTimeEmbedding,
ImageProjection,
ImageTimeEmbedding,
# PositionNet,
TextImageProjection,
TextImageTimeEmbedding,
TextTimeEmbedding,
TimestepEmbedding,
Timesteps,
)
from diffusers.models.modeling_utils import ModelMixin
from diffusers.utils import (
USE_PEFT_BACKEND,
BaseOutput,
deprecate,
logging,
scale_lora_layers,
unscale_lora_layers,
)
from .unet_2d_blocks import (
UNetMidBlock2D,
UNetMidBlock2DCrossAttn,
get_down_block,
get_up_block,
)
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
@dataclass
class UNet2DConditionOutput(BaseOutput):
"""
The output of [`UNet2DConditionModel`].
Args:
sample (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`):
The hidden states output conditioned on `encoder_hidden_states` input. Output of last layer of model.
"""
sample: torch.FloatTensor = None
ref_features: Tuple[torch.FloatTensor] = None
class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin):
r"""
A conditional 2D UNet model that takes a noisy sample, conditional state, and a timestep and returns a sample
shaped output.
This model inherits from [`ModelMixin`]. Check the superclass documentation for it's generic methods implemented
for all models (such as downloading or saving).
Parameters:
sample_size (`int` or `Tuple[int, int]`, *optional*, defaults to `None`):
Height and width of input/output sample.
in_channels (`int`, *optional*, defaults to 4): Number of channels in the input sample.
out_channels (`int`, *optional*, defaults to 4): Number of channels in the output.
center_input_sample (`bool`, *optional*, defaults to `False`): Whether to center the input sample.
flip_sin_to_cos (`bool`, *optional*, defaults to `False`):
Whether to flip the sin to cos in the time embedding.
freq_shift (`int`, *optional*, defaults to 0): The frequency shift to apply to the time embedding.
down_block_types (`Tuple[str]`, *optional*, defaults to `("CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "DownBlock2D")`):
The tuple of downsample blocks to use.
mid_block_type (`str`, *optional*, defaults to `"UNetMidBlock2DCrossAttn"`):
Block type for middle of UNet, it can be one of `UNetMidBlock2DCrossAttn`, `UNetMidBlock2D`, or
`UNetMidBlock2DSimpleCrossAttn`. If `None`, the mid block layer is skipped.
up_block_types (`Tuple[str]`, *optional*, defaults to `("UpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "CrossAttnUpBlock2D")`):
The tuple of upsample blocks to use.
only_cross_attention(`bool` or `Tuple[bool]`, *optional*, default to `False`):
Whether to include self-attention in the basic transformer blocks, see
[`~models.attention.BasicTransformerBlock`].
block_out_channels (`Tuple[int]`, *optional*, defaults to `(320, 640, 1280, 1280)`):
The tuple of output channels for each block.
layers_per_block (`int`, *optional*, defaults to 2): The number of layers per block.
downsample_padding (`int`, *optional*, defaults to 1): The padding to use for the downsampling convolution.
mid_block_scale_factor (`float`, *optional*, defaults to 1.0): The scale factor to use for the mid block.
dropout (`float`, *optional*, defaults to 0.0): The dropout probability to use.
act_fn (`str`, *optional*, defaults to `"silu"`): The activation function to use.
norm_num_groups (`int`, *optional*, defaults to 32): The number of groups to use for the normalization.
If `None`, normalization and activation layers is skipped in post-processing.
norm_eps (`float`, *optional*, defaults to 1e-5): The epsilon to use for the normalization.
cross_attention_dim (`int` or `Tuple[int]`, *optional*, defaults to 1280):
The dimension of the cross attention features.
transformer_layers_per_block (`int`, `Tuple[int]`, or `Tuple[Tuple]` , *optional*, defaults to 1):
The number of transformer blocks of type [`~models.attention.BasicTransformerBlock`]. Only relevant for
[`~models.unet_2d_blocks.CrossAttnDownBlock2D`], [`~models.unet_2d_blocks.CrossAttnUpBlock2D`],
[`~models.unet_2d_blocks.UNetMidBlock2DCrossAttn`].
reverse_transformer_layers_per_block : (`Tuple[Tuple]`, *optional*, defaults to None):
The number of transformer blocks of type [`~models.attention.BasicTransformerBlock`], in the upsampling
blocks of the U-Net. Only relevant if `transformer_layers_per_block` is of type `Tuple[Tuple]` and for
[`~models.unet_2d_blocks.CrossAttnDownBlock2D`], [`~models.unet_2d_blocks.CrossAttnUpBlock2D`],
[`~models.unet_2d_blocks.UNetMidBlock2DCrossAttn`].
encoder_hid_dim (`int`, *optional*, defaults to None):
If `encoder_hid_dim_type` is defined, `encoder_hidden_states` will be projected from `encoder_hid_dim`
dimension to `cross_attention_dim`.
encoder_hid_dim_type (`str`, *optional*, defaults to `None`):
If given, the `encoder_hidden_states` and potentially other embeddings are down-projected to text
embeddings of dimension `cross_attention` according to `encoder_hid_dim_type`.
attention_head_dim (`int`, *optional*, defaults to 8): The dimension of the attention heads.
num_attention_heads (`int`, *optional*):
The number of attention heads. If not defined, defaults to `attention_head_dim`
resnet_time_scale_shift (`str`, *optional*, defaults to `"default"`): Time scale shift config
for ResNet blocks (see [`~models.resnet.ResnetBlock2D`]). Choose from `default` or `scale_shift`.
class_embed_type (`str`, *optional*, defaults to `None`):
The type of class embedding to use which is ultimately summed with the time embeddings. Choose from `None`,
`"timestep"`, `"identity"`, `"projection"`, or `"simple_projection"`.
addition_embed_type (`str`, *optional*, defaults to `None`):
Configures an optional embedding which will be summed with the time embeddings. Choose from `None` or
"text". "text" will use the `TextTimeEmbedding` layer.
addition_time_embed_dim: (`int`, *optional*, defaults to `None`):
Dimension for the timestep embeddings.
num_class_embeds (`int`, *optional*, defaults to `None`):
Input dimension of the learnable embedding matrix to be projected to `time_embed_dim`, when performing
class conditioning with `class_embed_type` equal to `None`.
time_embedding_type (`str`, *optional*, defaults to `positional`):
The type of position embedding to use for timesteps. Choose from `positional` or `fourier`.
time_embedding_dim (`int`, *optional*, defaults to `None`):
An optional override for the dimension of the projected time embedding.
time_embedding_act_fn (`str`, *optional*, defaults to `None`):
Optional activation function to use only once on the time embeddings before they are passed to the rest of
the UNet. Choose from `silu`, `mish`, `gelu`, and `swish`.
timestep_post_act (`str`, *optional*, defaults to `None`):
The second activation function to use in timestep embedding. Choose from `silu`, `mish` and `gelu`.
time_cond_proj_dim (`int`, *optional*, defaults to `None`):
The dimension of `cond_proj` layer in the timestep embedding.
conv_in_kernel (`int`, *optional*, default to `3`): The kernel size of `conv_in` layer. conv_out_kernel (`int`,
*optional*, default to `3`): The kernel size of `conv_out` layer. projection_class_embeddings_input_dim (`int`,
*optional*): The dimension of the `class_labels` input when
`class_embed_type="projection"`. Required when `class_embed_type="projection"`.
class_embeddings_concat (`bool`, *optional*, defaults to `False`): Whether to concatenate the time
embeddings with the class embeddings.
mid_block_only_cross_attention (`bool`, *optional*, defaults to `None`):
Whether to use cross attention with the mid block when using the `UNetMidBlock2DSimpleCrossAttn`. If
`only_cross_attention` is given as a single boolean and `mid_block_only_cross_attention` is `None`, the
`only_cross_attention` value is used as the value for `mid_block_only_cross_attention`. Default to `False`
otherwise.
"""
_supports_gradient_checkpointing = True
@register_to_config
def __init__(
self,
sample_size: Optional[int] = None,
in_channels: int = 4,
out_channels: int = 4,
center_input_sample: bool = False,
flip_sin_to_cos: bool = True,
freq_shift: int = 0,
down_block_types: Tuple[str] = (
"CrossAttnDownBlock2D",
"CrossAttnDownBlock2D",
"CrossAttnDownBlock2D",
"DownBlock2D",
),
mid_block_type: Optional[str] = "UNetMidBlock2DCrossAttn",
up_block_types: Tuple[str] = (
"UpBlock2D",
"CrossAttnUpBlock2D",
"CrossAttnUpBlock2D",
"CrossAttnUpBlock2D",
),
only_cross_attention: Union[bool, Tuple[bool]] = False,
block_out_channels: Tuple[int] = (320, 640, 1280, 1280),
layers_per_block: Union[int, Tuple[int]] = 2,
downsample_padding: int = 1,
mid_block_scale_factor: float = 1,
dropout: float = 0.0,
act_fn: str = "silu",
norm_num_groups: Optional[int] = 32,
norm_eps: float = 1e-5,
cross_attention_dim: Union[int, Tuple[int]] = 1280,
transformer_layers_per_block: Union[int, Tuple[int], Tuple[Tuple]] = 1,
reverse_transformer_layers_per_block: Optional[Tuple[Tuple[int]]] = None,
encoder_hid_dim: Optional[int] = None,
encoder_hid_dim_type: Optional[str] = None,
attention_head_dim: Union[int, Tuple[int]] = 8,
num_attention_heads: Optional[Union[int, Tuple[int]]] = None,
dual_cross_attention: bool = False,
use_linear_projection: bool = False,
class_embed_type: Optional[str] = None,
addition_embed_type: Optional[str] = None,
addition_time_embed_dim: Optional[int] = None,
num_class_embeds: Optional[int] = None,
upcast_attention: bool = False,
resnet_time_scale_shift: str = "default",
resnet_skip_time_act: bool = False,
resnet_out_scale_factor: int = 1.0,
time_embedding_type: str = "positional",
time_embedding_dim: Optional[int] = None,
time_embedding_act_fn: Optional[str] = None,
timestep_post_act: Optional[str] = None,
time_cond_proj_dim: Optional[int] = None,
conv_in_kernel: int = 3,
conv_out_kernel: int = 3,
projection_class_embeddings_input_dim: Optional[int] = None,
attention_type: str = "default",
class_embeddings_concat: bool = False,
mid_block_only_cross_attention: Optional[bool] = None,
cross_attention_norm: Optional[str] = None,
addition_embed_type_num_heads=64,
):
super().__init__()
self.sample_size = sample_size
if num_attention_heads is not None:
raise ValueError(
"At the moment it is not possible to define the number of attention heads via `num_attention_heads` because of a naming issue as described in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131. Passing `num_attention_heads` will only be supported in diffusers v0.19."
)
# If `num_attention_heads` is not defined (which is the case for most models)
# it will default to `attention_head_dim`. This looks weird upon first reading it and it is.
# The reason for this behavior is to correct for incorrectly named variables that were introduced
# when this library was created. The incorrect naming was only discovered much later in https://github.com/huggingface/diffusers/issues/2011#issuecomment-1547958131
# Changing `attention_head_dim` to `num_attention_heads` for 40,000+ configurations is too backwards breaking
# which is why we correct for the naming here.
num_attention_heads = num_attention_heads or attention_head_dim
# Check inputs
if len(down_block_types) != len(up_block_types):
raise ValueError(
f"Must provide the same number of `down_block_types` as `up_block_types`. `down_block_types`: {down_block_types}. `up_block_types`: {up_block_types}."
)
if len(block_out_channels) != len(down_block_types):
raise ValueError(
f"Must provide the same number of `block_out_channels` as `down_block_types`. `block_out_channels`: {block_out_channels}. `down_block_types`: {down_block_types}."
)
if not isinstance(only_cross_attention, bool) and len(
only_cross_attention
) != len(down_block_types):
raise ValueError(
f"Must provide the same number of `only_cross_attention` as `down_block_types`. `only_cross_attention`: {only_cross_attention}. `down_block_types`: {down_block_types}."
)
if not isinstance(num_attention_heads, int) and len(num_attention_heads) != len(
down_block_types
):
raise ValueError(
f"Must provide the same number of `num_attention_heads` as `down_block_types`. `num_attention_heads`: {num_attention_heads}. `down_block_types`: {down_block_types}."
)
if not isinstance(attention_head_dim, int) and len(attention_head_dim) != len(
down_block_types
):
raise ValueError(
f"Must provide the same number of `attention_head_dim` as `down_block_types`. `attention_head_dim`: {attention_head_dim}. `down_block_types`: {down_block_types}."
)
if isinstance(cross_attention_dim, list) and len(cross_attention_dim) != len(
down_block_types
):
raise ValueError(
f"Must provide the same number of `cross_attention_dim` as `down_block_types`. `cross_attention_dim`: {cross_attention_dim}. `down_block_types`: {down_block_types}."
)
if not isinstance(layers_per_block, int) and len(layers_per_block) != len(
down_block_types
):
raise ValueError(
f"Must provide the same number of `layers_per_block` as `down_block_types`. `layers_per_block`: {layers_per_block}. `down_block_types`: {down_block_types}."
)
if (
isinstance(transformer_layers_per_block, list)
and reverse_transformer_layers_per_block is None
):
for layer_number_per_block in transformer_layers_per_block:
if isinstance(layer_number_per_block, list):
raise ValueError(
"Must provide 'reverse_transformer_layers_per_block` if using asymmetrical UNet."
)
# input
conv_in_padding = (conv_in_kernel - 1) // 2
self.conv_in = nn.Conv2d(
in_channels,
block_out_channels[0],
kernel_size=conv_in_kernel,
padding=conv_in_padding,
)
# time
if time_embedding_type == "fourier":
time_embed_dim = time_embedding_dim or block_out_channels[0] * 2
if time_embed_dim % 2 != 0:
raise ValueError(
f"`time_embed_dim` should be divisible by 2, but is {time_embed_dim}."
)
self.time_proj = GaussianFourierProjection(
time_embed_dim // 2,
set_W_to_weight=False,
log=False,
flip_sin_to_cos=flip_sin_to_cos,
)
timestep_input_dim = time_embed_dim
elif time_embedding_type == "positional":
time_embed_dim = time_embedding_dim or block_out_channels[0] * 4
self.time_proj = Timesteps(
block_out_channels[0], flip_sin_to_cos, freq_shift
)
timestep_input_dim = block_out_channels[0]
else:
raise ValueError(
f"{time_embedding_type} does not exist. Please make sure to use one of `fourier` or `positional`."
)
self.time_embedding = TimestepEmbedding(
timestep_input_dim,
time_embed_dim,
act_fn=act_fn,
post_act_fn=timestep_post_act,
cond_proj_dim=time_cond_proj_dim,
)
if encoder_hid_dim_type is None and encoder_hid_dim is not None:
encoder_hid_dim_type = "text_proj"
self.register_to_config(encoder_hid_dim_type=encoder_hid_dim_type)
logger.info(
"encoder_hid_dim_type defaults to 'text_proj' as `encoder_hid_dim` is defined."
)
if encoder_hid_dim is None and encoder_hid_dim_type is not None:
raise ValueError(
f"`encoder_hid_dim` has to be defined when `encoder_hid_dim_type` is set to {encoder_hid_dim_type}."
)
if encoder_hid_dim_type == "text_proj":
self.encoder_hid_proj = nn.Linear(encoder_hid_dim, cross_attention_dim)
elif encoder_hid_dim_type == "text_image_proj":
# image_embed_dim DOESN'T have to be `cross_attention_dim`. To not clutter the __init__ too much
# they are set to `cross_attention_dim` here as this is exactly the required dimension for the currently only use
# case when `addition_embed_type == "text_image_proj"` (Kadinsky 2.1)`
self.encoder_hid_proj = TextImageProjection(
text_embed_dim=encoder_hid_dim,
image_embed_dim=cross_attention_dim,
cross_attention_dim=cross_attention_dim,
)
elif encoder_hid_dim_type == "image_proj":
# Kandinsky 2.2
self.encoder_hid_proj = ImageProjection(
image_embed_dim=encoder_hid_dim,
cross_attention_dim=cross_attention_dim,
)
elif encoder_hid_dim_type is not None:
raise ValueError(
f"encoder_hid_dim_type: {encoder_hid_dim_type} must be None, 'text_proj' or 'text_image_proj'."
)
else:
self.encoder_hid_proj = None
# class embedding
if class_embed_type is None and num_class_embeds is not None:
self.class_embedding = nn.Embedding(num_class_embeds, time_embed_dim)
elif class_embed_type == "timestep":
self.class_embedding = TimestepEmbedding(
timestep_input_dim, time_embed_dim, act_fn=act_fn
)
elif class_embed_type == "identity":
self.class_embedding = nn.Identity(time_embed_dim, time_embed_dim)
elif class_embed_type == "projection":
if projection_class_embeddings_input_dim is None:
raise ValueError(
"`class_embed_type`: 'projection' requires `projection_class_embeddings_input_dim` be set"
)
# The projection `class_embed_type` is the same as the timestep `class_embed_type` except
# 1. the `class_labels` inputs are not first converted to sinusoidal embeddings
# 2. it projects from an arbitrary input dimension.
#
# Note that `TimestepEmbedding` is quite general, being mainly linear layers and activations.
# When used for embedding actual timesteps, the timesteps are first converted to sinusoidal embeddings.
# As a result, `TimestepEmbedding` can be passed arbitrary vectors.
self.class_embedding = TimestepEmbedding(
projection_class_embeddings_input_dim, time_embed_dim
)
elif class_embed_type == "simple_projection":
if projection_class_embeddings_input_dim is None:
raise ValueError(
"`class_embed_type`: 'simple_projection' requires `projection_class_embeddings_input_dim` be set"
)
self.class_embedding = nn.Linear(
projection_class_embeddings_input_dim, time_embed_dim
)
else:
self.class_embedding = None
if addition_embed_type == "text":
if encoder_hid_dim is not None:
text_time_embedding_from_dim = encoder_hid_dim
else:
text_time_embedding_from_dim = cross_attention_dim
self.add_embedding = TextTimeEmbedding(
text_time_embedding_from_dim,
time_embed_dim,
num_heads=addition_embed_type_num_heads,
)
elif addition_embed_type == "text_image":
# text_embed_dim and image_embed_dim DON'T have to be `cross_attention_dim`. To not clutter the __init__ too much
# they are set to `cross_attention_dim` here as this is exactly the required dimension for the currently only use
# case when `addition_embed_type == "text_image"` (Kadinsky 2.1)`
self.add_embedding = TextImageTimeEmbedding(
text_embed_dim=cross_attention_dim,
image_embed_dim=cross_attention_dim,
time_embed_dim=time_embed_dim,
)
elif addition_embed_type == "text_time":
self.add_time_proj = Timesteps(
addition_time_embed_dim, flip_sin_to_cos, freq_shift
)
self.add_embedding = TimestepEmbedding(
projection_class_embeddings_input_dim, time_embed_dim
)
elif addition_embed_type == "image":
# Kandinsky 2.2
self.add_embedding = ImageTimeEmbedding(
image_embed_dim=encoder_hid_dim, time_embed_dim=time_embed_dim
)
elif addition_embed_type == "image_hint":
# Kandinsky 2.2 ControlNet
self.add_embedding = ImageHintTimeEmbedding(
image_embed_dim=encoder_hid_dim, time_embed_dim=time_embed_dim
)
elif addition_embed_type is not None:
raise ValueError(
f"addition_embed_type: {addition_embed_type} must be None, 'text' or 'text_image'."
)
if time_embedding_act_fn is None:
self.time_embed_act = None
else:
self.time_embed_act = get_activation(time_embedding_act_fn)
self.down_blocks = nn.ModuleList([])
self.up_blocks = nn.ModuleList([])
if isinstance(only_cross_attention, bool):
if mid_block_only_cross_attention is None:
mid_block_only_cross_attention = only_cross_attention
only_cross_attention = [only_cross_attention] * len(down_block_types)
if mid_block_only_cross_attention is None:
mid_block_only_cross_attention = False
if isinstance(num_attention_heads, int):
num_attention_heads = (num_attention_heads,) * len(down_block_types)
if isinstance(attention_head_dim, int):
attention_head_dim = (attention_head_dim,) * len(down_block_types)
if isinstance(cross_attention_dim, int):
cross_attention_dim = (cross_attention_dim,) * len(down_block_types)
if isinstance(layers_per_block, int):
layers_per_block = [layers_per_block] * len(down_block_types)
if isinstance(transformer_layers_per_block, int):
transformer_layers_per_block = [transformer_layers_per_block] * len(
down_block_types
)
if class_embeddings_concat:
# The time embeddings are concatenated with the class embeddings. The dimension of the
# time embeddings passed to the down, middle, and up blocks is twice the dimension of the
# regular time embeddings
blocks_time_embed_dim = time_embed_dim * 2
else:
blocks_time_embed_dim = time_embed_dim
# down
output_channel = block_out_channels[0]
for i, down_block_type in enumerate(down_block_types):
input_channel = output_channel
output_channel = block_out_channels[i]
is_final_block = i == len(block_out_channels) - 1
down_block = get_down_block(
down_block_type,
num_layers=layers_per_block[i],
transformer_layers_per_block=transformer_layers_per_block[i],
in_channels=input_channel,
out_channels=output_channel,
temb_channels=blocks_time_embed_dim,
add_downsample=not is_final_block,
resnet_eps=norm_eps,
resnet_act_fn=act_fn,
resnet_groups=norm_num_groups,
cross_attention_dim=cross_attention_dim[i],
num_attention_heads=num_attention_heads[i],
downsample_padding=downsample_padding,
dual_cross_attention=dual_cross_attention,
use_linear_projection=use_linear_projection,
only_cross_attention=only_cross_attention[i],
upcast_attention=upcast_attention,
resnet_time_scale_shift=resnet_time_scale_shift,
attention_type=attention_type,
resnet_skip_time_act=resnet_skip_time_act,
resnet_out_scale_factor=resnet_out_scale_factor,
cross_attention_norm=cross_attention_norm,
attention_head_dim=attention_head_dim[i]
if attention_head_dim[i] is not None
else output_channel,
dropout=dropout,
)
self.down_blocks.append(down_block)
# mid
if mid_block_type == "UNetMidBlock2DCrossAttn":
self.mid_block = UNetMidBlock2DCrossAttn(
transformer_layers_per_block=transformer_layers_per_block[-1],
in_channels=block_out_channels[-1],
temb_channels=blocks_time_embed_dim,
dropout=dropout,
resnet_eps=norm_eps,
resnet_act_fn=act_fn,
output_scale_factor=mid_block_scale_factor,
resnet_time_scale_shift=resnet_time_scale_shift,
cross_attention_dim=cross_attention_dim[-1],
num_attention_heads=num_attention_heads[-1],
resnet_groups=norm_num_groups,
dual_cross_attention=dual_cross_attention,
use_linear_projection=use_linear_projection,
upcast_attention=upcast_attention,
attention_type=attention_type,
)
elif mid_block_type == "UNetMidBlock2DSimpleCrossAttn":
raise NotImplementedError(f"Unsupport mid_block_type: {mid_block_type}")
elif mid_block_type == "UNetMidBlock2D":
self.mid_block = UNetMidBlock2D(
in_channels=block_out_channels[-1],
temb_channels=blocks_time_embed_dim,
dropout=dropout,
num_layers=0,
resnet_eps=norm_eps,
resnet_act_fn=act_fn,
output_scale_factor=mid_block_scale_factor,
resnet_groups=norm_num_groups,
resnet_time_scale_shift=resnet_time_scale_shift,
add_attention=False,
)
elif mid_block_type is None:
self.mid_block = None
else:
raise ValueError(f"unknown mid_block_type : {mid_block_type}")
# count how many layers upsample the images
self.num_upsamplers = 0
# up
reversed_block_out_channels = list(reversed(block_out_channels))
reversed_num_attention_heads = list(reversed(num_attention_heads))
reversed_layers_per_block = list(reversed(layers_per_block))
reversed_cross_attention_dim = list(reversed(cross_attention_dim))
reversed_transformer_layers_per_block = (
list(reversed(transformer_layers_per_block))
if reverse_transformer_layers_per_block is None
else reverse_transformer_layers_per_block
)
only_cross_attention = list(reversed(only_cross_attention))
output_channel = reversed_block_out_channels[0]
for i, up_block_type in enumerate(up_block_types):
is_final_block = i == len(block_out_channels) - 1
prev_output_channel = output_channel
output_channel = reversed_block_out_channels[i]
input_channel = reversed_block_out_channels[
min(i + 1, len(block_out_channels) - 1)
]
# add upsample block for all BUT final layer
if not is_final_block:
add_upsample = True
self.num_upsamplers += 1
else:
add_upsample = False
up_block = get_up_block(
up_block_type,
num_layers=reversed_layers_per_block[i] + 1,
transformer_layers_per_block=reversed_transformer_layers_per_block[i],
in_channels=input_channel,
out_channels=output_channel,
prev_output_channel=prev_output_channel,
temb_channels=blocks_time_embed_dim,
add_upsample=add_upsample,
resnet_eps=norm_eps,
resnet_act_fn=act_fn,
resolution_idx=i,
resnet_groups=norm_num_groups,
cross_attention_dim=reversed_cross_attention_dim[i],
num_attention_heads=reversed_num_attention_heads[i],
dual_cross_attention=dual_cross_attention,
use_linear_projection=use_linear_projection,
only_cross_attention=only_cross_attention[i],
upcast_attention=upcast_attention,
resnet_time_scale_shift=resnet_time_scale_shift,
attention_type=attention_type,
resnet_skip_time_act=resnet_skip_time_act,
resnet_out_scale_factor=resnet_out_scale_factor,
cross_attention_norm=cross_attention_norm,
attention_head_dim=attention_head_dim[i]
if attention_head_dim[i] is not None
else output_channel,
dropout=dropout,
)
self.up_blocks.append(up_block)
prev_output_channel = output_channel
# out
if norm_num_groups is not None:
self.conv_norm_out = nn.GroupNorm(
num_channels=block_out_channels[0],
num_groups=norm_num_groups,
eps=norm_eps,
)
self.conv_act = get_activation(act_fn)
else:
self.conv_norm_out = None
self.conv_act = None
self.conv_norm_out = None
conv_out_padding = (conv_out_kernel - 1) // 2
# self.conv_out = nn.Conv2d(
# block_out_channels[0],
# out_channels,
# kernel_size=conv_out_kernel,
# padding=conv_out_padding,
# )
if attention_type in ["gated", "gated-text-image"]:
positive_len = 768
if isinstance(cross_attention_dim, int):
positive_len = cross_attention_dim
elif isinstance(cross_attention_dim, tuple) or isinstance(
cross_attention_dim, list
):
positive_len = cross_attention_dim[0]
feature_type = "text-only" if attention_type == "gated" else "text-image"
# self.position_net = PositionNet(
# positive_len=positive_len,
# out_dim=cross_attention_dim,
# feature_type=feature_type,
# )
@property
def attn_processors(self) -> Dict[str, AttentionProcessor]:
r"""
Returns:
`dict` of attention processors: A dictionary containing all attention processors used in the model with
indexed by its weight name.
"""
# set recursively
processors = {}
def fn_recursive_add_processors(
name: str,
module: torch.nn.Module,
processors: Dict[str, AttentionProcessor],
):
if hasattr(module, "get_processor"):
processors[f"{name}.processor"] = module.get_processor(
return_deprecated_lora=True
)
for sub_name, child in module.named_children():
fn_recursive_add_processors(f"{name}.{sub_name}", child, processors)
return processors
for name, module in self.named_children():
fn_recursive_add_processors(name, module, processors)
return processors
def set_attn_processor(
self,
processor: Union[AttentionProcessor, Dict[str, AttentionProcessor]],
_remove_lora=False,
):
r"""
Sets the attention processor to use to compute attention.
Parameters:
processor (`dict` of `AttentionProcessor` or only `AttentionProcessor`):
The instantiated processor class or a dictionary of processor classes that will be set as the processor
for **all** `Attention` layers.
If `processor` is a dict, the key needs to define the path to the corresponding cross attention
processor. This is strongly recommended when setting trainable attention processors.
"""
count = len(self.attn_processors.keys())
if isinstance(processor, dict) and len(processor) != count:
raise ValueError(
f"A dict of processors was passed, but the number of processors {len(processor)} does not match the"
f" number of attention layers: {count}. Please make sure to pass {count} processor classes."
)
def fn_recursive_attn_processor(name: str, module: torch.nn.Module, processor):
if hasattr(module, "set_processor"):
if not isinstance(processor, dict):
module.set_processor(processor, _remove_lora=_remove_lora)
else:
module.set_processor(
processor.pop(f"{name}.processor"), _remove_lora=_remove_lora
)
for sub_name, child in module.named_children():
fn_recursive_attn_processor(f"{name}.{sub_name}", child, processor)
for name, module in self.named_children():
fn_recursive_attn_processor(name, module, processor)
def set_default_attn_processor(self):
"""
Disables custom attention processors and sets the default attention implementation.
"""
if all(
proc.__class__ in ADDED_KV_ATTENTION_PROCESSORS
for proc in self.attn_processors.values()
):
processor = AttnAddedKVProcessor()
elif all(
proc.__class__ in CROSS_ATTENTION_PROCESSORS
for proc in self.attn_processors.values()
):
processor = AttnProcessor()
else:
raise ValueError(
f"Cannot call `set_default_attn_processor` when attention processors are of type {next(iter(self.attn_processors.values()))}"
)
self.set_attn_processor(processor, _remove_lora=True)
def set_attention_slice(self, slice_size):
r"""
Enable sliced attention computation.
When this option is enabled, the attention module splits the input tensor in slices to compute attention in
several steps. This is useful for saving some memory in exchange for a small decrease in speed.
Args:
slice_size (`str` or `int` or `list(int)`, *optional*, defaults to `"auto"`):
When `"auto"`, input to the attention heads is halved, so attention is computed in two steps. If
`"max"`, maximum amount of memory is saved by running only one slice at a time. If a number is
provided, uses as many slices as `attention_head_dim // slice_size`. In this case, `attention_head_dim`
must be a multiple of `slice_size`.
"""
sliceable_head_dims = []
def fn_recursive_retrieve_sliceable_dims(module: torch.nn.Module):
if hasattr(module, "set_attention_slice"):
sliceable_head_dims.append(module.sliceable_head_dim)
for child in module.children():
fn_recursive_retrieve_sliceable_dims(child)
# retrieve number of attention layers
for module in self.children():
fn_recursive_retrieve_sliceable_dims(module)
num_sliceable_layers = len(sliceable_head_dims)
if slice_size == "auto":
# half the attention head size is usually a good trade-off between
# speed and memory
slice_size = [dim // 2 for dim in sliceable_head_dims]
elif slice_size == "max":
# make smallest slice possible
slice_size = num_sliceable_layers * [1]
slice_size = (
num_sliceable_layers * [slice_size]
if not isinstance(slice_size, list)
else slice_size
)
if len(slice_size) != len(sliceable_head_dims):
raise ValueError(
f"You have provided {len(slice_size)}, but {self.config} has {len(sliceable_head_dims)} different"
f" attention layers. Make sure to match `len(slice_size)` to be {len(sliceable_head_dims)}."
)
for i in range(len(slice_size)):
size = slice_size[i]
dim = sliceable_head_dims[i]
if size is not None and size > dim:
raise ValueError(f"size {size} has to be smaller or equal to {dim}.")
# Recursively walk through all the children.
# Any children which exposes the set_attention_slice method
# gets the message
def fn_recursive_set_attention_slice(
module: torch.nn.Module, slice_size: List[int]
):
if hasattr(module, "set_attention_slice"):
module.set_attention_slice(slice_size.pop())
for child in module.children():
fn_recursive_set_attention_slice(child, slice_size)
reversed_slice_size = list(reversed(slice_size))
for module in self.children():
fn_recursive_set_attention_slice(module, reversed_slice_size)
def _set_gradient_checkpointing(self, module, value=False):
if hasattr(module, "gradient_checkpointing"):
module.gradient_checkpointing = value
def enable_freeu(self, s1, s2, b1, b2):
r"""Enables the FreeU mechanism from https://arxiv.org/abs/2309.11497.
The suffixes after the scaling factors represent the stage blocks where they are being applied.
Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of values that
are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
Args:
s1 (`float`):
Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
mitigate the "oversmoothing effect" in the enhanced denoising process.
s2 (`float`):
Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
mitigate the "oversmoothing effect" in the enhanced denoising process.
b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
"""
for i, upsample_block in enumerate(self.up_blocks):
setattr(upsample_block, "s1", s1)
setattr(upsample_block, "s2", s2)
setattr(upsample_block, "b1", b1)
setattr(upsample_block, "b2", b2)
def disable_freeu(self):
"""Disables the FreeU mechanism."""
freeu_keys = {"s1", "s2", "b1", "b2"}
for i, upsample_block in enumerate(self.up_blocks):
for k in freeu_keys:
if (
hasattr(upsample_block, k)
or getattr(upsample_block, k, None) is not None
):
setattr(upsample_block, k, None)
def forward(
self,
sample: torch.FloatTensor,
timestep: Union[torch.Tensor, float, int],
encoder_hidden_states: torch.Tensor,
class_labels: Optional[torch.Tensor] = None,
timestep_cond: Optional[torch.Tensor] = None,
attention_mask: Optional[torch.Tensor] = None,
cross_attention_kwargs: Optional[Dict[str, Any]] = None,
added_cond_kwargs: Optional[Dict[str, torch.Tensor]] = None,
down_block_additional_residuals: Optional[Tuple[torch.Tensor]] = None,
mid_block_additional_residual: Optional[torch.Tensor] = None,
down_intrablock_additional_residuals: Optional[Tuple[torch.Tensor]] = None,
encoder_attention_mask: Optional[torch.Tensor] = None,
return_dict: bool = True,
) -> Union[UNet2DConditionOutput, Tuple]:
r"""
The [`UNet2DConditionModel`] forward method.
Args:
sample (`torch.FloatTensor`):
The noisy input tensor with the following shape `(batch, channel, height, width)`.
timestep (`torch.FloatTensor` or `float` or `int`): The number of timesteps to denoise an input.
encoder_hidden_states (`torch.FloatTensor`):
The encoder hidden states with shape `(batch, sequence_length, feature_dim)`.
class_labels (`torch.Tensor`, *optional*, defaults to `None`):
Optional class labels for conditioning. Their embeddings will be summed with the timestep embeddings.
timestep_cond: (`torch.Tensor`, *optional*, defaults to `None`):
Conditional embeddings for timestep. If provided, the embeddings will be summed with the samples passed
through the `self.time_embedding` layer to obtain the timestep embeddings.
attention_mask (`torch.Tensor`, *optional*, defaults to `None`):
An attention mask of shape `(batch, key_tokens)` is applied to `encoder_hidden_states`. If `1` the mask
is kept, otherwise if `0` it is discarded. Mask will be converted into a bias, which adds large
negative values to the attention scores corresponding to "discard" tokens.
cross_attention_kwargs (`dict`, *optional*):
A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
`self.processor` in
[diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
added_cond_kwargs: (`dict`, *optional*):
A kwargs dictionary containing additional embeddings that if specified are added to the embeddings that
are passed along to the UNet blocks.
down_block_additional_residuals: (`tuple` of `torch.Tensor`, *optional*):
A tuple of tensors that if specified are added to the residuals of down unet blocks.
mid_block_additional_residual: (`torch.Tensor`, *optional*):
A tensor that if specified is added to the residual of the middle unet block.
encoder_attention_mask (`torch.Tensor`):
A cross-attention mask of shape `(batch, sequence_length)` is applied to `encoder_hidden_states`. If
`True` the mask is kept, otherwise if `False` it is discarded. Mask will be converted into a bias,
which adds large negative values to the attention scores corresponding to "discard" tokens.
return_dict (`bool`, *optional*, defaults to `True`):
Whether or not to return a [`~models.unet_2d_condition.UNet2DConditionOutput`] instead of a plain
tuple.
cross_attention_kwargs (`dict`, *optional*):
A kwargs dictionary that if specified is passed along to the [`AttnProcessor`].
added_cond_kwargs: (`dict`, *optional*):
A kwargs dictionary containin additional embeddings that if specified are added to the embeddings that
are passed along to the UNet blocks.
down_block_additional_residuals (`tuple` of `torch.Tensor`, *optional*):
additional residuals to be added to UNet long skip connections from down blocks to up blocks for
example from ControlNet side model(s)
mid_block_additional_residual (`torch.Tensor`, *optional*):
additional residual to be added to UNet mid block output, for example from ControlNet side model
down_intrablock_additional_residuals (`tuple` of `torch.Tensor`, *optional*):
additional residuals to be added within UNet down blocks, for example from T2I-Adapter side model(s)
Returns:
[`~models.unet_2d_condition.UNet2DConditionOutput`] or `tuple`:
If `return_dict` is True, an [`~models.unet_2d_condition.UNet2DConditionOutput`] is returned, otherwise
a `tuple` is returned where the first element is the sample tensor.
"""
# By default samples have to be AT least a multiple of the overall upsampling factor.
# The overall upsampling factor is equal to 2 ** (# num of upsampling layers).
# However, the upsampling interpolation output size can be forced to fit any upsampling size
# on the fly if necessary.
default_overall_up_factor = 2**self.num_upsamplers
# upsample size should be forwarded when sample is not a multiple of `default_overall_up_factor`
forward_upsample_size = False
upsample_size = None
for dim in sample.shape[-2:]:
if dim % default_overall_up_factor != 0:
# Forward upsample size to force interpolation output size.
forward_upsample_size = True
break
# ensure attention_mask is a bias, and give it a singleton query_tokens dimension
# expects mask of shape:
# [batch, key_tokens]
# adds singleton query_tokens dimension:
# [batch, 1, key_tokens]
# this helps to broadcast it as a bias over attention scores, which will be in one of the following shapes:
# [batch, heads, query_tokens, key_tokens] (e.g. torch sdp attn)
# [batch * heads, query_tokens, key_tokens] (e.g. xformers or classic attn)
if attention_mask is not None:
# assume that mask is expressed as:
# (1 = keep, 0 = discard)
# convert mask into a bias that can be added to attention scores:
# (keep = +0, discard = -10000.0)
attention_mask = (1 - attention_mask.to(sample.dtype)) * -10000.0
attention_mask = attention_mask.unsqueeze(1)
# convert encoder_attention_mask to a bias the same way we do for attention_mask
if encoder_attention_mask is not None:
encoder_attention_mask = (
1 - encoder_attention_mask.to(sample.dtype)
) * -10000.0
encoder_attention_mask = encoder_attention_mask.unsqueeze(1)
# 0. center input if necessary
if self.config.center_input_sample:
sample = 2 * sample - 1.0
# 1. time
timesteps = timestep
if not torch.is_tensor(timesteps):
# TODO: this requires sync between CPU and GPU. So try to pass timesteps as tensors if you can
# This would be a good case for the `match` statement (Python 3.10+)
is_mps = sample.device.type == "mps"
if isinstance(timestep, float):
dtype = torch.float32 if is_mps else torch.float64
else:
dtype = torch.int32 if is_mps else torch.int64
timesteps = torch.tensor([timesteps], dtype=dtype, device=sample.device)
elif len(timesteps.shape) == 0:
timesteps = timesteps[None].to(sample.device)
# broadcast to batch dimension in a way that's compatible with ONNX/Core ML
timesteps = timesteps.expand(sample.shape[0])
t_emb = self.time_proj(timesteps)
# `Timesteps` does not contain any weights and will always return f32 tensors
# but time_embedding might actually be running in fp16. so we need to cast here.
# there might be better ways to encapsulate this.
t_emb = t_emb.to(dtype=sample.dtype)
emb = self.time_embedding(t_emb, timestep_cond)
aug_emb = None
if self.class_embedding is not None:
if class_labels is None:
raise ValueError(
"class_labels should be provided when num_class_embeds > 0"
)
if self.config.class_embed_type == "timestep":
class_labels = self.time_proj(class_labels)
# `Timesteps` does not contain any weights and will always return f32 tensors
# there might be better ways to encapsulate this.
class_labels = class_labels.to(dtype=sample.dtype)
class_emb = self.class_embedding(class_labels).to(dtype=sample.dtype)
if self.config.class_embeddings_concat:
emb = torch.cat([emb, class_emb], dim=-1)
else:
emb = emb + class_emb
if self.config.addition_embed_type == "text":
aug_emb = self.add_embedding(encoder_hidden_states)
elif self.config.addition_embed_type == "text_image":
# Kandinsky 2.1 - style
if "image_embeds" not in added_cond_kwargs:
raise ValueError(
f"{self.__class__} has the config param `addition_embed_type` set to 'text_image' which requires the keyword argument `image_embeds` to be passed in `added_cond_kwargs`"
)
image_embs = added_cond_kwargs.get("image_embeds")
text_embs = added_cond_kwargs.get("text_embeds", encoder_hidden_states)
aug_emb = self.add_embedding(text_embs, image_embs)
elif self.config.addition_embed_type == "text_time":
# SDXL - style
if "text_embeds" not in added_cond_kwargs:
raise ValueError(
f"{self.__class__} has the config param `addition_embed_type` set to 'text_time' which requires the keyword argument `text_embeds` to be passed in `added_cond_kwargs`"
)
text_embeds = added_cond_kwargs.get("text_embeds")
if "time_ids" not in added_cond_kwargs:
raise ValueError(
f"{self.__class__} has the config param `addition_embed_type` set to 'text_time' which requires the keyword argument `time_ids` to be passed in `added_cond_kwargs`"
)
time_ids = added_cond_kwargs.get("time_ids")
time_embeds = self.add_time_proj(time_ids.flatten())
time_embeds = time_embeds.reshape((text_embeds.shape[0], -1))
add_embeds = torch.concat([text_embeds, time_embeds], dim=-1)
add_embeds = add_embeds.to(emb.dtype)
aug_emb = self.add_embedding(add_embeds)
elif self.config.addition_embed_type == "image":
# Kandinsky 2.2 - style
if "image_embeds" not in added_cond_kwargs:
raise ValueError(
f"{self.__class__} has the config param `addition_embed_type` set to 'image' which requires the keyword argument `image_embeds` to be passed in `added_cond_kwargs`"
)
image_embs = added_cond_kwargs.get("image_embeds")
aug_emb = self.add_embedding(image_embs)
elif self.config.addition_embed_type == "image_hint":
# Kandinsky 2.2 - style
if (
"image_embeds" not in added_cond_kwargs
or "hint" not in added_cond_kwargs
):
raise ValueError(
f"{self.__class__} has the config param `addition_embed_type` set to 'image_hint' which requires the keyword arguments `image_embeds` and `hint` to be passed in `added_cond_kwargs`"
)
image_embs = added_cond_kwargs.get("image_embeds")
hint = added_cond_kwargs.get("hint")
aug_emb, hint = self.add_embedding(image_embs, hint)
sample = torch.cat([sample, hint], dim=1)
emb = emb + aug_emb if aug_emb is not None else emb
if self.time_embed_act is not None:
emb = self.time_embed_act(emb)
if (
self.encoder_hid_proj is not None
and self.config.encoder_hid_dim_type == "text_proj"
):
encoder_hidden_states = self.encoder_hid_proj(encoder_hidden_states)
elif (
self.encoder_hid_proj is not None
and self.config.encoder_hid_dim_type == "text_image_proj"
):
# Kadinsky 2.1 - style
if "image_embeds" not in added_cond_kwargs:
raise ValueError(
f"{self.__class__} has the config param `encoder_hid_dim_type` set to 'text_image_proj' which requires the keyword argument `image_embeds` to be passed in `added_conditions`"
)
image_embeds = added_cond_kwargs.get("image_embeds")
encoder_hidden_states = self.encoder_hid_proj(
encoder_hidden_states, image_embeds
)
elif (
self.encoder_hid_proj is not None
and self.config.encoder_hid_dim_type == "image_proj"
):
# Kandinsky 2.2 - style
if "image_embeds" not in added_cond_kwargs:
raise ValueError(
f"{self.__class__} has the config param `encoder_hid_dim_type` set to 'image_proj' which requires the keyword argument `image_embeds` to be passed in `added_conditions`"
)
image_embeds = added_cond_kwargs.get("image_embeds")
encoder_hidden_states = self.encoder_hid_proj(image_embeds)
elif (
self.encoder_hid_proj is not None
and self.config.encoder_hid_dim_type == "ip_image_proj"
):
if "image_embeds" not in added_cond_kwargs:
raise ValueError(
f"{self.__class__} has the config param `encoder_hid_dim_type` set to 'ip_image_proj' which requires the keyword argument `image_embeds` to be passed in `added_conditions`"
)
image_embeds = added_cond_kwargs.get("image_embeds")
image_embeds = self.encoder_hid_proj(image_embeds).to(
encoder_hidden_states.dtype
)
encoder_hidden_states = torch.cat(
[encoder_hidden_states, image_embeds], dim=1
)
# 2. pre-process
sample = self.conv_in(sample)
# # 2.5 GLIGEN position net
# if (
# cross_attention_kwargs is not None
# and cross_attention_kwargs.get("gligen", None) is not None
# ):
# cross_attention_kwargs = cross_attention_kwargs.copy()
# gligen_args = cross_attention_kwargs.pop("gligen")
# cross_attention_kwargs["gligen"] = {
# "objs": self.position_net(**gligen_args)
# }
# 3. down
lora_scale = (
cross_attention_kwargs.get("scale", 1.0)
if cross_attention_kwargs is not None
else 1.0
)
if USE_PEFT_BACKEND:
# weight the lora layers by setting `lora_scale` for each PEFT layer
scale_lora_layers(self, lora_scale)
is_controlnet = (
mid_block_additional_residual is not None
and down_block_additional_residuals is not None
)
# using new arg down_intrablock_additional_residuals for T2I-Adapters, to distinguish from controlnets
is_adapter = down_intrablock_additional_residuals is not None
# maintain backward compatibility for legacy usage, where
# T2I-Adapter and ControlNet both use down_block_additional_residuals arg
# but can only use one or the other
if (
not is_adapter
and mid_block_additional_residual is None
and down_block_additional_residuals is not None
):
deprecate(
"T2I should not use down_block_additional_residuals",
"1.3.0",
"Passing intrablock residual connections with `down_block_additional_residuals` is deprecated \
and will be removed in diffusers 1.3.0. `down_block_additional_residuals` should only be used \
for ControlNet. Please make sure use `down_intrablock_additional_residuals` instead. ",
standard_warn=False,
)
down_intrablock_additional_residuals = down_block_additional_residuals
is_adapter = True
down_block_res_samples = (sample,)
tot_referece_features = ()
for downsample_block in self.down_blocks:
if (
hasattr(downsample_block, "has_cross_attention")
and downsample_block.has_cross_attention
):
# For t2i-adapter CrossAttnDownBlock2D
additional_residuals = {}
if is_adapter and len(down_intrablock_additional_residuals) > 0:
additional_residuals[
"additional_residuals"
] = down_intrablock_additional_residuals.pop(0)
sample, res_samples = downsample_block(
hidden_states=sample,
temb=emb,
encoder_hidden_states=encoder_hidden_states,
attention_mask=attention_mask,
cross_attention_kwargs=cros
gitextract_lf1bmt9g/
├── LICENSE
├── README.md
├── checkpoints/
│ └── place_checkpoints_here.txt
├── environment.yaml
├── infer.py
├── inference/
│ └── depthlab_pipeline.py
├── output/
│ └── in-the-wild_example/
│ └── depth_npy/
│ └── RGB_pred.npy
├── scripts/
│ └── infer.sh
├── src/
│ └── models/
│ ├── attention.py
│ ├── mutual_self_attention.py
│ ├── projection.py
│ ├── transformer_2d.py
│ ├── unet_2d_blocks.py
│ ├── unet_2d_condition.py
│ └── unet_2d_condition_main.py
├── test_cases/
│ ├── know_depth.npy
│ └── mask.npy
└── utils/
├── image_util.py
└── seed_all.py
SYMBOL INDEX (101 symbols across 11 files)
FILE: infer.py
function load_and_process_mask (line 24) | def load_and_process_mask(mask_path):
FILE: inference/depthlab_pipeline.py
class DepthPipelineOutput (line 31) | class DepthPipelineOutput(BaseOutput):
class DepthLabPipeline (line 48) | class DepthLabPipeline(DiffusionPipeline):
method __init__ (line 53) | def __init__(self,
method __call__ (line 80) | def __call__(self,
method get_timesteps (line 252) | def get_timesteps(self, num_inference_steps, strength, device):
method single_infer (line 262) | def single_infer(self,
method encode_RGB (line 377) | def encode_RGB(self, rgb_in: torch.Tensor) -> torch.Tensor:
method decode_depth (line 399) | def decode_depth(self, depth_latent: torch.Tensor) -> torch.Tensor:
method encode_depth (line 423) | def encode_depth(self, depth_latent: torch.Tensor) -> torch.Tensor:
FILE: src/models/attention.py
class BasicTransformerBlock (line 12) | class BasicTransformerBlock(nn.Module):
method __init__ (line 47) | def __init__(
method set_chunk_feed_forward (line 173) | def set_chunk_feed_forward(self, chunk_size: Optional[int], dim: int =...
method forward (line 178) | def forward(
class TemporalBasicTransformerBlock (line 298) | class TemporalBasicTransformerBlock(nn.Module):
method __init__ (line 299) | def __init__(
method forward (line 381) | def forward(
FILE: src/models/mutual_self_attention.py
function torch_dfs (line 12) | def torch_dfs(model: torch.nn.Module):
class ReferenceAttentionControl (line 19) | class ReferenceAttentionControl:
method __init__ (line 20) | def __init__(
method register_reference_hooks (line 52) | def register_reference_hooks(
method update (line 286) | def update(self, writer, dtype=torch.float16):
method clear (line 325) | def clear(self):
FILE: src/models/projection.py
class My_proj (line 4) | class My_proj(nn.Module):
method __init__ (line 5) | def __init__(self, input=1024, dtype=torch.float32):
method forward (line 10) | def forward(self, x):
FILE: src/models/transformer_2d.py
class Transformer2DModelOutput (line 18) | class Transformer2DModelOutput(BaseOutput):
class Transformer2DModel (line 32) | class Transformer2DModel(ModelMixin, ConfigMixin):
method __init__ (line 63) | def __init__(
method _set_gradient_checkpointing (line 209) | def _set_gradient_checkpointing(self, module, value=False):
method forward (line 213) | def forward(
FILE: src/models/unet_2d_blocks.py
function get_down_block (line 20) | def get_down_block(
function get_up_block (line 102) | def get_up_block(
function get_mid_block (line 184) | def get_mid_block(
class AutoencoderTinyBlock (line 243) | class AutoencoderTinyBlock(nn.Module):
method __init__ (line 259) | def __init__(self, in_channels: int, out_channels: int, act_fn: str):
method forward (line 276) | def forward(self, x: torch.FloatTensor) -> torch.FloatTensor:
class UNetMidBlock2D (line 280) | class UNetMidBlock2D(nn.Module):
method __init__ (line 311) | def __init__(
method forward (line 401) | def forward(
class UNetMidBlock2DCrossAttn (line 413) | class UNetMidBlock2DCrossAttn(nn.Module):
method __init__ (line 414) | def __init__(
method forward (line 509) | def forward(
class CrossAttnDownBlock2D (line 567) | class CrossAttnDownBlock2D(nn.Module):
method __init__ (line 568) | def __init__(
method forward (line 663) | def forward(
class DownBlock2D (line 738) | class DownBlock2D(nn.Module):
method __init__ (line 739) | def __init__(
method forward (line 794) | def forward(
class CrossAttnUpBlock2D (line 836) | class CrossAttnUpBlock2D(nn.Module):
method __init__ (line 837) | def __init__(
method forward (line 929) | def forward(
class UpBlock2D (line 1019) | class UpBlock2D(nn.Module):
method __init__ (line 1020) | def __init__(
method forward (line 1071) | def forward(
FILE: src/models/unet_2d_condition.py
class UNet2DConditionOutput (line 51) | class UNet2DConditionOutput(BaseOutput):
class UNet2DConditionModel (line 64) | class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoade...
method __init__ (line 161) | def __init__(
method attn_processors (line 672) | def attn_processors(self) -> Dict[str, AttentionProcessor]:
method set_attn_processor (line 701) | def set_attn_processor(
method set_default_attn_processor (line 741) | def set_default_attn_processor(self):
method set_attention_slice (line 762) | def set_attention_slice(self, slice_size):
method _set_gradient_checkpointing (line 833) | def _set_gradient_checkpointing(self, module, value=False):
method enable_freeu (line 837) | def enable_freeu(self, s1, s2, b1, b2):
method disable_freeu (line 861) | def disable_freeu(self):
method forward (line 872) | def forward(
FILE: src/models/unet_2d_condition_main.py
class UNet2DConditionOutput (line 51) | class UNet2DConditionOutput(BaseOutput):
class UNet2DConditionModel_main (line 63) | class UNet2DConditionModel_main(ModelMixin, ConfigMixin, UNet2DCondition...
method __init__ (line 160) | def __init__(
method _check_config (line 476) | def _check_config(
method _set_time_proj (line 528) | def _set_time_proj(
method _set_encoder_hid_proj (line 556) | def _set_encoder_hid_proj(
method _set_class_embedding (line 596) | def _set_class_embedding(
method _set_add_embedding (line 633) | def _set_add_embedding(
method _set_pos_net_if_use_gligen (line 673) | def _set_pos_net_if_use_gligen(self, attention_type: str, cross_attent...
method attn_processors (line 687) | def attn_processors(self) -> Dict[str, AttentionProcessor]:
method set_attn_processor (line 710) | def set_attn_processor(self, processor: Union[AttentionProcessor, Dict...
method set_default_attn_processor (line 744) | def set_default_attn_processor(self):
method set_attention_slice (line 759) | def set_attention_slice(self, slice_size: Union[str, int, List[int]] =...
method _set_gradient_checkpointing (line 824) | def _set_gradient_checkpointing(self, module, value=False):
method enable_freeu (line 828) | def enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
method disable_freeu (line 852) | def disable_freeu(self):
method fuse_qkv_projections (line 860) | def fuse_qkv_projections(self):
method unfuse_qkv_projections (line 883) | def unfuse_qkv_projections(self):
method unload_lora (line 896) | def unload_lora(self):
method get_time_embed (line 907) | def get_time_embed(
method get_class_embed (line 933) | def get_class_embed(self, sample: torch.Tensor, class_labels: Optional...
method get_aug_embed (line 949) | def get_aug_embed(
method process_encoder_hidden_states (line 1001) | def process_encoder_hidden_states(
method forward (line 1033) | def forward(
FILE: utils/image_util.py
class DepthFileNameMode (line 13) | class DepthFileNameMode(Enum):
function get_filled_depth (line 21) | def get_filled_depth(depth, mask, method):
function resize_max_res (line 30) | def resize_max_res(img: Image.Image, max_edge_resolution: int, resample=...
function resize_max_res_cv2 (line 54) | def resize_max_res_cv2(img: np.ndarray, max_edge_resolution: int, interp...
function resize_max_res_tensor (line 78) | def resize_max_res_tensor(input_tensor,recom_resolution=768):
function colorize_depth_maps (line 106) | def colorize_depth_maps(
function chw2hwc (line 146) | def chw2hwc(chw):
function Disparity_Normalization_mask_scale (line 154) | def Disparity_Normalization_mask_scale(disparity, min_value, max_value, ...
function get_pred_name (line 160) | def get_pred_name(rgb_basename, name_mode, suffix=".png"):
function get_filled_for_latents (line 176) | def get_filled_for_latents(mask, sparse_depth):
FILE: utils/seed_all.py
function seed_all (line 26) | def seed_all(seed: int = 0):
Condensed preview — 19 files, each showing path, character count, and a content snippet. Download the .json file or copy for the full structured content (286K chars).
[
{
"path": "LICENSE",
"chars": 11357,
"preview": " Apache License\n Version 2.0, January 2004\n "
},
{
"path": "README.md",
"chars": 5263,
"preview": "# DepthLab: From Partial to Complete\n\nThis repository represents the official implementation of the paper titled \"DepthL"
},
{
"path": "checkpoints/place_checkpoints_here.txt",
"chars": 0,
"preview": ""
},
{
"path": "environment.yaml",
"chars": 914,
"preview": "name: DepthLab\nchannels:\n - defaults\ndependencies:\n - pip=24.2=py39h06a4308_0\n - python=3.9.19=h955ad1f_1\n - pip:\n "
},
{
"path": "infer.py",
"chars": 9487,
"preview": "import argparse\nimport logging\nimport os\nimport random\nimport numpy as np\nimport torch\nfrom PIL import Image\nfrom tqdm.a"
},
{
"path": "inference/depthlab_pipeline.py",
"chars": 16447,
"preview": "import os, sys\nparent_dir = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))\nif parent_dir not in sys.path:\n"
},
{
"path": "scripts/infer.sh",
"chars": 963,
"preview": "#!/usr/bin/env bash\nset -e\nset -x\npretrained_model_name_or_path='./checkpoints/marigold-depth-v1-0'\nimage_encoder_path='"
},
{
"path": "src/models/attention.py",
"chars": 17711,
"preview": "# Adapted from https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention.py\n\nfrom typing import "
},
{
"path": "src/models/mutual_self_attention.py",
"chars": 13773,
"preview": "# Adapted from https://github.com/magic-research/magic-animate/blob/main/magicanimate/models/mutual_self_attention.py\nfr"
},
{
"path": "src/models/projection.py",
"chars": 319,
"preview": "import torch\nimport torch.nn as nn\n\nclass My_proj(nn.Module):\n def __init__(self, input=1024, dtype=torch.float32):\n "
},
{
"path": "src/models/transformer_2d.py",
"chars": 18943,
"preview": "# Adapted from https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/transformer_2d.py\nfrom dataclasse"
},
{
"path": "src/models/unet_2d_blocks.py",
"chars": 43626,
"preview": "# Adapted from https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/unet_2d_blocks.py\nfrom typing imp"
},
{
"path": "src/models/unet_2d_condition.py",
"chars": 63613,
"preview": "# Adapted from https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/unet_2d_condition.py\nfrom datacla"
},
{
"path": "src/models/unet_2d_condition_main.py",
"chars": 66629,
"preview": "# Adapted from https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/unet_2d_condition.py\nfrom datacla"
},
{
"path": "utils/image_util.py",
"chars": 6775,
"preview": "import matplotlib\nimport numpy as np\nimport torch\nfrom PIL import Image\nimport cv2\nfrom torchvision.transforms.functiona"
},
{
"path": "utils/seed_all.py",
"chars": 1250,
"preview": "# Copyright 2023 Bingxin Ke, ETH Zurich. All rights reserved.\n#\n# Licensed under the Apache License, Version 2.0 (the \"L"
}
]
// ... and 3 more files (download for full content)
About this extraction
This page contains the full source code of the ant-research/DepthLab GitHub repository, extracted and formatted as plain text for AI agents and large language models (LLMs). The extraction includes 19 files (270.6 KB), approximately 59.1k tokens, and a symbol index with 101 extracted functions, classes, methods, constants, and types. Use this with OpenClaw, Claude, ChatGPT, Cursor, Windsurf, or any other AI tool that accepts text input. You can copy the full output to your clipboard or download it as a .txt file.
Extracted by GitExtract — free GitHub repo to text converter for AI. Built by Nikandr Surkov.