SYMBOL INDEX (4565 symbols across 509 files)

FILE: llava-train_videochat/llava/conversation.py
  class SeparatorStyle (line 11) | class SeparatorStyle(Enum):
  class Conversation (line 26) | class Conversation:
    method get_prompt (line 47) | def get_prompt(self):
    method append_message (line 178) | def append_message(self, role, message):
    method process_image (line 181) | def process_image(self, image, image_process_mode, return_pil=False, i...
    method get_images (line 227) | def get_images(self, return_pil=False, return_path=False):
    method is_image_file (line 242) | def is_image_file(self, filename):
    method is_video_file (line 246) | def is_video_file(self, filename):
    method to_gradio_chatbot (line 250) | def to_gradio_chatbot(self):
    method copy (line 288) | def copy(self):
    method dict (line 291) | def dict(self):

FILE: llava-train_videochat/llava/dist_utils.py
  function get_rank (line 16) | def get_rank() -> int:
  function get_world_size (line 23) | def get_world_size() -> int:
  function setup_for_distributed (line 30) | def setup_for_distributed(is_master):
  function init_distributed_mode (line 44) | def init_distributed_mode(use_dynamic_port: bool = True):

FILE: llava-train_videochat/llava/mm_utils.py
  function resize_and_center_crop (line 12) | def resize_and_center_crop(image, shortest_edge_length):
  function auto_pad_images (line 33) | def auto_pad_images(image, grid_params):
  function extract_patches (line 63) | def extract_patches(image, patch_size, overlap_ratio):
  function process_highres_image_crop_split (line 87) | def process_highres_image_crop_split(image, data_args, processor=None):
  function process_highres_image (line 98) | def process_highres_image(image, processor, grid_pinpoints):
  function select_best_resolution (line 119) | def select_best_resolution(original_size, possible_resolutions, max_reso...
  function resize_and_pad_image (line 157) | def resize_and_pad_image(image, target_resolution):
  function divide_to_patches (line 196) | def divide_to_patches(image, patch_size):
  function get_anyres_image_grid_shape (line 218) | def get_anyres_image_grid_shape(image_size, grid_pinpoints, patch_size, ...
  function process_anyres_image (line 251) | def process_anyres_image(image, processor, grid_pinpoints):
  function process_anyres_image_nopad (line 306) | def process_anyres_image_nopad(image, processor, grid_pinpoints):
  function process_anyres_video_nopad (line 365) | def process_anyres_video_nopad(video, processor, grid_pinpoints, max_res...
  function load_image_from_base64 (line 410) | def load_image_from_base64(image):
  function expand2square (line 414) | def expand2square(pil_img, background_color):
  function process_images (line 428) | def process_images(images, image_processor, model_cfg):
  function tokenizer_image_token (line 460) | def tokenizer_image_token(prompt, tokenizer, image_token_index=IMAGE_TOK...
  function get_model_name_from_path (line 482) | def get_model_name_from_path(model_path):
  class KeywordsStoppingCriteria (line 491) | class KeywordsStoppingCriteria(StoppingCriteria):
    method __init__ (line 492) | def __init__(self, keywords, tokenizer, input_ids):
    method __call__ (line 503) | def __call__(self, output_ids: torch.LongTensor, scores: torch.FloatTe...

FILE: llava-train_videochat/llava/model/apply_delta.py
  function apply_delta (line 14) | def apply_delta(base_model_path, target_model_path, delta_path):

FILE: llava-train_videochat/llava/model/builder.py
  function load_pretrained_model (line 27) | def load_pretrained_model(model_path, model_base, model_name, load_8bit=...

FILE: llava-train_videochat/llava/model/consolidate.py
  function consolidate_ckpt (line 14) | def consolidate_ckpt(src_path, dst_path):

FILE: llava-train_videochat/llava/model/language_model/llava_qwen.py
  class LlavaQwenConfig (line 35) | class LlavaQwenConfig(Qwen2Config):
  class LlavaQwenModel (line 39) | class LlavaQwenModel(LlavaMetaModel, Qwen2Model):
    method __init__ (line 42) | def __init__(self, config: Qwen2Config):
  class LlavaQwenForCausalLM (line 46) | class LlavaQwenForCausalLM(Qwen2ForCausalLM, LlavaMetaForCausalLM):
    method __init__ (line 49) | def __init__(self, config):
    method get_model (line 60) | def get_model(self):
    method forward (line 63) | def forward(
    method generate (line 119) | def generate(
    method prepare_inputs_for_generation (line 139) | def prepare_inputs_for_generation(self, input_ids, past_key_values=Non...

FILE: llava-train_videochat/llava/model/language_model/llava_qwen_flash.py
  class LlavaQwenConfig_Flash (line 35) | class LlavaQwenConfig_Flash(Qwen2Config):
  class LlavaQwenModel_Flash (line 39) | class LlavaQwenModel_Flash(LlavaMetaModel, Qwen2Model_Flash):
    method __init__ (line 42) | def __init__(self, config: Qwen2Config):
  class LlavaQwenForCausalLM_Flash (line 46) | class LlavaQwenForCausalLM_Flash(Qwen2ForCausalLM_Flash, LlavaMetaForCau...
    method __init__ (line 49) | def __init__(self, config):
    method get_model (line 60) | def get_model(self):
    method forward (line 63) | def forward(
    method generate (line 120) | def generate(
    method prepare_inputs_for_generation (line 143) | def prepare_inputs_for_generation(self, input_ids, past_key_values=Non...

FILE: llava-train_videochat/llava/model/language_model/modeling_qwen2_flash.py
  function _get_unpad_data (line 70) | def _get_unpad_data(attention_mask):
  class Qwen2RMSNorm (line 83) | class Qwen2RMSNorm(nn.Module):
    method __init__ (line 84) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 92) | def forward(self, hidden_states):
  class Qwen2RotaryEmbedding (line 101) | class Qwen2RotaryEmbedding(nn.Module):
    method __init__ (line 102) | def __init__(self, dim, max_position_embeddings=2048, base=10000, devi...
    method _set_cos_sin_cache (line 116) | def _set_cos_sin_cache(self, seq_len, device, dtype):
    method forward (line 126) | def forward(self, x, seq_len=None):
  function rotate_half (line 138) | def rotate_half(x):
  function apply_rotary_pos_emb (line 146) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
  class Qwen2MLP (line 175) | class Qwen2MLP(nn.Module):
    method __init__ (line 176) | def __init__(self, config):
    method forward (line 186) | def forward(self, x):
  function repeat_kv (line 191) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class Qwen2Attention (line 203) | class Qwen2Attention(nn.Module):
    method __init__ (line 209) | def __init__(self, config: Qwen2Config, layer_idx: Optional[int] = None):
    method forward (line 246) | def forward(
  class Qwen2FlashAttention2 (line 328) | class Qwen2FlashAttention2(Qwen2Attention):
    method __init__ (line 338) | def __init__(self, *args, **kwargs):
    method forward (line 346) | def forward(
    method _flash_attention_forward (line 482) | def _flash_attention_forward(
    method _upad_input (line 587) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class Qwen2SdpaAttention (line 631) | class Qwen2SdpaAttention(Qwen2Attention):
    method forward (line 639) | def forward(
  class Qwen2DecoderLayer (line 725) | class Qwen2DecoderLayer(nn.Module):
    method __init__ (line 726) | def __init__(self, config: Qwen2Config, layer_idx: int):
    method forward (line 741) | def forward(
  class Qwen2PreTrainedModel (line 823) | class Qwen2PreTrainedModel(PreTrainedModel):
    method _init_weights (line 833) | def _init_weights(self, module):
  class Qwen2Model_Flash (line 919) | class Qwen2Model_Flash(Qwen2PreTrainedModel):
    method __init__ (line 927) | def __init__(self, config: Qwen2Config):
    method get_input_embeddings (line 944) | def get_input_embeddings(self):
    method set_input_embeddings (line 947) | def set_input_embeddings(self, value):
    method forward (line 951) | def forward(
    method flash_rank_drop (line 1159) | def flash_rank_drop(
  class Qwen2ForCausalLM_Flash (line 1382) | class Qwen2ForCausalLM_Flash(Qwen2PreTrainedModel):
    method __init__ (line 1385) | def __init__(self, config):
    method get_input_embeddings (line 1394) | def get_input_embeddings(self):
    method set_input_embeddings (line 1397) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1400) | def get_output_embeddings(self):
    method set_output_embeddings (line 1403) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1406) | def set_decoder(self, decoder):
    method get_decoder (line 1409) | def get_decoder(self):
    method forward (line 1414) | def forward(
    method prepare_inputs_for_generation (line 1502) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1560) | def _reorder_cache(past_key_values, beam_idx):

FILE: llava-train_videochat/llava/model/llava_arch.py
  class LlavaMetaModel (line 35) | class LlavaMetaModel:
    method __init__ (line 37) | def __init__(self, config):
    method get_vision_tower (line 50) | def get_vision_tower(self):
    method initialize_vision_modules (line 56) | def initialize_vision_modules(self, model_args, fsdp=None):
  function unpad_image (line 116) | def unpad_image(tensor, original_size, is_frame=False):
  class LlavaMetaForCausalLM (line 160) | class LlavaMetaForCausalLM(ABC):
    method get_model (line 163) | def get_model(self):
    method get_vision_tower (line 166) | def get_vision_tower(self):
    method get_4dPool (line 169) | def get_4dPool(self, image_feature):
    method get_2dPool (line 194) | def get_2dPool(self, image_feature):
    method encode_image (line 220) | def encode_image(self, images_list):
    method encode_image_video (line 230) | def encode_image_video(self, images_list, video_idx_in_batch):
    method encode_video (line 255) | def encode_video(self, images_list, video_idx_in_batch):
    method encode_video_image (line 309) | def encode_video_image(self, images_list, video_idx_in_batch):
    method add_token_per_frame (line 364) | def add_token_per_frame(self, image_feature):
    method add_different_token_per_frame (line 373) | def add_different_token_per_frame(self, image_feature):
    method prepare_inputs_labels_for_multimodal (line 377) | def prepare_inputs_labels_for_multimodal(self, input_ids, position_ids...
    method initialize_vision_tokenizer (line 801) | def initialize_vision_tokenizer(self, model_args, tokenizer):

FILE: llava-train_videochat/llava/model/make_delta.py
  function make_delta (line 14) | def make_delta(base_model_path, target_model_path, delta_path, hub_repo_...

FILE: llava-train_videochat/llava/model/multimodal_encoder/builder.py
  function build_vision_tower (line 11) | def build_vision_tower(vision_tower_cfg, **kwargs):

FILE: llava-train_videochat/llava/model/multimodal_encoder/clip_encoder.py
  class CLIPVisionTower (line 12) | class CLIPVisionTower(nn.Module):
    method __init__ (line 13) | def __init__(self, vision_tower, args, delay_load=False):
    method load_model (line 35) | def load_model(self, device_map=None):
    method feature_select (line 46) | def feature_select(self, image_forward_outs):
    method forward (line 68) | def forward(self, images):
    method dummy_feature (line 82) | def dummy_feature(self):
    method dtype (line 86) | def dtype(self):
    method device (line 90) | def device(self):
    method config (line 94) | def config(self):
    method hidden_size (line 101) | def hidden_size(self):
    method num_patches_per_side (line 110) | def num_patches_per_side(self):
    method num_patches (line 114) | def num_patches(self):
    method image_size (line 121) | def image_size(self):
  class CLIPVisionTowerS2 (line 125) | class CLIPVisionTowerS2(CLIPVisionTower):
    method __init__ (line 126) | def __init__(self, vision_tower, args, delay_load=False):
    method load_model (line 141) | def load_model(self, device_map=None):
    method forward_feature (line 155) | def forward_feature(self, images):
    method forward (line 160) | def forward(self, images):
    method hidden_size (line 172) | def hidden_size(self):

FILE: llava-train_videochat/llava/model/multimodal_encoder/internvideo2/flash_attention_class.py
  class FlashAttention (line 10) | class FlashAttention(nn.Module):
    method __init__ (line 21) | def __init__(self, softmax_scale=None, attention_dropout=0.0, device=N...
    method forward (line 26) | def forward(self, qkv, key_padding_mask=None, causal=False, cu_seqlens...

FILE: llava-train_videochat/llava/model/multimodal_encoder/internvideo2/pos_embed.py
  function get_3d_sincos_pos_embed (line 12) | def get_3d_sincos_pos_embed(embed_dim, grid_size, t_size, cls_token=False):
  function get_2d_sincos_pos_embed (line 66) | def get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):
  function get_1d_sincos_pos_embed (line 86) | def get_1d_sincos_pos_embed(embed_dim, t_size, cls_token=False):
  function get_2d_sincos_pos_embed_from_grid (line 101) | def get_2d_sincos_pos_embed_from_grid(embed_dim, grid):
  function get_1d_sincos_pos_embed_from_grid (line 116) | def get_1d_sincos_pos_embed_from_grid(embed_dim, pos):
  function interpolate_pos_embed_internvideo2 (line 137) | def interpolate_pos_embed_internvideo2(checkpoint_model, model, orig_t_s...
  function interpolate_pos_embed_internvideo2_new (line 227) | def interpolate_pos_embed_internvideo2_new(checkpoint_model, model, orig...

FILE: llava-train_videochat/llava/model/multimodal_encoder/internvideo2/vit_scale_clean.py
  class CrossAttention (line 20) | class CrossAttention(nn.Module):
    method __init__ (line 21) | def __init__(
    method forward (line 52) | def forward(self, x, k=None, v=None):
  class AttentiveBlock (line 85) | class AttentiveBlock(nn.Module):
    method __init__ (line 87) | def __init__(self, dim, num_heads, qkv_bias=False, qk_scale=None, drop...
    method forward (line 100) | def forward(self, x_q, x_kv, pos_q, pos_k, bool_masked_pos, rel_pos_bi...
  class AttentionPoolingBlock (line 109) | class AttentionPoolingBlock(AttentiveBlock):
    method forward (line 111) | def forward(self, x):
  class RMSNorm (line 119) | class RMSNorm(nn.Module):
    method __init__ (line 120) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 125) | def forward(self, hidden_states):
  class LayerScale (line 133) | class LayerScale(nn.Module):
    method __init__ (line 134) | def __init__(self, dim, init_values=1e-5, inplace=False, force_fp32=Fa...
    method forward (line 141) | def forward(self, x):
  class Attention (line 151) | class Attention(nn.Module):
    method __init__ (line 152) | def __init__(self, dim, num_heads=8, qkv_bias=False, attn_drop=0., pro...
    method _naive_attn (line 175) | def _naive_attn(self, x):
    method _flash_attn (line 197) | def _flash_attn(self, x, key_padding_mask=None, need_weights=False):
    method forward (line 219) | def forward(self, x):
  class Mlp (line 224) | class Mlp(nn.Module):
    method __init__ (line 228) | def __init__(self, in_features, hidden_features=None, out_features=Non...
    method forward (line 242) | def forward(self, x):
  class Block (line 251) | class Block(nn.Module):
    method __init__ (line 253) | def __init__(
    method forward (line 284) | def forward(self, x, residual=None):
  class PatchEmbed (line 306) | class PatchEmbed(nn.Module):
    method __init__ (line 310) | def __init__(
    method forward (line 334) | def forward(self, x):
  class PretrainVisionTransformer_clean (line 343) | class PretrainVisionTransformer_clean(nn.Module):
    method __init__ (line 344) | def __init__(
    method init_pos_embed (line 453) | def init_pos_embed(self):
    method _init_weights (line 476) | def _init_weights(self, m):
    method fix_init_weight (line 485) | def fix_init_weight(self):
    method dtype (line 494) | def dtype(self):
    method get_num_layers (line 497) | def get_num_layers(self):
    method no_weight_decay (line 501) | def no_weight_decay(self):
    method forward (line 512) | def forward(self, x, mask=None, use_image=False):
  function pretrain_internvideo2_giant_patch14_224_clean (line 572) | def pretrain_internvideo2_giant_patch14_224_clean(config):
  function pretrain_internvideo2_6b_patch14_224_clean (line 608) | def pretrain_internvideo2_6b_patch14_224_clean(config):

FILE: llava-train_videochat/llava/model/multimodal_encoder/internvideo2_encoder.py
  class InternVideo2ImageProcessor (line 30) | class InternVideo2ImageProcessor:
    method __init__ (line 31) | def __init__(self, image_mean=(0.485, 0.456, 0.406), image_std=(0.229,...
    method preprocess (line 43) | def preprocess(self, images, return_tensors, target_size=None):
  class InternVideo2VisionConfig (line 69) | class InternVideo2VisionConfig:
    method __init__ (line 72) | def __init__(
  function build_vit (line 101) | def build_vit(config, pt_type='origin'):
  class InternVideo2VisionTower (line 145) | class InternVideo2VisionTower(nn.Module):
    method __init__ (line 146) | def __init__(self, vision_tower, vision_tower_cfg, delay_load=False, p...
    method load_model (line 172) | def load_model(self, device_map=None):
    method forward (line 182) | def forward(self, images):
    method dummy_feature (line 195) | def dummy_feature(self):
    method dtype (line 199) | def dtype(self):
    method device (line 204) | def device(self):
    method hidden_size (line 209) | def hidden_size(self):
    method num_patches (line 213) | def num_patches(self):
    method num_patches_per_side (line 217) | def num_patches_per_side(self):
    method image_size (line 222) | def image_size(self):

FILE: llava-train_videochat/llava/model/multimodal_encoder/siglip_encoder.py
  class SigLipImageProcessor (line 34) | class SigLipImageProcessor:
    method __init__ (line 35) | def __init__(self, image_mean=(0.5, 0.5, 0.5), image_std=(0.5, 0.5, 0....
    method preprocess (line 47) | def preprocess(self, images, return_tensors):
  class SigLipVisionConfig (line 70) | class SigLipVisionConfig(PretrainedConfig):
    method __init__ (line 73) | def __init__(
    method from_pretrained (line 103) | def from_pretrained(cls, pretrained_model_name_or_path: Union[str, os....
  class SigLipVisionModelOutput (line 120) | class SigLipVisionModelOutput(ModelOutput):
  class SigLipVisionEmbeddings (line 148) | class SigLipVisionEmbeddings(nn.Module):
    method __init__ (line 149) | def __init__(self, config: SigLipVisionConfig):
    method forward (line 169) | def forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
  class SigLipAttention (line 177) | class SigLipAttention(nn.Module):
    method __init__ (line 181) | def __init__(self, config):
    method forward (line 197) | def forward(
  class SigLipMLP (line 243) | class SigLipMLP(nn.Module):
    method __init__ (line 244) | def __init__(self, config):
    method forward (line 251) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
  class SigLipEncoderLayer (line 259) | class SigLipEncoderLayer(nn.Module):
    method __init__ (line 260) | def __init__(self, config: SigLipVisionConfig):
    method forward (line 269) | def forward(
  class SigLipPreTrainedModel (line 308) | class SigLipPreTrainedModel(PreTrainedModel):
    method _init_weights (line 318) | def _init_weights(self, module):
  class SigLipEncoder (line 324) | class SigLipEncoder(nn.Module):
    method __init__ (line 333) | def __init__(self, config: SigLipVisionConfig):
    method forward (line 340) | def forward(
  class SigLipVisionTransformer (line 408) | class SigLipVisionTransformer(nn.Module):
    method __init__ (line 409) | def __init__(self, config: SigLipVisionConfig):
    method forward (line 419) | def forward(
  class SigLipMultiheadAttentionPoolingHead (line 459) | class SigLipMultiheadAttentionPoolingHead(nn.Module):
    method __init__ (line 462) | def __init__(self, config: SigLipVisionConfig):
    method forward (line 470) | def forward(self, hidden_state):
  class SigLipVisionModel (line 483) | class SigLipVisionModel(SigLipPreTrainedModel):
    method __init__ (line 488) | def __init__(self, config: SigLipVisionConfig):
    method get_input_embeddings (line 496) | def get_input_embeddings(self) -> nn.Module:
    method forward (line 499) | def forward(
  class SigLipVisionTower (line 538) | class SigLipVisionTower(nn.Module):
    method __init__ (line 539) | def __init__(self, vision_tower, vision_tower_cfg, delay_load=False):
    method load_model (line 563) | def load_model(self, device_map=None):
    method forward (line 576) | def forward(self, images):
    method dummy_feature (line 592) | def dummy_feature(self):
    method dtype (line 596) | def dtype(self):
    method device (line 601) | def device(self):
    method hidden_size (line 606) | def hidden_size(self):
    method num_patches (line 610) | def num_patches(self):
    method num_patches_per_side (line 614) | def num_patches_per_side(self):
    method image_size (line 619) | def image_size(self):

FILE: llava-train_videochat/llava/model/multimodal_encoder/umt/vit.py
  class DropPath (line 19) | class DropPath(nn.Module):
    method __init__ (line 22) | def __init__(self, drop_prob=None):
    method forward (line 26) | def forward(self, x):
    method extra_repr (line 29) | def extra_repr(self) -> str:
  class Mlp (line 33) | class Mlp(nn.Module):
    method __init__ (line 34) | def __init__(self, in_features, hidden_features=None, out_features=Non...
    method forward (line 43) | def forward(self, x):
  class Attention (line 51) | class Attention(nn.Module):
    method __init__ (line 52) | def __init__(
    method forward (line 85) | def forward(self, x):
  class Block (line 117) | class Block(nn.Module):
    method __init__ (line 118) | def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_sc...
    method forward (line 138) | def forward(self, x):
  class PatchEmbed (line 148) | class PatchEmbed(nn.Module):
    method __init__ (line 151) | def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=...
    method forward (line 167) | def forward(self, x, **kwargs):
  function get_sinusoid_encoding_table (line 177) | def get_sinusoid_encoding_table(n_position, d_hid, ckpt_num_frame=-1, cu...
  function get_sinusoid_encoding_table2 (line 211) | def get_sinusoid_encoding_table2(n_position=784, d_hid=1024, cur_frame=8...
  class PretrainVisionTransformerEncoder (line 258) | class PretrainVisionTransformerEncoder(nn.Module):
    method __init__ (line 261) | def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=...
    method no_weight_decay (line 312) | def no_weight_decay(self):
    method forward_features (line 315) | def forward_features(self, x, use_image=False):
    method forward (line 336) | def forward(self, x, use_image=False):
  class PretrainVisionTransformer (line 341) | class PretrainVisionTransformer(nn.Module):
    method __init__ (line 344) | def __init__(self,
    method _init_weights (line 400) | def _init_weights(self, m):
    method no_weight_decay (line 410) | def no_weight_decay(self):
    method forward (line 413) | def forward(self, x, use_image=False):

FILE: llava-train_videochat/llava/model/multimodal_encoder/umt_encoder.py
  class UMTImageProcessor (line 30) | class UMTImageProcessor:
    method __init__ (line 31) | def __init__(self, image_mean=(0.485, 0.456, 0.406), image_std=(0.229,...
    method preprocess (line 43) | def preprocess(self, images, return_tensors, target_size=None):
  class UMTVisionConfig (line 69) | class UMTVisionConfig:
    method __init__ (line 72) | def __init__(
  function build_vit (line 95) | def build_vit(config, pt_type='origin'):
  class UMTVisionTower (line 134) | class UMTVisionTower(nn.Module):
    method __init__ (line 135) | def __init__(self, vision_tower, vision_tower_cfg, delay_load=False, p...
    method load_model (line 160) | def load_model(self, device_map=None):
    method forward (line 170) | def forward(self, images):
    method dummy_feature (line 185) | def dummy_feature(self):
    method dtype (line 189) | def dtype(self):
    method device (line 194) | def device(self):
    method hidden_size (line 199) | def hidden_size(self):
    method num_patches (line 203) | def num_patches(self):
    method num_patches_per_side (line 207) | def num_patches_per_side(self):
    method image_size (line 212) | def image_size(self):

FILE: llava-train_videochat/llava/model/multimodal_projector/builder.py
  class IdentityMap (line 7) | class IdentityMap(nn.Module):
    method __init__ (line 8) | def __init__(self):
    method forward (line 11) | def forward(self, x, *args, **kwargs):
    method config (line 15) | def config(self):
  class SimpleResBlock (line 19) | class SimpleResBlock(nn.Module):
    method __init__ (line 20) | def __init__(self, channels):
    method forward (line 25) | def forward(self, x):
  function build_vision_projector (line 30) | def build_vision_projector(config, delay_load=False, **kwargs):

FILE: llava-train_videochat/llava/model/multimodal_projector/tome16_mlp_hd64.py
  function bipartite_soft_matching (line 16) | def bipartite_soft_matching(
  function merge_wavg (line 72) | def merge_wavg(
  class ToMe16_mlp_hd64 (line 91) | class ToMe16_mlp_hd64(nn.Module):
    method __init__ (line 92) | def __init__(self, config, vision_cfg):
    method merge_tokens (line 107) | def merge_tokens(self, x, target_num_token):
    method forward (line 143) | def forward(self, x, compress=False, local_num_frames=-1):
    method config (line 169) | def config(self):

FILE: llava-train_videochat/llava/model/utils.py
  function auto_upgrade (line 4) | def auto_upgrade(config):

FILE: llava-train_videochat/llava/serialize_utils.py
  function get_world_size (line 25) | def get_world_size() -> int:
  function get_rank (line 33) | def get_rank() -> int:
  function get_rank (line 40) | def get_rank() -> int:
  function get_local_rank (line 47) | def get_local_rank() -> int:
  function get_local_size (line 62) | def get_local_size() -> int:
  function _get_global_gloo_group (line 67) | def _get_global_gloo_group():
  function all_gather (line 78) | def all_gather(data, group=None):
  class NumpySerializedList (line 105) | class NumpySerializedList:
    method __init__ (line 106) | def __init__(self, lst: list):
    method __len__ (line 122) | def __len__(self):
    method __getitem__ (line 125) | def __getitem__(self, idx):
  class TorchSerializedList (line 132) | class TorchSerializedList(NumpySerializedList):
    method __init__ (line 133) | def __init__(self, lst: list):
    method __getitem__ (line 138) | def __getitem__(self, idx):
  function local_scatter (line 145) | def local_scatter(array: Optional[List[Any]]):
  class TorchShmSerializedList (line 167) | class TorchShmSerializedList(TorchSerializedList):
    method __init__ (line 168) | def __init__(self, lst: list):
  function local_broadcast_process_authkey (line 194) | def local_broadcast_process_authkey():

FILE: llava-train_videochat/llava/train/llava_trainer.py
  function maybe_zero_3 (line 27) | def maybe_zero_3(param, ignore_status=False, name=None):
  function get_mm_adapter_state_maybe_zero_3 (line 42) | def get_mm_adapter_state_maybe_zero_3(named_params, keys_to_match):
  function split_to_even_chunks (line 48) | def split_to_even_chunks(indices, lengths, num_chunks):
  function get_variable_length_grouped_indices (line 70) | def get_variable_length_grouped_indices(lengths, batch_size, world_size,...
  function get_modality_length_grouped_indices (line 86) | def get_modality_length_grouped_indices(lengths, batch_size, world_size,...
  function get_length_grouped_indices (line 126) | def get_length_grouped_indices(lengths, batch_size, world_size, generato...
  function get_length_grouped_indices_auto_single (line 149) | def get_length_grouped_indices_auto_single(lengths, batch_size, world_si...
  function get_modality_length_grouped_indices_auto (line 164) | def get_modality_length_grouped_indices_auto(lengths, batch_size, world_...
  class LengthGroupedSampler (line 193) | class LengthGroupedSampler(Sampler):
    method __init__ (line 199) | def __init__(
    method __len__ (line 220) | def __len__(self):
    method __iter__ (line 223) | def __iter__(self):
  class LLaVATrainer (line 237) | class LLaVATrainer(Trainer):
    method create_accelerator_and_postprocess (line 239) | def create_accelerator_and_postprocess(self):
    method _get_train_sampler (line 270) | def _get_train_sampler(self) -> Optional[torch.utils.data.Sampler]:
    method get_train_dataloader (line 316) | def get_train_dataloader(self) -> DataLoader:
    method create_optimizer (line 353) | def create_optimizer(self):
    method _save_checkpoint (line 432) | def _save_checkpoint(self, model, trial, metrics=None):
    method _save (line 456) | def _save(self, output_dir: Optional[str] = None, state_dict=None):

FILE: llava-train_videochat/llava/train/llava_trainer_eval.py
  class LLaVAEvalTrainer (line 7) | class LLaVAEvalTrainer(LLaVATrainer):
    method evaluate (line 8) | def evaluate(self, evaluate_args):

FILE: llava-train_videochat/llava/train/train.py
  class ModelArguments (line 64) | class ModelArguments:
  class DataArguments (line 135) | class DataArguments:
  class TrainingArguments (line 159) | class TrainingArguments(transformers.TrainingArguments):
  function maybe_zero_3 (line 208) | def maybe_zero_3(param, ignore_status=False, name=None):
  function get_peft_state_maybe_zero_3 (line 224) | def get_peft_state_maybe_zero_3(named_params, bias):
  function get_peft_state_non_lora_maybe_zero_3 (line 249) | def get_peft_state_non_lora_maybe_zero_3(named_params, require_grad_only...
  function get_mm_adapter_state_maybe_zero_3 (line 257) | def get_mm_adapter_state_maybe_zero_3(named_params, keys_to_match):
  function find_all_linear_names (line 263) | def find_all_linear_names(model):
  function safe_save_model_for_hf_trainer (line 279) | def safe_save_model_for_hf_trainer(trainer: transformers.Trainer, output...
  function smart_tokenizer_and_embedding_resize (line 323) | def smart_tokenizer_and_embedding_resize(
  function _tokenize_fn (line 346) | def _tokenize_fn(strings: Sequence[str], tokenizer: transformers.PreTrai...
  function _mask_targets (line 368) | def _mask_targets(target, tokenized_lens, speakers):
  function _add_speaker_and_signal (line 379) | def _add_speaker_and_signal(header, source, get_conversation=True):
  function preprocess_multimodal (line 399) | def preprocess_multimodal(sources: Sequence[str], data_args: DataArgumen...
  function preprocess_llama_2 (line 431) | def preprocess_llama_2(sources, tokenizer: transformers.PreTrainedTokeni...
  function preprocess_gemma (line 506) | def preprocess_gemma(sources: List[List[Dict[str, str]]], tokenizer: tra...
  function preprocess_qwen (line 585) | def preprocess_qwen(sources, tokenizer: transformers.PreTrainedTokenizer...
  function preprocess_internlm2 (line 665) | def preprocess_internlm2(sources, tokenizer: transformers.PreTrainedToke...
  function preprocess_llama3 (line 742) | def preprocess_llama3(
  function preprocess_v1 (line 828) | def preprocess_v1(sources, tokenizer: transformers.PreTrainedTokenizer, ...
  function preprocess_mpt (line 907) | def preprocess_mpt(sources, tokenizer: transformers.PreTrainedTokenizer,...
  function preprocess_plain (line 988) | def preprocess_plain(
  function preprocess (line 1010) | def preprocess(sources: Sequence[str], tokenizer: transformers.PreTraine...
  class LazySupervisedDataset (line 1065) | class LazySupervisedDataset(Dataset):
    method __init__ (line 1066) | def __init__(self, data_path: str, tokenizer: transformers.PreTrainedT...
    method __len__ (line 1257) | def __len__(self):
    method lengths (line 1261) | def lengths(self):
    method modality_lengths (line 1274) | def modality_lengths(self):
    method process_image (line 1285) | def process_image(self, image_file, overwrite_image_aspect_ratio=None):
    method process_video (line 1344) | def process_video(self, video_file, data_anno, data_args):
    method __getitem__ (line 1413) | def __getitem__(self, i) -> Dict[str, torch.Tensor]:
    method _get_item (line 1448) | def _get_item(self, i) -> Dict[str, torch.Tensor]:
  class DataCollatorForSupervisedDataset (line 1538) | class DataCollatorForSupervisedDataset(object):
    method pad_sequence (line 1543) | def pad_sequence(self, input_ids, batch_first, padding_value):
    method __call__ (line 1551) | def __call__(self, instances: Sequence[Dict]) -> Dict[str, torch.Tensor]:
  function make_supervised_data_module (line 1588) | def make_supervised_data_module(tokenizer: transformers.PreTrainedTokeni...
  function get_model (line 1595) | def get_model(model_args, training_args, bnb_model_from_pretrained_args):
  function train (line 1814) | def train(attn_implementation=None):

FILE: llava-train_videochat/llava/utils.py
  function process_video_with_decord (line 25) | def process_video_with_decord(video_file, data_args):
  function process_video_with_pyav (line 41) | def process_video_with_pyav(video_file, data_args):
  function rank0_print (line 66) | def rank0_print(*args):
  function rank_print (line 74) | def rank_print(*args):
  function build_logger (line 80) | def build_logger(logger_name, logger_filename):
  class StreamToLogger (line 122) | class StreamToLogger(object):
    method __init__ (line 127) | def __init__(self, logger, log_level=logging.INFO):
    method __getattr__ (line 133) | def __getattr__(self, attr):
    method write (line 136) | def write(self, buf):
    method flush (line 150) | def flush(self):
  function disable_torch_init (line 156) | def disable_torch_init():
  function violates_moderation (line 166) | def violates_moderation(text):
  function pretty_print_semaphore (line 188) | def pretty_print_semaphore(semaphore):

FILE: llava-train_videochat/llava/video_utils.py
  function get_index (line 18) | def get_index(num_frames, num_segments):
  function lazy_load_s3video (line 27) | def lazy_load_s3video(s3path_video, num_frames, video_start, video_end, ...
  function pts_to_secs (line 71) | def pts_to_secs(pts: int, time_base: float, start_pts: int) -> float:
  function get_pyav_video_duration (line 86) | def get_pyav_video_duration(video_reader):
  function get_frame_indices (line 97) | def get_frame_indices(num_frames, vlen, sample='middle', fix_start=None,...
  function read_frames_av (line 167) | def read_frames_av(video_path, num_frames, sample='rand', client=None, f...
  function read_frames_gif (line 196) | def read_frames_gif(
  function read_frames_decord (line 248) | def read_frames_decord(
  function read_frames_img (line 298) | def read_frames_img(
  function read_frames_fake (line 400) | def read_frames_fake(

FILE: lmms-eval_videochat/lmms_eval/__main__.py
  function _handle_non_serializable (line 30) | def _handle_non_serializable(o):
  function parse_eval_args (line 39) | def parse_eval_args() -> argparse.Namespace:
  function cli_evaluate (line 152) | def cli_evaluate(args: Union[argparse.Namespace, None] = None) -> None:
  function cli_evaluate_single (line 228) | def cli_evaluate_single(args: Union[argparse.Namespace, None] = None) ->...
  function print_results (line 345) | def print_results(args, results):

FILE: lmms-eval_videochat/lmms_eval/api/filter.py
  class Filter (line 8) | class Filter:
    method __init__ (line 17) | def __init__(self, *args, **kwargs) -> None:
    method apply (line 22) | def apply(self, resps, docs):
  class FilterEnsemble (line 33) | class FilterEnsemble:
    method apply (line 44) | def apply(self, instances: List[Instance], docs: List[Dataset]) -> None:

FILE: lmms-eval_videochat/lmms_eval/api/instance.py
  class Instance (line 6) | class Instance:
    method __post_init__ (line 20) | def __post_init__(self) -> None:
    method args (line 25) | def args(self):

FILE: lmms-eval_videochat/lmms_eval/api/metrics.py
  function bypass_agg (line 17) | def bypass_agg(arr):
  function mean (line 22) | def mean(arr):
  function median (line 27) | def median(arr):
  function perplexity (line 34) | def perplexity(items):
  function weighted_perplexity (line 41) | def weighted_perplexity(items):
  function bits_per_byte (line 46) | def bits_per_byte(items):
  function f1_score (line 51) | def f1_score(items):
  function matthews_corrcoef (line 61) | def matthews_corrcoef(items):
  function bleu (line 70) | def bleu(items):
  function chrf (line 88) | def chrf(items):
  function ter (line 103) | def ter(items):
  function acc_fn (line 124) | def acc_fn(items):  # This is a passthrough function
  function acc_norm_fn (line 134) | def acc_norm_fn(items):  # This is a passthrough function
  function acc_mutual_info_fn (line 144) | def acc_mutual_info_fn(items):  # This is a passthrough function
  function exact_match_fn (line 156) | def exact_match_fn(**kwargs):
  function perplexity_fn (line 166) | def perplexity_fn(items):  # This is a passthrough function
  function levenshtein_distance (line 170) | def levenshtein_distance(s1, s2):
  function anls (line 192) | def anls(
  function pop_stddev (line 216) | def pop_stddev(arr):
  function sample_stddev (line 221) | def sample_stddev(arr):
  function mean_stderr (line 226) | def mean_stderr(arr):
  function bypass (line 236) | def bypass(items):
  function mcc_fn (line 246) | def mcc_fn(items):  # This is a passthrough function
  function f1_fn (line 256) | def f1_fn(items):  # This is a passthrough function
  function bleu_fn (line 266) | def bleu_fn(items):  # This is a passthrough function
  function chrf_fn (line 276) | def chrf_fn(items):  # This is a passthrough function
  function ter_fn (line 286) | def ter_fn(items):  # This is a passthrough function
  function acc_all (line 296) | def acc_all(items):
  function acc_all_stderr (line 315) | def acc_all_stderr(items):
  function metric_max_over_ground_truths (line 333) | def metric_max_over_ground_truths(metric_fn, prediction, ground_truths):
  function weighted_mean (line 342) | def weighted_mean(items):
  function is_non_str_iterable (line 347) | def is_non_str_iterable(obj):
  function _sacreformat (line 351) | def _sacreformat(refs, preds):
  class _bootstrap_internal (line 382) | class _bootstrap_internal:
    method __init__ (line 383) | def __init__(self, f, n) -> None:
    method __call__ (line 387) | def __call__(self, v):
  function bootstrap_stderr (line 397) | def bootstrap_stderr(f, xs, iters):
  function stderr_for_metric (line 426) | def stderr_for_metric(metric, bootstrap_iters):

FILE: lmms-eval_videochat/lmms_eval/api/model.py
  class lmms (line 18) | class lmms(abc.ABC):
    method __init__ (line 19) | def __init__(self) -> None:
    method loglikelihood (line 31) | def loglikelihood(self, requests: List[Instance]) -> List[Tuple[float,...
    method generate_until (line 59) | def generate_until(self, requests) -> List[str]:
    method create_from_arg_string (line 78) | def create_from_arg_string(cls: Type[T], arg_string: str, additional_c...
    method rank (line 95) | def rank(self):
    method world_size (line 102) | def world_size(self):
    method set_cache_hook (line 108) | def set_cache_hook(self, cache_hook) -> None:
  function hash_args (line 113) | def hash_args(attr, args):
  class CacheHook (line 118) | class CacheHook:
    method __init__ (line 119) | def __init__(self, cachinglm) -> None:
    method add_partial (line 126) | def add_partial(self, attr, req, res) -> None:
  class CachingLMM (line 133) | class CachingLMM:
    method __init__ (line 134) | def __init__(self, lm, cache_db) -> None:
    method __getattr__ (line 151) | def __getattr__(self, attr):
    method get_cache_hook (line 202) | def get_cache_hook(self):

FILE: lmms-eval_videochat/lmms_eval/api/registry.py
  function register_model (line 11) | def register_model(*names):
  function get_model (line 27) | def get_model(model_name):
  function register_task (line 41) | def register_task(name):
  function register_group (line 53) | def register_group(name):
  function register_metric (line 82) | def register_metric(**args):
  function get_metric (line 109) | def get_metric(name: str, hf_evaluate_metric=False) -> Callable:
  function register_aggregation (line 125) | def register_aggregation(name):
  function get_aggregation (line 135) | def get_aggregation(name):
  function get_metric_aggregation (line 144) | def get_metric_aggregation(name):
  function is_higher_better (line 153) | def is_higher_better(metric_name):

FILE: lmms-eval_videochat/lmms_eval/api/samplers.py
  class ContextSampler (line 1) | class ContextSampler:
    method __init__ (line 2) | def __init__(self, docs, task, fewshot_indices=None, rnd=None) -> None:
    method get_context (line 20) | def get_context(self, doc, num_fewshot):
    method sample (line 50) | def sample(self, n):
  class FirstNSampler (line 58) | class FirstNSampler(ContextSampler):
    method sample (line 59) | def sample(self, n) -> None:
  class BalancedSampler (line 68) | class BalancedSampler(ContextSampler):
    method sample (line 69) | def sample(self, n) -> None:
  class ManualSampler (line 78) | class ManualSampler(ContextSampler):
    method sample (line 79) | def sample(self, n) -> None:
  function get_sampler (line 90) | def get_sampler(name):

FILE: lmms-eval_videochat/lmms_eval/api/task.py
  class TaskConfig (line 55) | class TaskConfig(dict):
    method __post_init__ (line 103) | def __post_init__(self) -> None:
    method __getitem__ (line 130) | def __getitem__(self, item):
    method __setitem__ (line 133) | def __setitem__(self, item, value):
    method to_dict (line 136) | def to_dict(self):
  class Task (line 157) | class Task(abc.ABC):
    method __init__ (line 178) | def __init__(
    method download (line 215) | def download(self, data_dir=None, cache_dir=None, download_mode=None) ...
    method config (line 268) | def config(self):
    method has_training_docs (line 273) | def has_training_docs(self):
    method has_validation_docs (line 278) | def has_validation_docs(self):
    method has_test_docs (line 283) | def has_test_docs(self):
    method training_docs (line 287) | def training_docs(self):
    method validation_docs (line 294) | def validation_docs(self):
    method test_docs (line 301) | def test_docs(self):
    method fewshot_docs (line 308) | def fewshot_docs(self):
    method _process_doc (line 322) | def _process_doc(self, doc):
    method instances (line 334) | def instances(self):
    method fewshot_examples (line 340) | def fewshot_examples(self, k, rnd):
    method doc_to_decontamination_query (line 346) | def doc_to_decontamination_query(self, doc) -> None:
    method doc_to_text (line 351) | def doc_to_text(self, doc):
    method doc_to_target (line 355) | def doc_to_target(self, doc):
    method build_all_requests (line 359) | def build_all_requests(self, limit=None, rank=None, world_size=None) -...
    method construct_requests (line 399) | def construct_requests(self, doc_id, ctx, **kwargs):
    method process_results (line 418) | def process_results(self, doc, results):
    method aggregation (line 431) | def aggregation(self):
    method higher_is_better (line 440) | def higher_is_better(self):
    method count_bytes (line 449) | def count_bytes(cls, doc):
    method fewshot_context (line 454) | def fewshot_context(
    method apply_filters (line 504) | def apply_filters(self):
    method dump_config (line 512) | def dump_config(self) -> dict:
    method override_metric (line 522) | def override_metric(self, metric_name: str) -> None:
  class ConfigurableTask (line 546) | class ConfigurableTask(Task):
    method __init__ (line 551) | def __init__(self, model_name) -> None:  # TODO no super() call here
    method _prepare_model_specific_config (line 639) | def _prepare_model_specific_config(self):
    method _prepare_metric_and_aggregation (line 665) | def _prepare_metric_and_aggregation(self):
    method download (line 717) | def download(self, dataset_kwargs=None) -> None:
    method has_training_docs (line 915) | def has_training_docs(self) -> bool:
    method has_validation_docs (line 921) | def has_validation_docs(self) -> bool:
    method has_test_docs (line 927) | def has_test_docs(self) -> bool:
    method training_docs (line 933) | def training_docs(self) -> datasets.Dataset:
    method validation_docs (line 937) | def validation_docs(self) -> datasets.Dataset:
    method test_docs (line 941) | def test_docs(self) -> datasets.Dataset:
    method fewshot_docs (line 945) | def fewshot_docs(self):
    method fewshot_context (line 954) | def fewshot_context(self, doc_id, num_fewshot, split):
    method apply_filters (line 983) | def apply_filters(self):
    method should_decontaminate (line 991) | def should_decontaminate(self):
    method doc_to_decontamination_query (line 994) | def doc_to_decontamination_query(self, doc):
    method _process_doc (line 1007) | def _process_doc(self, doc):
    method doc_to_text (line 1018) | def doc_to_text(self, doc):
    method doc_to_target (line 1055) | def doc_to_target(self, doc: dict) -> Union[int, str, list]:
    method doc_to_visual (line 1092) | def doc_to_visual(self, doc: dict) -> Union[int, str, list]:
    method doc_to_choice (line 1109) | def doc_to_choice(self, doc: Any) -> List[str]:
    method construct_requests (line 1131) | def construct_requests(self, doc_id: int, ctx: str, **kwargs) -> Union...
    method process_results (line 1185) | def process_results(self, doc, results, full_docs=None):
    method aggregation (line 1330) | def aggregation(self):
    method higher_is_better (line 1333) | def higher_is_better(self):

FILE: lmms-eval_videochat/lmms_eval/evaluator.py
  function simple_evaluate (line 35) | def simple_evaluate(
  function evaluate (line 169) | def evaluate(

FILE: lmms-eval_videochat/lmms_eval/filters/__init__.py
  function get_filter (line 24) | def get_filter(filter_name):
  function build_filter_ensemble (line 31) | def build_filter_ensemble(filter_name, components):

FILE: lmms-eval_videochat/lmms_eval/filters/decontamination.py
  class DecontaminationFilter (line 4) | class DecontaminationFilter(Filter):
    method __init__ (line 11) | def __init__(self, path) -> None:
    method apply (line 19) | def apply(self, resps, docs) -> None:

FILE: lmms-eval_videochat/lmms_eval/filters/extraction.py
  class WhitespaceFilter (line 7) | class WhitespaceFilter(Filter):
    method __init__ (line 10) | def __init__(self) -> None:
    method apply (line 13) | def apply(self, resps, docs):
  class RegexFilter (line 29) | class RegexFilter(Filter):
    method __init__ (line 32) | def __init__(
    method apply (line 47) | def apply(self, resps, docs):
  class MultiChoiceRegexFilter (line 73) | class MultiChoiceRegexFilter(RegexFilter):
    method __init__ (line 81) | def __init__(
    method apply (line 104) | def apply(self, resps, docs):
  class ExtendedRegexFilter (line 175) | class ExtendedRegexFilter(RegexFilter):
    method __init__ (line 178) | def __init__(
    method filter_ignores (line 192) | def filter_ignores(self, st):
    method find_match (line 205) | def find_match(self, regex, resp, convert_dict={}):
  class SimpleMultiChoiceRegexFilter (line 218) | class SimpleMultiChoiceRegexFilter(ExtendedRegexFilter):
    method __init__ (line 219) | def __init__(self, *args, **kwargs):
    method apply (line 231) | def apply(self, resps, docs):

FILE: lmms-eval_videochat/lmms_eval/filters/selection.py
  class TakeFirstFilter (line 6) | class TakeFirstFilter(Filter):
    method __init__ (line 7) | def __init__(self) -> None:
    method apply (line 12) | def apply(self, resps, docs):
  class TakeKFilter (line 19) | class TakeKFilter(Filter):
    method __init__ (line 20) | def __init__(self, *args, **kwargs) -> None:
    method apply (line 25) | def apply(self, resps, docs):
  class MajorityVoteFilter (line 31) | class MajorityVoteFilter(Filter):
    method __init__ (line 32) | def __init__(self) -> None:
    method apply (line 37) | def apply(self, resps, docs):

FILE: lmms-eval_videochat/lmms_eval/filters/transformation.py
  class LowercaseFilter (line 4) | class LowercaseFilter(Filter):
    method __init__ (line 5) | def __init__(self) -> None:
    method apply (line 8) | def apply(self, resps, docs):
  class UppercaseFilter (line 15) | class UppercaseFilter(Filter):
    method __init__ (line 16) | def __init__(self) -> None:
    method apply (line 19) | def apply(self, resps, docs):
  class MapFilter (line 26) | class MapFilter(Filter):
    method __init__ (line 27) | def __init__(self, mapping_dict: dict = {}, default_value=None) -> None:
    method apply (line 44) | def apply(self, resps, docs):

FILE: lmms-eval_videochat/lmms_eval/logging_utils.py
  function remove_none_pattern (line 27) | def remove_none_pattern(input_string):
  function _handle_non_serializable (line 40) | def _handle_non_serializable(o: Any) -> Union[int, str, list]:
  function get_wandb_printer (line 59) | def get_wandb_printer() -> Literal["Printer"]:
  class WandbLogger (line 69) | class WandbLogger:
    method __init__ (line 70) | def __init__(self, args):
    method finish (line 82) | def finish(self):
    method init_run (line 86) | def init_run(self):
    method post_init (line 100) | def post_init(self, results: Dict[str, Any]) -> None:
    method _get_config (line 105) | def _get_config(self) -> Dict[str, Any]:
    method _sanitize_results_dict (line 116) | def _sanitize_results_dict(self) -> Tuple[Dict[str, str], Dict[str, An...
    method _log_results_as_table (line 156) | def _log_results_as_table(self) -> None:
    method _log_results_as_artifact (line 216) | def _log_results_as_artifact(self) -> None:
    method log_eval_result (line 224) | def log_eval_result(self) -> None:
    method _generate_dataset (line 240) | def _generate_dataset(self, data: List[Dict[str, Any]], config: Dict[s...
    method _log_samples_as_artifact (line 306) | def _log_samples_as_artifact(self, data: List[Dict[str, Any]], task_na...
    method log_eval_samples (line 320) | def log_eval_samples(self, samples: Dict[str, List[Dict[str, Any]]]) -...

FILE: lmms-eval_videochat/lmms_eval/models/videochat_flash.py
  class VideoChat_Flash (line 40) | class VideoChat_Flash(lmms):
    method __init__ (line 45) | def __init__(
    method config (line 128) | def config(self):
    method tokenizer (line 133) | def tokenizer(self):
    method model (line 137) | def model(self):
    method eot_token_id (line 145) | def eot_token_id(self):
    method max_length (line 150) | def max_length(self):
    method batch_size (line 156) | def batch_size(self):
    method device (line 160) | def device(self):
    method rank (line 164) | def rank(self):
    method world_size (line 168) | def world_size(self):
    method tok_encode (line 171) | def tok_encode(self, string: str, left_truncate_len=None, add_special_...
    method tok_decode (line 180) | def tok_decode(self, tokens):
    method loglikelihood (line 186) | def loglikelihood(self, requests: List[Instance]) -> List[Tuple[float,...
    method flatten (line 190) | def flatten(self, input):
    method generate_until (line 197) | def generate_until(self, requests: List[Instance]) -> List[str]:

FILE: lmms-eval_videochat/lmms_eval/tasks/__init__.py
  function register_configurable_task (line 22) | def register_configurable_task(config: Dict[str, str]) -> int:
  function register_configurable_group (line 47) | def register_configurable_group(config: Dict[str, str]) -> int:
  function get_task_name_from_config (line 61) | def get_task_name_from_config(task_config: Dict[str, str]) -> str:
  function include_task_folder (line 68) | def include_task_folder(task_dir: str, register_task: bool = True) -> None:
  function include_path (line 107) | def include_path(task_dir):
  function initialize_tasks (line 114) | def initialize_tasks(verbosity="INFO"):
  function get_task (line 127) | def get_task(task_name, model_name):
  function get_task_name_from_object (line 136) | def get_task_name_from_object(task_object):
  function get_task_dict (line 147) | def get_task_dict(task_name_list: List[Union[str, Dict, Task]], model_na...

FILE: lmms-eval_videochat/lmms_eval/tasks/_task_utils/file_utils.py
  function generate_submission_file (line 4) | def generate_submission_file(file_name, args, subpath="submissions"):

FILE: lmms-eval_videochat/lmms_eval/tasks/_task_utils/video_loader.py
  function get_cache_dir (line 4) | def get_cache_dir(config, sub_dir="videos"):
  function _get_video_file (line 12) | def _get_video_file(prefix: str, video_name: str, suffix: str):
  function get_video (line 21) | def get_video(prefix: str, video_name: str, suffix: str = "mp4"):

FILE: lmms-eval_videochat/lmms_eval/tasks/_task_utils/vqa_eval_metric.py
  class EvalAIAnswerProcessor (line 4) | class EvalAIAnswerProcessor:
    method __init__ (line 175) | def __init__(self, *args, **kwargs):
    method word_tokenize (line 178) | def word_tokenize(self, word):
    method process_punctuation (line 183) | def process_punctuation(self, in_text):
    method process_digit_article (line 193) | def process_digit_article(self, in_text):
    method __call__ (line 208) | def __call__(self, item):

FILE: lmms-eval_videochat/lmms_eval/tasks/longvideobench/utils.py
  function timestamp_to_seconds (line 22) | def timestamp_to_seconds(timestamp):
  function load_video (line 30) | def load_video(video_file, duration, max_num_frames=16):
  function compute_frame_timestamps (line 50) | def compute_frame_timestamps(duration, max_num_frames=16):
  function insert_subtitles_into_frames (line 57) | def insert_subtitles_into_frames(frame_timestamps, subtitles, starting_t...
  function longvideobench_doc_to_text (line 119) | def longvideobench_doc_to_text(doc, lmms_eval_specific_kwargs):
  function longvideobench_doc_to_visual_v (line 164) | def longvideobench_doc_to_visual_v(doc):
  function longvideobench_doc_to_visual_i (line 180) | def longvideobench_doc_to_visual_i(doc):
  function get_multi_choice_info (line 197) | def get_multi_choice_info(options):
  function parse_multi_choice_response (line 214) | def parse_multi_choice_response(response, all_choices, index2ans):
  function evaluate_longvideobench (line 275) | def evaluate_longvideobench(samples):
  function eval_multi_choice (line 294) | def eval_multi_choice(gold_i, pred_i):
  function calculate_ins_level_acc (line 308) | def calculate_ins_level_acc(results):
  function longvideobench_process_results (line 322) | def longvideobench_process_results(doc, results):
  function longvideobench_aggregate_results (line 344) | def longvideobench_aggregate_results(results):
  function longvideobench_aggregate_results_for_submission (line 370) | def longvideobench_aggregate_results_for_submission(results, args):

FILE: lmms-eval_videochat/lmms_eval/tasks/lvbench/utils.py
  function lvbench_mc_doc_to_visual (line 30) | def lvbench_mc_doc_to_visual(doc, lmms_eval_specific_kwargs=None):
  function lvbench_mc_doc_to_text (line 51) | def lvbench_mc_doc_to_text(doc, lmms_eval_specific_kwargs=None):
  function mcq_acc (line 63) | def mcq_acc(answer, pred):
  function lvbench_mc_process_results (line 115) | def lvbench_mc_process_results(doc, results):
  function lvbench_mc_aggregate_results (line 144) | def lvbench_mc_aggregate_results(results):

FILE: lmms-eval_videochat/lmms_eval/tasks/mlvu_mc/utils.py
  function mlvu_mc_doc_to_visual (line 43) | def mlvu_mc_doc_to_visual(doc, lmms_eval_specific_kwargs=None):
  function mlvu_mc_frames_doc_to_visual (line 70) | def mlvu_mc_frames_doc_to_visual(doc, lmms_eval_specific_kwargs=None):
  function mlvu_mc_doc_to_text (line 96) | def mlvu_mc_doc_to_text(doc, lmms_eval_specific_kwargs=None):
  function mcq_acc (line 112) | def mcq_acc(answer, pred):
  function mlvu_mc_process_results (line 164) | def mlvu_mc_process_results(doc, results):
  function mlvu_mc_aggregate_results (line 193) | def mlvu_mc_aggregate_results(results):

FILE: lmms-eval_videochat/lmms_eval/tasks/mvbench/utils.py
  function mvbench_doc_to_visual (line 56) | def mvbench_doc_to_visual(doc, lmms_eval_specific_kwargs=None):
  function mvbench_frames_doc_to_visual (line 82) | def mvbench_frames_doc_to_visual(doc, lmms_eval_specific_kwargs=None):
  function mvbench_doc_to_text (line 108) | def mvbench_doc_to_text(doc, lmms_eval_specific_kwargs=None):
  function mcq_acc (line 125) | def mcq_acc(answer, pred):
  function mvbench_process_results (line 177) | def mvbench_process_results(doc, results):
  function mvbench_aggregate_results (line 206) | def mvbench_aggregate_results(results):

FILE: lmms-eval_videochat/lmms_eval/tasks/perceptiontest/val/utils.py
  function perceptiontest_val_doc_to_visual (line 35) | def perceptiontest_val_doc_to_visual(doc):
  function perceptiontest_val_doc_to_text (line 48) | def perceptiontest_val_doc_to_text(doc, lmms_eval_specific_kwargs=None):
  function perceptiontest_val_doc_to_answer (line 75) | def perceptiontest_val_doc_to_answer(doc):
  function perceptiontest_val_process_results_mc_ppl (line 80) | def perceptiontest_val_process_results_mc_ppl(doc, result):
  function perceptiontest_val_process_results_mc (line 110) | def perceptiontest_val_process_results_mc(doc, result):
  function perceptiontest_val_aggregate_accuracy (line 139) | def perceptiontest_val_aggregate_accuracy(results, args):
  function perceptiontest_val_doc_to_choice (line 152) | def perceptiontest_val_doc_to_choice(doc):

FILE: lmms-eval_videochat/lmms_eval/tasks/temporal_grounding/eval_tvg.py
  function read_json (line 11) | def read_json(path):
  function write_json (line 17) | def write_json(path, data):
  function extract_time (line 24) | def extract_time(paragraph):
  function iou (line 101) | def iou(A, B):

FILE: lmms-eval_videochat/lmms_eval/tasks/temporal_grounding/utils.py
  function temporal_grounding_doc_to_visual (line 32) | def temporal_grounding_doc_to_visual(doc, lmms_eval_specific_kwargs=None):
  function temporal_grounding_doc_to_text (line 45) | def temporal_grounding_doc_to_text(doc, lmms_eval_specific_kwargs=None):
  function temporal_grounding_doc_to_answer (line 59) | def temporal_grounding_doc_to_answer(doc):
  function temporal_grounding_process_results_generation (line 64) | def temporal_grounding_process_results_generation(doc, result):
  function temporal_grounding_aggregate_charades (line 69) | def temporal_grounding_aggregate_charades(results, args):
  function temporal_grounding_aggregate_submissions (line 72) | def temporal_grounding_aggregate_submissions(results, args, task):

FILE: lmms-eval_videochat/lmms_eval/tasks/videomme/utils.py
  function parse_subtitle_time (line 97) | def parse_subtitle_time(time_str):
  function load_subtitles (line 103) | def load_subtitles(subtitle_path):
  function convert_time_to_frame (line 119) | def convert_time_to_frame(time_in_seconds, fps):
  function extract_subtitles (line 123) | def extract_subtitles(video_path, subtitle_path):
  function parse_subtitle_time (line 138) | def parse_subtitle_time(time_str):
  function load_subtitles (line 144) | def load_subtitles(subtitle_path):
  function convert_time_to_frame (line 160) | def convert_time_to_frame(time_in_seconds, fps):
  function extract_subtitles (line 164) | def extract_subtitles(video_path, subtitle_path):
  function videomme_doc_to_visual (line 179) | def videomme_doc_to_visual(doc):
  function videomme_doc_to_text (line 195) | def videomme_doc_to_text(doc, lmms_eval_specific_kwargs=None):
  function videomme_doc_to_text_subtitle (line 221) | def videomme_doc_to_text_subtitle(doc, lmms_eval_specific_kwargs=None):
  function extract_characters_regex (line 282) | def extract_characters_regex(s):
  function videomme_process_results (line 313) | def videomme_process_results(doc, results):
  function videomme_aggregate_results (line 334) | def videomme_aggregate_results(results):

FILE: lmms-eval_videochat/lmms_eval/utils.py
  function is_json (line 45) | def is_json(string):
  function escaped_split (line 53) | def escaped_split(text, sep_char, maxsplit=-1):
  function handle_arg_string (line 74) | def handle_arg_string(arg):
  function simple_parse_args_string (line 87) | def simple_parse_args_string(args_string):
  function join_iters (line 111) | def join_iters(iters):
  function chunks (line 116) | def chunks(iter, n: int = 0, fn=None):
  function group (line 154) | def group(arr, fn):
  class MultiChoice (line 163) | class MultiChoice:
    method __init__ (line 164) | def __init__(self, choices) -> None:
    method __contains__ (line 168) | def __contains__(self, values) -> bool:
    method __iter__ (line 177) | def __iter__(self) -> Iterator:
  function pattern_match (line 184) | def pattern_match(patterns, source_list):
  function general_detokenize (line 195) | def general_detokenize(string):
  function get_rolling_token_windows (line 205) | def get_rolling_token_windows(token_list, prefix_token, max_seq_len, con...
  function make_disjoint_window (line 246) | def make_disjoint_window(pair):
  class Reorderer (line 252) | class Reorderer:
    method __init__ (line 253) | def __init__(self, arr: List[Any], fn: Callable) -> None:
    method get_reordered (line 270) | def get_reordered(self):
    method get_original (line 278) | def get_original(self, newarr):
  class Grouper (line 300) | class Grouper:
    method __init__ (line 307) | def __init__(self, arr, fn) -> None:
    method get_grouped (line 325) | def get_grouped(self):
    method get_original (line 336) | def get_original(self, grouped_dict):
  function make_table (line 358) | def make_table(result_dict, column: str = "results"):
  function positional_deprecated (line 434) | def positional_deprecated(fn):
  function find_test_root (line 450) | def find_test_root(start_path: pathlib.Path) -> pathlib.Path:
  function run_task_tests (line 466) | def run_task_tests(task_list: List[str]):
  function get_git_commit_hash (line 486) | def get_git_commit_hash():
  function get_datetime_str (line 500) | def get_datetime_str(timezone="Asia/Singapore"):
  function import_function (line 511) | def import_function(loader, node):
  function load_yaml_config (line 532) | def load_yaml_config(yaml_path=None, yaml_config=None, yaml_dir=None):
  function regex_replace (line 571) | def regex_replace(string, pattern, repl, count: int = 0):
  function apply_template (line 580) | def apply_template(template: str, doc: dict) -> str:
  function create_iterator (line 585) | def create_iterator(raw_iterator, rank, world_size, limit=None):
  function pad_and_concat (line 594) | def pad_and_concat(
  function clear_torch_cache (line 643) | def clear_torch_cache() -> None:
  function get_dtype (line 648) | def get_dtype(dtype: Union[str, torch.dtype]) -> torch.dtype:
  class MultiTokenEOSCriteria (line 659) | class MultiTokenEOSCriteria(transformers.StoppingCriteria):
    method __init__ (line 662) | def __init__(
    method __call__ (line 683) | def __call__(self, input_ids, scores, **kwargs) -> bool:
  function stop_sequences_criteria (line 694) | def stop_sequences_criteria(
  function divide (line 708) | def divide(iterable, n) -> List[Iterator]:
  class Collator (line 759) | class Collator:
    method __init__ (line 766) | def __init__(
    method group_by_index (line 782) | def group_by_index(self) -> None:
    method get_batched (line 785) | def get_batched(self, n: int = 1, batch_fn: Optional[Callable] = None)...
    method _reorder (line 809) | def _reorder(self, arr: Union[List, Tuple[Tuple[int, Any], ...]]) -> L...
    method get_original (line 823) | def get_original(self, newarr: List) -> List:
    method __len__ (line 844) | def __len__(self):
    method group (line 848) | def group(arr: Iterable, fn: Callable, values: bool = False) -> Iterable:
    method get_chunks (line 878) | def get_chunks(_iter, n: int = 0, fn=None):

FILE: xtuner-eval_niah/llava/conversation.py
  class SeparatorStyle (line 11) | class SeparatorStyle(Enum):
  class Conversation (line 26) | class Conversation:
    method get_prompt (line 47) | def get_prompt(self):
    method append_message (line 178) | def append_message(self, role, message):
    method process_image (line 181) | def process_image(self, image, image_process_mode, return_pil=False, i...
    method get_images (line 227) | def get_images(self, return_pil=False, return_path=False):
    method is_image_file (line 242) | def is_image_file(self, filename):
    method is_video_file (line 246) | def is_video_file(self, filename):
    method to_gradio_chatbot (line 250) | def to_gradio_chatbot(self):
    method copy (line 288) | def copy(self):
    method dict (line 291) | def dict(self):

FILE: xtuner-eval_niah/llava/dist_utils.py
  function get_rank (line 16) | def get_rank() -> int:
  function get_world_size (line 23) | def get_world_size() -> int:
  function setup_for_distributed (line 30) | def setup_for_distributed(is_master):
  function init_distributed_mode (line 44) | def init_distributed_mode(use_dynamic_port: bool = True):

FILE: xtuner-eval_niah/llava/mm_utils.py
  function resize_and_center_crop (line 12) | def resize_and_center_crop(image, shortest_edge_length):
  function auto_pad_images (line 33) | def auto_pad_images(image, grid_params):
  function extract_patches (line 63) | def extract_patches(image, patch_size, overlap_ratio):
  function process_highres_image_crop_split (line 87) | def process_highres_image_crop_split(image, data_args, processor=None):
  function process_highres_image (line 98) | def process_highres_image(image, processor, grid_pinpoints):
  function select_best_resolution (line 119) | def select_best_resolution(original_size, possible_resolutions, max_reso...
  function resize_and_pad_image (line 157) | def resize_and_pad_image(image, target_resolution):
  function divide_to_patches (line 196) | def divide_to_patches(image, patch_size):
  function get_anyres_image_grid_shape (line 218) | def get_anyres_image_grid_shape(image_size, grid_pinpoints, patch_size, ...
  function process_anyres_image (line 251) | def process_anyres_image(image, processor, grid_pinpoints):
  function process_anyres_image_nopad (line 306) | def process_anyres_image_nopad(image, processor, grid_pinpoints):
  function process_anyres_video_nopad (line 365) | def process_anyres_video_nopad(video, processor, grid_pinpoints, max_res...
  function load_image_from_base64 (line 410) | def load_image_from_base64(image):
  function expand2square (line 414) | def expand2square(pil_img, background_color):
  function process_images (line 428) | def process_images(images, image_processor, model_cfg):
  function tokenizer_image_token (line 460) | def tokenizer_image_token(prompt, tokenizer, image_token_index=IMAGE_TOK...
  function get_model_name_from_path (line 482) | def get_model_name_from_path(model_path):
  class KeywordsStoppingCriteria (line 491) | class KeywordsStoppingCriteria(StoppingCriteria):
    method __init__ (line 492) | def __init__(self, keywords, tokenizer, input_ids):
    method __call__ (line 503) | def __call__(self, output_ids: torch.LongTensor, scores: torch.FloatTe...

FILE: xtuner-eval_niah/llava/model/apply_delta.py
  function apply_delta (line 14) | def apply_delta(base_model_path, target_model_path, delta_path):

FILE: xtuner-eval_niah/llava/model/builder.py
  function load_pretrained_model (line 27) | def load_pretrained_model(model_path, model_base, model_name, load_8bit=...

FILE: xtuner-eval_niah/llava/model/consolidate.py
  function consolidate_ckpt (line 14) | def consolidate_ckpt(src_path, dst_path):

FILE: xtuner-eval_niah/llava/model/language_model/llava_qwen.py
  class LlavaQwenConfig (line 35) | class LlavaQwenConfig(Qwen2Config):
  class LlavaQwenModel (line 39) | class LlavaQwenModel(LlavaMetaModel, Qwen2Model):
    method __init__ (line 42) | def __init__(self, config: Qwen2Config):
  class LlavaQwenForCausalLM (line 46) | class LlavaQwenForCausalLM(Qwen2ForCausalLM, LlavaMetaForCausalLM):
    method __init__ (line 49) | def __init__(self, config):
    method get_model (line 60) | def get_model(self):
    method forward (line 63) | def forward(
    method generate (line 119) | def generate(
    method prepare_inputs_for_generation (line 139) | def prepare_inputs_for_generation(self, input_ids, past_key_values=Non...

FILE: xtuner-eval_niah/llava/model/language_model/llava_qwen_flash.py
  class LlavaQwenConfig_Flash (line 35) | class LlavaQwenConfig_Flash(Qwen2Config):
  class LlavaQwenModel_Flash (line 39) | class LlavaQwenModel_Flash(LlavaMetaModel, Qwen2Model_Flash):
    method __init__ (line 42) | def __init__(self, config: Qwen2Config):
  class LlavaQwenForCausalLM_Flash (line 46) | class LlavaQwenForCausalLM_Flash(Qwen2ForCausalLM_Flash, LlavaMetaForCau...
    method __init__ (line 49) | def __init__(self, config):
    method get_model (line 60) | def get_model(self):
    method forward (line 63) | def forward(
    method generate (line 120) | def generate(
    method prepare_inputs_for_generation (line 143) | def prepare_inputs_for_generation(self, input_ids, past_key_values=Non...

FILE: xtuner-eval_niah/llava/model/language_model/modeling_qwen2_flash.py
  function _get_unpad_data (line 70) | def _get_unpad_data(attention_mask):
  class Qwen2RMSNorm (line 83) | class Qwen2RMSNorm(nn.Module):
    method __init__ (line 84) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 92) | def forward(self, hidden_states):
  class Qwen2RotaryEmbedding (line 101) | class Qwen2RotaryEmbedding(nn.Module):
    method __init__ (line 102) | def __init__(self, dim, max_position_embeddings=2048, base=10000, devi...
    method _set_cos_sin_cache (line 116) | def _set_cos_sin_cache(self, seq_len, device, dtype):
    method forward (line 126) | def forward(self, x, seq_len=None):
  function rotate_half (line 138) | def rotate_half(x):
  function apply_rotary_pos_emb (line 146) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
  class Qwen2MLP (line 175) | class Qwen2MLP(nn.Module):
    method __init__ (line 176) | def __init__(self, config):
    method forward (line 186) | def forward(self, x):
  function repeat_kv (line 191) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class Qwen2Attention (line 203) | class Qwen2Attention(nn.Module):
    method __init__ (line 209) | def __init__(self, config: Qwen2Config, layer_idx: Optional[int] = None):
    method forward (line 246) | def forward(
  class Qwen2FlashAttention2 (line 328) | class Qwen2FlashAttention2(Qwen2Attention):
    method __init__ (line 338) | def __init__(self, *args, **kwargs):
    method forward (line 346) | def forward(
    method _flash_attention_forward (line 482) | def _flash_attention_forward(
    method _upad_input (line 587) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class Qwen2SdpaAttention (line 631) | class Qwen2SdpaAttention(Qwen2Attention):
    method forward (line 639) | def forward(
  class Qwen2DecoderLayer (line 725) | class Qwen2DecoderLayer(nn.Module):
    method __init__ (line 726) | def __init__(self, config: Qwen2Config, layer_idx: int):
    method forward (line 741) | def forward(
  class Qwen2PreTrainedModel (line 823) | class Qwen2PreTrainedModel(PreTrainedModel):
    method _init_weights (line 833) | def _init_weights(self, module):
  class Qwen2Model_Flash (line 919) | class Qwen2Model_Flash(Qwen2PreTrainedModel):
    method __init__ (line 927) | def __init__(self, config: Qwen2Config):
    method get_input_embeddings (line 944) | def get_input_embeddings(self):
    method set_input_embeddings (line 947) | def set_input_embeddings(self, value):
    method forward (line 951) | def forward(
    method flash_rank_drop (line 1159) | def flash_rank_drop(
  class Qwen2ForCausalLM_Flash (line 1382) | class Qwen2ForCausalLM_Flash(Qwen2PreTrainedModel):
    method __init__ (line 1385) | def __init__(self, config):
    method get_input_embeddings (line 1394) | def get_input_embeddings(self):
    method set_input_embeddings (line 1397) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1400) | def get_output_embeddings(self):
    method set_output_embeddings (line 1403) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1406) | def set_decoder(self, decoder):
    method get_decoder (line 1409) | def get_decoder(self):
    method forward (line 1414) | def forward(
    method prepare_inputs_for_generation (line 1502) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1560) | def _reorder_cache(past_key_values, beam_idx):

FILE: xtuner-eval_niah/llava/model/llava_arch.py
  class LlavaMetaModel (line 35) | class LlavaMetaModel:
    method __init__ (line 37) | def __init__(self, config):
    method get_vision_tower (line 50) | def get_vision_tower(self):
    method initialize_vision_modules (line 56) | def initialize_vision_modules(self, model_args, fsdp=None):
  function unpad_image (line 116) | def unpad_image(tensor, original_size, is_frame=False):
  class LlavaMetaForCausalLM (line 160) | class LlavaMetaForCausalLM(ABC):
    method get_model (line 163) | def get_model(self):
    method get_vision_tower (line 166) | def get_vision_tower(self):
    method get_4dPool (line 169) | def get_4dPool(self, image_feature):
    method get_2dPool (line 194) | def get_2dPool(self, image_feature):
    method encode_image (line 220) | def encode_image(self, images_list):
    method encode_image_video (line 230) | def encode_image_video(self, images_list, video_idx_in_batch):
    method encode_video (line 255) | def encode_video(self, images_list, video_idx_in_batch):
    method encode_video_image (line 309) | def encode_video_image(self, images_list, video_idx_in_batch):
    method add_token_per_frame (line 364) | def add_token_per_frame(self, image_feature):
    method add_different_token_per_frame (line 373) | def add_different_token_per_frame(self, image_feature):
    method prepare_inputs_labels_for_multimodal (line 377) | def prepare_inputs_labels_for_multimodal(self, input_ids, position_ids...
    method initialize_vision_tokenizer (line 801) | def initialize_vision_tokenizer(self, model_args, tokenizer):

FILE: xtuner-eval_niah/llava/model/make_delta.py
  function make_delta (line 14) | def make_delta(base_model_path, target_model_path, delta_path, hub_repo_...

FILE: xtuner-eval_niah/llava/model/multimodal_encoder/builder.py
  function build_vision_tower (line 11) | def build_vision_tower(vision_tower_cfg, **kwargs):

FILE: xtuner-eval_niah/llava/model/multimodal_encoder/clip_encoder.py
  class CLIPVisionTower (line 12) | class CLIPVisionTower(nn.Module):
    method __init__ (line 13) | def __init__(self, vision_tower, args, delay_load=False):
    method load_model (line 35) | def load_model(self, device_map=None):
    method feature_select (line 46) | def feature_select(self, image_forward_outs):
    method forward (line 68) | def forward(self, images):
    method dummy_feature (line 82) | def dummy_feature(self):
    method dtype (line 86) | def dtype(self):
    method device (line 90) | def device(self):
    method config (line 94) | def config(self):
    method hidden_size (line 101) | def hidden_size(self):
    method num_patches_per_side (line 110) | def num_patches_per_side(self):
    method num_patches (line 114) | def num_patches(self):
    method image_size (line 121) | def image_size(self):
  class CLIPVisionTowerS2 (line 125) | class CLIPVisionTowerS2(CLIPVisionTower):
    method __init__ (line 126) | def __init__(self, vision_tower, args, delay_load=False):
    method load_model (line 141) | def load_model(self, device_map=None):
    method forward_feature (line 155) | def forward_feature(self, images):
    method forward (line 160) | def forward(self, images):
    method hidden_size (line 172) | def hidden_size(self):

FILE: xtuner-eval_niah/llava/model/multimodal_encoder/internvideo2/flash_attention_class.py
  class FlashAttention (line 10) | class FlashAttention(nn.Module):
    method __init__ (line 21) | def __init__(self, softmax_scale=None, attention_dropout=0.0, device=N...
    method forward (line 26) | def forward(self, qkv, key_padding_mask=None, causal=False, cu_seqlens...

FILE: xtuner-eval_niah/llava/model/multimodal_encoder/internvideo2/pos_embed.py
  function get_3d_sincos_pos_embed (line 12) | def get_3d_sincos_pos_embed(embed_dim, grid_size, t_size, cls_token=False):
  function get_2d_sincos_pos_embed (line 66) | def get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):
  function get_1d_sincos_pos_embed (line 86) | def get_1d_sincos_pos_embed(embed_dim, t_size, cls_token=False):
  function get_2d_sincos_pos_embed_from_grid (line 101) | def get_2d_sincos_pos_embed_from_grid(embed_dim, grid):
  function get_1d_sincos_pos_embed_from_grid (line 116) | def get_1d_sincos_pos_embed_from_grid(embed_dim, pos):
  function interpolate_pos_embed_internvideo2 (line 137) | def interpolate_pos_embed_internvideo2(checkpoint_model, model, orig_t_s...
  function interpolate_pos_embed_internvideo2_new (line 227) | def interpolate_pos_embed_internvideo2_new(checkpoint_model, model, orig...

FILE: xtuner-eval_niah/llava/model/multimodal_encoder/internvideo2/vit_scale_clean.py
  class CrossAttention (line 20) | class CrossAttention(nn.Module):
    method __init__ (line 21) | def __init__(
    method forward (line 52) | def forward(self, x, k=None, v=None):
  class AttentiveBlock (line 85) | class AttentiveBlock(nn.Module):
    method __init__ (line 87) | def __init__(self, dim, num_heads, qkv_bias=False, qk_scale=None, drop...
    method forward (line 100) | def forward(self, x_q, x_kv, pos_q, pos_k, bool_masked_pos, rel_pos_bi...
  class AttentionPoolingBlock (line 109) | class AttentionPoolingBlock(AttentiveBlock):
    method forward (line 111) | def forward(self, x):
  class RMSNorm (line 119) | class RMSNorm(nn.Module):
    method __init__ (line 120) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 125) | def forward(self, hidden_states):
  class LayerScale (line 133) | class LayerScale(nn.Module):
    method __init__ (line 134) | def __init__(self, dim, init_values=1e-5, inplace=False, force_fp32=Fa...
    method forward (line 141) | def forward(self, x):
  class Attention (line 151) | class Attention(nn.Module):
    method __init__ (line 152) | def __init__(self, dim, num_heads=8, qkv_bias=False, attn_drop=0., pro...
    method _naive_attn (line 175) | def _naive_attn(self, x):
    method _flash_attn (line 197) | def _flash_attn(self, x, key_padding_mask=None, need_weights=False):
    method forward (line 219) | def forward(self, x):
  class Mlp (line 224) | class Mlp(nn.Module):
    method __init__ (line 228) | def __init__(self, in_features, hidden_features=None, out_features=Non...
    method forward (line 242) | def forward(self, x):
  class Block (line 251) | class Block(nn.Module):
    method __init__ (line 253) | def __init__(
    method forward (line 284) | def forward(self, x, residual=None):
  class PatchEmbed (line 306) | class PatchEmbed(nn.Module):
    method __init__ (line 310) | def __init__(
    method forward (line 334) | def forward(self, x):
  class PretrainVisionTransformer_clean (line 343) | class PretrainVisionTransformer_clean(nn.Module):
    method __init__ (line 344) | def __init__(
    method init_pos_embed (line 453) | def init_pos_embed(self):
    method _init_weights (line 476) | def _init_weights(self, m):
    method fix_init_weight (line 485) | def fix_init_weight(self):
    method dtype (line 494) | def dtype(self):
    method get_num_layers (line 497) | def get_num_layers(self):
    method no_weight_decay (line 501) | def no_weight_decay(self):
    method forward (line 512) | def forward(self, x, mask=None, use_image=False):
  function pretrain_internvideo2_giant_patch14_224_clean (line 572) | def pretrain_internvideo2_giant_patch14_224_clean(config):
  function pretrain_internvideo2_6b_patch14_224_clean (line 608) | def pretrain_internvideo2_6b_patch14_224_clean(config):

FILE: xtuner-eval_niah/llava/model/multimodal_encoder/internvideo2_encoder.py
  class InternVideo2ImageProcessor (line 30) | class InternVideo2ImageProcessor:
    method __init__ (line 31) | def __init__(self, image_mean=(0.485, 0.456, 0.406), image_std=(0.229,...
    method preprocess (line 43) | def preprocess(self, images, return_tensors, target_size=None):
  class InternVideo2VisionConfig (line 69) | class InternVideo2VisionConfig:
    method __init__ (line 72) | def __init__(
  function build_vit (line 101) | def build_vit(config, pt_type='origin'):
  class InternVideo2VisionTower (line 145) | class InternVideo2VisionTower(nn.Module):
    method __init__ (line 146) | def __init__(self, vision_tower, vision_tower_cfg, delay_load=False, p...
    method load_model (line 172) | def load_model(self, device_map=None):
    method forward (line 182) | def forward(self, images):
    method dummy_feature (line 195) | def dummy_feature(self):
    method dtype (line 199) | def dtype(self):
    method device (line 204) | def device(self):
    method hidden_size (line 209) | def hidden_size(self):
    method num_patches (line 213) | def num_patches(self):
    method num_patches_per_side (line 217) | def num_patches_per_side(self):
    method image_size (line 222) | def image_size(self):

FILE: xtuner-eval_niah/llava/model/multimodal_encoder/siglip_encoder.py
  class SigLipImageProcessor (line 34) | class SigLipImageProcessor:
    method __init__ (line 35) | def __init__(self, image_mean=(0.5, 0.5, 0.5), image_std=(0.5, 0.5, 0....
    method preprocess (line 47) | def preprocess(self, images, return_tensors):
  class SigLipVisionConfig (line 70) | class SigLipVisionConfig(PretrainedConfig):
    method __init__ (line 73) | def __init__(
    method from_pretrained (line 103) | def from_pretrained(cls, pretrained_model_name_or_path: Union[str, os....
  class SigLipVisionModelOutput (line 120) | class SigLipVisionModelOutput(ModelOutput):
  class SigLipVisionEmbeddings (line 148) | class SigLipVisionEmbeddings(nn.Module):
    method __init__ (line 149) | def __init__(self, config: SigLipVisionConfig):
    method forward (line 169) | def forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
  class SigLipAttention (line 177) | class SigLipAttention(nn.Module):
    method __init__ (line 181) | def __init__(self, config):
    method forward (line 197) | def forward(
  class SigLipMLP (line 243) | class SigLipMLP(nn.Module):
    method __init__ (line 244) | def __init__(self, config):
    method forward (line 251) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
  class SigLipEncoderLayer (line 259) | class SigLipEncoderLayer(nn.Module):
    method __init__ (line 260) | def __init__(self, config: SigLipVisionConfig):
    method forward (line 269) | def forward(
  class SigLipPreTrainedModel (line 308) | class SigLipPreTrainedModel(PreTrainedModel):
    method _init_weights (line 318) | def _init_weights(self, module):
  class SigLipEncoder (line 324) | class SigLipEncoder(nn.Module):
    method __init__ (line 333) | def __init__(self, config: SigLipVisionConfig):
    method forward (line 340) | def forward(
  class SigLipVisionTransformer (line 408) | class SigLipVisionTransformer(nn.Module):
    method __init__ (line 409) | def __init__(self, config: SigLipVisionConfig):
    method forward (line 419) | def forward(
  class SigLipMultiheadAttentionPoolingHead (line 459) | class SigLipMultiheadAttentionPoolingHead(nn.Module):
    method __init__ (line 462) | def __init__(self, config: SigLipVisionConfig):
    method forward (line 470) | def forward(self, hidden_state):
  class SigLipVisionModel (line 483) | class SigLipVisionModel(SigLipPreTrainedModel):
    method __init__ (line 488) | def __init__(self, config: SigLipVisionConfig):
    method get_input_embeddings (line 496) | def get_input_embeddings(self) -> nn.Module:
    method forward (line 499) | def forward(
  class SigLipVisionTower (line 538) | class SigLipVisionTower(nn.Module):
    method __init__ (line 539) | def __init__(self, vision_tower, vision_tower_cfg, delay_load=False):
    method load_model (line 563) | def load_model(self, device_map=None):
    method forward (line 576) | def forward(self, images):
    method dummy_feature (line 592) | def dummy_feature(self):
    method dtype (line 596) | def dtype(self):
    method device (line 601) | def device(self):
    method hidden_size (line 606) | def hidden_size(self):
    method num_patches (line 610) | def num_patches(self):
    method num_patches_per_side (line 614) | def num_patches_per_side(self):
    method image_size (line 619) | def image_size(self):

FILE: xtuner-eval_niah/llava/model/multimodal_encoder/umt/vit.py
  class DropPath (line 19) | class DropPath(nn.Module):
    method __init__ (line 22) | def __init__(self, drop_prob=None):
    method forward (line 26) | def forward(self, x):
    method extra_repr (line 29) | def extra_repr(self) -> str:
  class Mlp (line 33) | class Mlp(nn.Module):
    method __init__ (line 34) | def __init__(self, in_features, hidden_features=None, out_features=Non...
    method forward (line 43) | def forward(self, x):
  class Attention (line 51) | class Attention(nn.Module):
    method __init__ (line 52) | def __init__(
    method forward (line 85) | def forward(self, x):
  class Block (line 117) | class Block(nn.Module):
    method __init__ (line 118) | def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_sc...
    method forward (line 138) | def forward(self, x):
  class PatchEmbed (line 148) | class PatchEmbed(nn.Module):
    method __init__ (line 151) | def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=...
    method forward (line 167) | def forward(self, x, **kwargs):
  function get_sinusoid_encoding_table (line 178) | def get_sinusoid_encoding_table(n_position, d_hid, ckpt_num_frame=-1, cu...
  function get_sinusoid_encoding_table2 (line 212) | def get_sinusoid_encoding_table2(n_position=784, d_hid=1024, cur_frame=8...
  class PretrainVisionTransformerEncoder (line 259) | class PretrainVisionTransformerEncoder(nn.Module):
    method __init__ (line 262) | def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=...
    method no_weight_decay (line 313) | def no_weight_decay(self):
    method forward_features (line 316) | def forward_features(self, x, use_image=False):
    method forward (line 337) | def forward(self, x, use_image=False):
  class PretrainVisionTransformer (line 342) | class PretrainVisionTransformer(nn.Module):
    method __init__ (line 345) | def __init__(self,
    method _init_weights (line 401) | def _init_weights(self, m):
    method no_weight_decay (line 411) | def no_weight_decay(self):
    method forward (line 414) | def forward(self, x, use_image=False):

FILE: xtuner-eval_niah/llava/model/multimodal_encoder/umt_encoder.py
  class UMTImageProcessor (line 30) | class UMTImageProcessor:
    method __init__ (line 31) | def __init__(self, image_mean=(0.485, 0.456, 0.406), image_std=(0.229,...
    method preprocess (line 43) | def preprocess(self, images, return_tensors, target_size=None):
  class UMTVisionConfig (line 69) | class UMTVisionConfig:
    method __init__ (line 72) | def __init__(
  function build_vit (line 95) | def build_vit(config, pt_type='origin'):
  class UMTVisionTower (line 134) | class UMTVisionTower(nn.Module):
    method __init__ (line 135) | def __init__(self, vision_tower, vision_tower_cfg, delay_load=False, p...
    method load_model (line 160) | def load_model(self, device_map=None):
    method forward (line 170) | def forward(self, images):
    method dummy_feature (line 185) | def dummy_feature(self):
    method dtype (line 189) | def dtype(self):
    method device (line 194) | def device(self):
    method hidden_size (line 199) | def hidden_size(self):
    method num_patches (line 203) | def num_patches(self):
    method num_patches_per_side (line 207) | def num_patches_per_side(self):
    method image_size (line 212) | def image_size(self):

FILE: xtuner-eval_niah/llava/model/multimodal_projector/builder.py
  class IdentityMap (line 7) | class IdentityMap(nn.Module):
    method __init__ (line 8) | def __init__(self):
    method forward (line 11) | def forward(self, x, *args, **kwargs):
    method config (line 15) | def config(self):
  class SimpleResBlock (line 19) | class SimpleResBlock(nn.Module):
    method __init__ (line 20) | def __init__(self, channels):
    method forward (line 25) | def forward(self, x):
  function build_vision_projector (line 30) | def build_vision_projector(config, delay_load=False, **kwargs):

FILE: xtuner-eval_niah/llava/model/multimodal_projector/tome16_mlp_hd64.py
  function bipartite_soft_matching (line 16) | def bipartite_soft_matching(
  function merge_wavg (line 72) | def merge_wavg(
  class ToMe16_mlp_hd64 (line 91) | class ToMe16_mlp_hd64(nn.Module):
    method __init__ (line 92) | def __init__(self, config, vision_cfg):
    method merge_tokens (line 107) | def merge_tokens(self, x, target_num_token):
    method forward (line 143) | def forward(self, x, compress=False, local_num_frames=-1):
    method config (line 169) | def config(self):

FILE: xtuner-eval_niah/llava/model/utils.py
  function auto_upgrade (line 4) | def auto_upgrade(config):

FILE: xtuner-eval_niah/llava/serialize_utils.py
  function get_world_size (line 25) | def get_world_size() -> int:
  function get_rank (line 33) | def get_rank() -> int:
  function get_rank (line 40) | def get_rank() -> int:
  function get_local_rank (line 47) | def get_local_rank() -> int:
  function get_local_size (line 62) | def get_local_size() -> int:
  function _get_global_gloo_group (line 67) | def _get_global_gloo_group():
  function all_gather (line 78) | def all_gather(data, group=None):
  class NumpySerializedList (line 105) | class NumpySerializedList:
    method __init__ (line 106) | def __init__(self, lst: list):
    method __len__ (line 122) | def __len__(self):
    method __getitem__ (line 125) | def __getitem__(self, idx):
  class TorchSerializedList (line 132) | class TorchSerializedList(NumpySerializedList):
    method __init__ (line 133) | def __init__(self, lst: list):
    method __getitem__ (line 138) | def __getitem__(self, idx):
  function local_scatter (line 145) | def local_scatter(array: Optional[List[Any]]):
  class TorchShmSerializedList (line 167) | class TorchShmSerializedList(TorchSerializedList):
    method __init__ (line 168) | def __init__(self, lst: list):
  function local_broadcast_process_authkey (line 194) | def local_broadcast_process_authkey():

FILE: xtuner-eval_niah/llava/train/llava_trainer.py
  function maybe_zero_3 (line 27) | def maybe_zero_3(param, ignore_status=False, name=None):
  function get_mm_adapter_state_maybe_zero_3 (line 42) | def get_mm_adapter_state_maybe_zero_3(named_params, keys_to_match):
  function split_to_even_chunks (line 48) | def split_to_even_chunks(indices, lengths, num_chunks):
  function get_variable_length_grouped_indices (line 70) | def get_variable_length_grouped_indices(lengths, batch_size, world_size,...
  function get_modality_length_grouped_indices (line 86) | def get_modality_length_grouped_indices(lengths, batch_size, world_size,...
  function get_length_grouped_indices (line 126) | def get_length_grouped_indices(lengths, batch_size, world_size, generato...
  function get_length_grouped_indices_auto_single (line 149) | def get_length_grouped_indices_auto_single(lengths, batch_size, world_si...
  function get_modality_length_grouped_indices_auto (line 164) | def get_modality_length_grouped_indices_auto(lengths, batch_size, world_...
  class LengthGroupedSampler (line 193) | class LengthGroupedSampler(Sampler):
    method __init__ (line 199) | def __init__(
    method __len__ (line 220) | def __len__(self):
    method __iter__ (line 223) | def __iter__(self):
  class LLaVATrainer (line 237) | class LLaVATrainer(Trainer):
    method create_accelerator_and_postprocess (line 239) | def create_accelerator_and_postprocess(self):
    method _get_train_sampler (line 270) | def _get_train_sampler(self) -> Optional[torch.utils.data.Sampler]:
    method get_train_dataloader (line 316) | def get_train_dataloader(self) -> DataLoader:
    method create_optimizer (line 353) | def create_optimizer(self):
    method _save_checkpoint (line 432) | def _save_checkpoint(self, model, trial, metrics=None):
    method _save (line 456) | def _save(self, output_dir: Optional[str] = None, state_dict=None):

FILE: xtuner-eval_niah/llava/train/llava_trainer_eval.py
  class LLaVAEvalTrainer (line 7) | class LLaVAEvalTrainer(LLaVATrainer):
    method evaluate (line 8) | def evaluate(self, evaluate_args):

FILE: xtuner-eval_niah/llava/train/train.py
  class ModelArguments (line 64) | class ModelArguments:
  class DataArguments (line 135) | class DataArguments:
  class TrainingArguments (line 159) | class TrainingArguments(transformers.TrainingArguments):
  function maybe_zero_3 (line 208) | def maybe_zero_3(param, ignore_status=False, name=None):
  function get_peft_state_maybe_zero_3 (line 224) | def get_peft_state_maybe_zero_3(named_params, bias):
  function get_peft_state_non_lora_maybe_zero_3 (line 249) | def get_peft_state_non_lora_maybe_zero_3(named_params, require_grad_only...
  function get_mm_adapter_state_maybe_zero_3 (line 257) | def get_mm_adapter_state_maybe_zero_3(named_params, keys_to_match):
  function find_all_linear_names (line 263) | def find_all_linear_names(model):
  function safe_save_model_for_hf_trainer (line 279) | def safe_save_model_for_hf_trainer(trainer: transformers.Trainer, output...
  function smart_tokenizer_and_embedding_resize (line 323) | def smart_tokenizer_and_embedding_resize(
  function _tokenize_fn (line 346) | def _tokenize_fn(strings: Sequence[str], tokenizer: transformers.PreTrai...
  function _mask_targets (line 368) | def _mask_targets(target, tokenized_lens, speakers):
  function _add_speaker_and_signal (line 379) | def _add_speaker_and_signal(header, source, get_conversation=True):
  function preprocess_multimodal (line 399) | def preprocess_multimodal(sources: Sequence[str], data_args: DataArgumen...
  function preprocess_llama_2 (line 431) | def preprocess_llama_2(sources, tokenizer: transformers.PreTrainedTokeni...
  function preprocess_gemma (line 506) | def preprocess_gemma(sources: List[List[Dict[str, str]]], tokenizer: tra...
  function preprocess_qwen (line 585) | def preprocess_qwen(sources, tokenizer: transformers.PreTrainedTokenizer...
  function preprocess_internlm2 (line 665) | def preprocess_internlm2(sources, tokenizer: transformers.PreTrainedToke...
  function preprocess_llama3 (line 742) | def preprocess_llama3(
  function preprocess_v1 (line 828) | def preprocess_v1(sources, tokenizer: transformers.PreTrainedTokenizer, ...
  function preprocess_mpt (line 907) | def preprocess_mpt(sources, tokenizer: transformers.PreTrainedTokenizer,...
  function preprocess_plain (line 988) | def preprocess_plain(
  function preprocess (line 1010) | def preprocess(sources: Sequence[str], tokenizer: transformers.PreTraine...
  class LazySupervisedDataset (line 1065) | class LazySupervisedDataset(Dataset):
    method __init__ (line 1066) | def __init__(self, data_path: str, tokenizer: transformers.PreTrainedT...
    method __len__ (line 1257) | def __len__(self):
    method lengths (line 1261) | def lengths(self):
    method modality_lengths (line 1274) | def modality_lengths(self):
    method process_image (line 1285) | def process_image(self, image_file, overwrite_image_aspect_ratio=None):
    method process_video (line 1344) | def process_video(self, video_file, data_anno, data_args):
    method __getitem__ (line 1413) | def __getitem__(self, i) -> Dict[str, torch.Tensor]:
    method _get_item (line 1448) | def _get_item(self, i) -> Dict[str, torch.Tensor]:
  class DataCollatorForSupervisedDataset (line 1538) | class DataCollatorForSupervisedDataset(object):
    method pad_sequence (line 1543) | def pad_sequence(self, input_ids, batch_first, padding_value):
    method __call__ (line 1551) | def __call__(self, instances: Sequence[Dict]) -> Dict[str, torch.Tensor]:
  function make_supervised_data_module (line 1588) | def make_supervised_data_module(tokenizer: transformers.PreTrainedTokeni...
  function get_model (line 1595) | def get_model(model_args, training_args, bnb_model_from_pretrained_args):
  function train (line 1814) | def train(attn_implementation=None):

FILE: xtuner-eval_niah/llava/utils.py
  function process_video_with_decord (line 25) | def process_video_with_decord(video_file, data_args):
  function process_video_with_pyav (line 41) | def process_video_with_pyav(video_file, data_args):
  function rank0_print (line 66) | def rank0_print(*args):
  function rank_print (line 74) | def rank_print(*args):
  function build_logger (line 80) | def build_logger(logger_name, logger_filename):
  class StreamToLogger (line 122) | class StreamToLogger(object):
    method __init__ (line 127) | def __init__(self, logger, log_level=logging.INFO):
    method __getattr__ (line 133) | def __getattr__(self, attr):
    method write (line 136) | def write(self, buf):
    method flush (line 150) | def flush(self):
  function disable_torch_init (line 156) | def disable_torch_init():
  function violates_moderation (line 166) | def violates_moderation(text):
  function pretty_print_semaphore (line 188) | def pretty_print_semaphore(semaphore):

FILE: xtuner-eval_niah/llava/video_utils.py
  function get_index (line 18) | def get_index(num_frames, num_segments):
  function lazy_load_s3video (line 27) | def lazy_load_s3video(s3path_video, num_frames, video_start, video_end, ...
  function pts_to_secs (line 71) | def pts_to_secs(pts: int, time_base: float, start_pts: int) -> float:
  function get_pyav_video_duration (line 86) | def get_pyav_video_duration(video_reader):
  function get_frame_indices (line 97) | def get_frame_indices(num_frames, vlen, sample='middle', fix_start=None,...
  function read_frames_av (line 167) | def read_frames_av(video_path, num_frames, sample='rand', client=None, f...
  function read_frames_gif (line 196) | def read_frames_gif(
  function read_frames_decord (line 248) | def read_frames_decord(
  function read_frames_img (line 298) | def read_frames_img(
  function read_frames_fake (line 400) | def read_frames_fake(

FILE: xtuner-eval_niah/longva/conversation.py
  class SeparatorStyle (line 11) | class SeparatorStyle(Enum):
  class Conversation (line 26) | class Conversation:
    method get_prompt (line 47) | def get_prompt(self):
    method append_message (line 178) | def append_message(self, role, message):
    method process_image (line 181) | def process_image(self, image, image_process_mode, return_pil=False, i...
    method get_images (line 227) | def get_images(self, return_pil=False, return_path=False):
    method to_gradio_chatbot (line 242) | def to_gradio_chatbot(self):
    method copy (line 266) | def copy(self):
    method dict (line 269) | def dict(self):

FILE: xtuner-eval_niah/longva/mm_utils.py
  function resize_and_center_crop (line 12) | def resize_and_center_crop(image, shortest_edge_length):
  function auto_pad_images (line 33) | def auto_pad_images(image, grid_params):
  function extract_patches (line 63) | def extract_patches(image, patch_size, overlap_ratio):
  function process_highres_image_crop_split (line 87) | def process_highres_image_crop_split(image, data_args, processor=None):
  function process_highres_image (line 98) | def process_highres_image(image, processor, grid_pinpoints):
  function select_best_resolution (line 119) | def select_best_resolution(original_size, possible_resolutions):
  function resize_and_pad_image (line 152) | def resize_and_pad_image(image, target_resolution):
  function divide_to_patches (line 191) | def divide_to_patches(image, patch_size):
  function get_anyres_image_grid_shape (line 213) | def get_anyres_image_grid_shape(image_size, grid_pinpoints, patch_size):
  function process_anyres_image (line 243) | def process_anyres_image(image, processor, grid_pinpoints):
  function load_image_from_base64 (line 296) | def load_image_from_base64(image):
  function expand2square (line 300) | def expand2square(pil_img, background_color):
  function process_images (line 314) | def process_images(images, image_processor, model_cfg):
  function tokenizer_image_token (line 341) | def tokenizer_image_token(prompt, tokenizer, image_token_index=IMAGE_TOK...
  function get_model_name_from_path (line 363) | def get_model_name_from_path(model_path):
  class KeywordsStoppingCriteria (line 372) | class KeywordsStoppingCriteria(StoppingCriteria):
    method __init__ (line 373) | def __init__(self, keywords, tokenizer, input_ids):
    method __call__ (line 384) | def __call__(self, output_ids: torch.LongTensor, scores: torch.FloatTe...

FILE: xtuner-eval_niah/longva/model/apply_delta.py
  function apply_delta (line 14) | def apply_delta(base_model_path, target_model_path, delta_path):

FILE: xtuner-eval_niah/longva/model/builder.py
  function load_pretrained_model (line 27) | def load_pretrained_model(model_path, model_base, model_name, load_8bit=...

FILE: xtuner-eval_niah/longva/model/consolidate.py
  function consolidate_ckpt (line 14) | def consolidate_ckpt(src_path, dst_path):

FILE: xtuner-eval_niah/longva/model/language_model/llava_llama.py
  class LlavaConfig (line 35) | class LlavaConfig(LlamaConfig):
  class LlavaLlamaModel (line 44) | class LlavaLlamaModel(LlavaMetaModel, LlamaModel):
    method __init__ (line 47) | def __init__(self, config: LlamaConfig):
  class LlavaLlamaForCausalLM (line 51) | class LlavaLlamaForCausalLM(LlamaForCausalLM, LlavaMetaForCausalLM):
    method __init__ (line 54) | def __init__(self, config):
    method get_model (line 66) | def get_model(self):
    method forward (line 69) | def forward(
    method generate (line 123) | def generate(
    method prepare_inputs_for_generation (line 144) | def prepare_inputs_for_generation(self, input_ids, past_key_values=Non...

FILE: xtuner-eval_niah/longva/model/language_model/llava_mistral.py
  class LlavaMistralConfig (line 30) | class LlavaMistralConfig(MistralConfig):
  class LlavaMistralModel (line 38) | class LlavaMistralModel(LlavaMetaModel, MistralModel):
    method __init__ (line 41) | def __init__(self, config: MistralConfig):
  class LlavaMistralForCausalLM (line 45) | class LlavaMistralForCausalLM(MistralForCausalLM, LlavaMetaForCausalLM):
    method __init__ (line 48) | def __init__(self, config):
    method get_model (line 59) | def get_model(self):
    method forward (line 62) | def forward(
    method generate (line 96) | def generate(
    method prepare_inputs_for_generation (line 115) | def prepare_inputs_for_generation(self, input_ids, past_key_values=Non...

FILE: xtuner-eval_niah/longva/model/language_model/llava_mpt.py
  class LlavaMptConfig (line 24) | class LlavaMptConfig(MptConfig):
  class LlavaMptModel (line 28) | class LlavaMptModel(LlavaMetaModel, MptModel):
    method __init__ (line 31) | def __init__(self, config: MptConfig):
    method embed_tokens (line 35) | def embed_tokens(self, x):
  class LlavaMptForCausalLM (line 39) | class LlavaMptForCausalLM(MptForCausalLM, LlavaMetaForCausalLM):
    method __init__ (line 43) | def __init__(self, config):
    method get_model (line 61) | def get_model(self):
    method _set_gradient_checkpointing (line 64) | def _set_gradient_checkpointing(self, module, value=False):
    method forward (line 68) | def forward(
    method prepare_inputs_for_generation (line 97) | def prepare_inputs_for_generation(self, input_ids, past_key_values=Non...

FILE: xtuner-eval_niah/longva/model/language_model/llava_qwen.py
  class LlavaQwenConfig (line 35) | class LlavaQwenConfig(Qwen2Config):
  class LlavaQwenModel (line 39) | class LlavaQwenModel(LlavaMetaModel, Qwen2Model):
    method __init__ (line 42) | def __init__(self, config: Qwen2Config):
  class LlavaQwenForCausalLM (line 46) | class LlavaQwenForCausalLM(Qwen2ForCausalLM, LlavaMetaForCausalLM):
    method __init__ (line 49) | def __init__(self, config):
    method get_model (line 60) | def get_model(self):
    method forward (line 63) | def forward(
    method generate (line 117) | def generate(
    method prepare_inputs_for_generation (line 137) | def prepare_inputs_for_generation(self, input_ids, past_key_values=Non...

FILE: xtuner-eval_niah/longva/model/language_model/modeling_llama.py
  function _get_unpad_data (line 61) | def _get_unpad_data(attention_mask):
  class LlamaRMSNorm (line 73) | class LlamaRMSNorm(nn.Module):
    method __init__ (line 74) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 82) | def forward(self, hidden_states):
  class LlamaRotaryEmbedding (line 93) | class LlamaRotaryEmbedding(nn.Module):
    method __init__ (line 94) | def __init__(self, dim, max_position_embeddings=2048, base=10000, devi...
    method sin_cached (line 113) | def sin_cached(self):
    method cos_cached (line 118) | def cos_cached(self):
    method forward (line 123) | def forward(self, x, position_ids, seq_len=None):
  class LlamaLinearScalingRotaryEmbedding (line 142) | class LlamaLinearScalingRotaryEmbedding(LlamaRotaryEmbedding):
    method forward (line 145) | def forward(self, x, position_ids, seq_len=None):
  class LlamaDynamicNTKScalingRotaryEmbedding (line 152) | class LlamaDynamicNTKScalingRotaryEmbedding(LlamaRotaryEmbedding):
    method forward (line 155) | def forward(self, x, position_ids, seq_len=None):
  function rotate_half (line 167) | def rotate_half(x):
  function apply_rotary_pos_emb (line 174) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  class LlamaMLP (line 201) | class LlamaMLP(nn.Module):
    method __init__ (line 202) | def __init__(self, config):
    method forward (line 212) | def forward(self, x):
  function repeat_kv (line 231) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class LlamaAttention (line 243) | class LlamaAttention(nn.Module):
    method __init__ (line 246) | def __init__(self, config: LlamaConfig, layer_idx: Optional[int] = None):
    method _init_rope (line 276) | def _init_rope(self):
    method forward (line 303) | def forward(
  class LlamaRingFlashAttention2 (line 385) | class LlamaRingFlashAttention2(LlamaAttention):
    method __init__ (line 392) | def __init__(self, *args, **kwargs):
    method forward (line 400) | def forward(
    method _flash_attention_forward (line 478) | def _flash_attention_forward(self, query_states, key_states, value_sta...
    method _upad_input (line 535) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class LlamaFlashAttention2 (line 566) | class LlamaFlashAttention2(LlamaAttention):
    method __init__ (line 573) | def __init__(self, *args, **kwargs):
    method forward (line 581) | def forward(
    method _flash_attention_forward (line 659) | def _flash_attention_forward(self, query_states, key_states, value_sta...
    method _upad_input (line 712) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class LlamaSdpaAttention (line 743) | class LlamaSdpaAttention(LlamaAttention):
    method forward (line 751) | def forward(
  class LlamaDecoderLayer (line 841) | class LlamaDecoderLayer(nn.Module):
    method __init__ (line 842) | def __init__(self, config: LlamaConfig, layer_idx: int):
    method forward (line 852) | def forward(
  class LlamaPreTrainedModel (line 935) | class LlamaPreTrainedModel(PreTrainedModel):
    method _init_weights (line 945) | def _init_weights(self, module):
    method _setup_cache (line 956) | def _setup_cache(self, cache_cls, max_batch_size, max_cache_len: Optio...
    method _reset_cache (line 972) | def _reset_cache(self):
  class LlamaModel (line 1051) | class LlamaModel(LlamaPreTrainedModel):
    method __init__ (line 1059) | def __init__(self, config: LlamaConfig):
    method get_input_embeddings (line 1076) | def get_input_embeddings(self):
    method set_input_embeddings (line 1079) | def set_input_embeddings(self, value):
    method forward (line 1083) | def forward(
    method _update_causal_mask (line 1191) | def _update_causal_mask(self, attention_mask, input_tensor):
  class LlamaForCausalLM (line 1228) | class LlamaForCausalLM(LlamaPreTrainedModel):
    method __init__ (line 1231) | def __init__(self, config):
    method get_input_embeddings (line 1240) | def get_input_embeddings(self):
    method set_input_embeddings (line 1243) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1246) | def get_output_embeddings(self):
    method set_output_embeddings (line 1249) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1252) | def set_decoder(self, decoder):
    method get_decoder (line 1255) | def get_decoder(self):
    method forward (line 1260) | def forward(
    method prepare_inputs_for_generation (line 1351) | def prepare_inputs_for_generation(self, input_ids, past_key_values=Non...
    method _reorder_cache (line 1423) | def _reorder_cache(past_key_values, beam_idx):
  class LlamaForSequenceClassification (line 1445) | class LlamaForSequenceClassification(LlamaPreTrainedModel):
    method __init__ (line 1446) | def __init__(self, config):
    method get_input_embeddings (line 1455) | def get_input_embeddings(self):
    method set_input_embeddings (line 1458) | def set_input_embeddings(self, value):
    method forward (line 1462) | def forward(
  class LlamaForQuestionAnswering (line 1560) | class LlamaForQuestionAnswering(LlamaPreTrainedModel):
    method __init__ (line 1564) | def __init__(self, config):
    method get_input_embeddings (line 1572) | def get_input_embeddings(self):
    method set_input_embeddings (line 1575) | def set_input_embeddings(self, value):
    method forward (line 1579) | def forward(

FILE: xtuner-eval_niah/longva/model/llava_arch.py
  class LlavaMetaModel (line 34) | class LlavaMetaModel:
    method __init__ (line 36) | def __init__(self, config):
    method get_vision_tower (line 48) | def get_vision_tower(self):
    method initialize_vision_modules (line 54) | def initialize_vision_modules(self, model_args, fsdp=None):
  function unpad_image (line 119) | def unpad_image(tensor, original_size):
  class LlavaMetaForCausalLM (line 154) | class LlavaMetaForCausalLM(ABC):
    method get_model (line 157) | def get_model(self):
    method get_vision_tower (line 160) | def get_vision_tower(self):
    method get_2dPool (line 163) | def get_2dPool(self, image_feature):
    method encode_images (line 179) | def encode_images(self, images):
    method encode_multimodals (line 186) | def encode_multimodals(self, videos_or_images, video_idx_in_batch, spl...
    method prepare_inputs_labels_for_multimodal (line 199) | def prepare_inputs_labels_for_multimodal(self, input_ids, position_ids...
    method initialize_vision_tokenizer (line 409) | def initialize_vision_tokenizer(self, model_args, tokenizer):

FILE: xtuner-eval_niah/longva/model/make_delta.py
  function make_delta (line 14) | def make_delta(base_model_path, target_model_path, delta_path, hub_repo_...

FILE: xtuner-eval_niah/longva/model/multimodal_encoder/builder.py
  function build_vision_tower (line 9) | def build_vision_tower(vision_tower_cfg, **kwargs):

FILE: xtuner-eval_niah/longva/model/multimodal_encoder/clip_encoder.py
  class CLIPVisionTower (line 12) | class CLIPVisionTower(nn.Module):
    method __init__ (line 13) | def __init__(self, vision_tower, args, delay_load=False):
    method load_model (line 35) | def load_model(self, device_map=None):
    method feature_select (line 46) | def feature_select(self, image_forward_outs):
    method forward (line 68) | def forward(self, images):
    method dummy_feature (line 82) | def dummy_feature(self):
    method dtype (line 86) | def dtype(self):
    method device (line 90) | def device(self):
    method config (line 94) | def config(self):
    method hidden_size (line 101) | def hidden_size(self):
    method num_patches_per_side (line 110) | def num_patches_per_side(self):
    method num_patches (line 114) | def num_patches(self):
    method image_size (line 121) | def image_size(self):
  class CLIPVisionTowerS2 (line 125) | class CLIPVisionTowerS2(CLIPVisionTower):
    method __init__ (line 126) | def __init__(self, vision_tower, args, delay_load=False):
    method load_model (line 141) | def load_model(self, device_map=None):
    method forward_feature (line 156) | def forward_feature(self, images):
    method forward (line 162) | def forward(self, images):
    method hidden_size (line 174) | def hidden_size(self):

FILE: xtuner-eval_niah/longva/model/multimodal_projector/builder.py
  class IdentityMap (line 8) | class IdentityMap(nn.Module):
    method __init__ (line 9) | def __init__(self):
    method forward (line 12) | def forward(self, x, *args, **kwargs):
    method config (line 16) | def config(self):
  class SimpleResBlock (line 20) | class SimpleResBlock(nn.Module):
    method __init__ (line 21) | def __init__(self, channels):
    method forward (line 27) | def forward(self, x):
  function build_vision_projector (line 32) | def build_vision_projector(config, delay_load=False, **kwargs):

FILE: xtuner-eval_niah/longva/model/multimodal_projector/pooler_projector.py
  class PoolerProjector (line 9) | class PoolerProjector(nn.Module):
    method __init__ (line 10) | def __init__(self, config, vision_cfg):
    method forward (line 22) | def forward(self, x, *args, **kwargs):
    method config (line 32) | def config(self):

FILE: xtuner-eval_niah/longva/model/multimodal_resampler/builder.py
  class IdentityMap (line 9) | class IdentityMap(torch.nn.Module):
    method __init__ (line 10) | def __init__(self):
    method forward (line 13) | def forward(self, x, *args, **kwargs):
    method config (line 17) | def config(self):
  function build_vision_resampler (line 21) | def build_vision_resampler(model_args, delay_load=False, **kwargs):

FILE: xtuner-eval_niah/longva/model/multimodal_resampler/masked_drop.py
  class MaskedDrop (line 7) | class MaskedDrop(nn.Module):
    method __init__ (line 8) | def __init__(self, model_args):
    method forward (line 17) | def forward(self, image_features, *args, **kwargs):
    method config (line 46) | def config(self):
    method random_masking (line 56) | def random_masking(self, x, len_keep):

FILE: xtuner-eval_niah/longva/model/multimodal_resampler/perceiver.py
  function exists (line 16) | def exists(val):
  function FeedForward (line 20) | def FeedForward(dim, mult=4):
  class PerceiverAttention (line 30) | class PerceiverAttention(nn.Module):
    method __init__ (line 31) | def __init__(self, *, dim, dim_head=64, heads=8):
    method forward (line 44) | def forward(self, x, latents):
  class PerceiverResamplerModule (line 73) | class PerceiverResamplerModule(nn.Module):
    method __init__ (line 74) | def __init__(
    method forward (line 104) | def forward(self, x):
  class PerceiverResampler (line 130) | class PerceiverResampler(nn.Module):
    method __init__ (line 131) | def __init__(self, model_args, vision_tower):
    method forward (line 144) | def forward(self, image_features, *args, **kwargs):
    method config (line 148) | def config(self):

FILE: xtuner-eval_niah/longva/model/multimodal_resampler/qformer.py
  function disabled_train (line 51) | def disabled_train(self, mode=True):
  class BertEmbeddings (line 57) | class BertEmbeddings(nn.Module):
    method __init__ (line 60) | def __init__(self, config):
    method forward (line 76) | def forward(
  class BertSelfAttention (line 107) | class BertSelfAttention(nn.Module):
    method __init__ (line 108) | def __init__(self, config, is_cross_attention):
    method save_attn_gradients (line 133) | def save_attn_gradients(self, attn_gradients):
    method get_attn_gradients (line 136) | def get_attn_gradients(self):
    method save_attention_map (line 139) | def save_attention_map(self, attention_map):
    method get_attention_map (line 142) | def get_attention_map(self):
    method transpose_for_scores (line 145) | def transpose_for_scores(self, x):
    method forward (line 153) | def forward(
  class BertSelfOutput (line 239) | class BertSelfOutput(nn.Module):
    method __init__ (line 240) | def __init__(self, config):
    method forward (line 246) | def forward(self, hidden_states, input_tensor):
  class BertAttention (line 253) | class BertAttention(nn.Module):
    method __init__ (line 254) | def __init__(self, config, is_cross_attention=False):
    method prune_heads (line 260) | def prune_heads(self, heads):
    method forward (line 281) | def forward(
  class BertIntermediate (line 306) | class BertIntermediate(nn.Module):
    method __init__ (line 307) | def __init__(self, config):
    method forward (line 315) | def forward(self, hidden_states):
  class BertOutput (line 321) | class BertOutput(nn.Module):
    method __init__ (line 322) | def __init__(self, config):
    method forward (line 328) | def forward(self, hidden_states, input_tensor):
  class BertLayer (line 335) | class BertLayer(nn.Module):
    method __init__ (line 336) | def __init__(self, config, layer_num):
    method forward (line 354) | def forward(
    method feed_forward_chunk (line 422) | def feed_forward_chunk(self, attention_output):
    method feed_forward_chunk_query (line 427) | def feed_forward_chunk_query(self, attention_output):
  class BertEncoder (line 433) | class BertEncoder(nn.Module):
    method __init__ (line 434) | def __init__(self, config):
    method forward (line 439) | def forward(
  class BertPooler (line 530) | class BertPooler(nn.Module):
    method __init__ (line 531) | def __init__(self, config):
    method forward (line 536) | def forward(self, hidden_states):
  class BertPredictionHeadTransform (line 545) | class BertPredictionHeadTransform(nn.Module):
    method __init__ (line 546) | def __init__(self, config):
    method forward (line 555) | def forward(self, hidden_states):
  class BertLMPredictionHead (line 562) | class BertLMPredictionHead(nn.Module):
    method __init__ (line 563) | def __init__(self, config):
    method forward (line 576) | def forward(self, hidden_states):
  class BertOnlyMLMHead (line 582) | class BertOnlyMLMHead(nn.Module):
    method __init__ (line 583) | def __init__(self, config):
    method forward (line 587) | def forward(self, sequence_output):
  class BertPreTrainedModel (line 592) | class BertPreTrainedModel(PreTrainedModel):
    method _init_weights (line 602) | def _init_weights(self, module):
  class BertModel (line 615) | class BertModel(BertPreTrainedModel):
    method __init__ (line 625) | def __init__(self, config, add_pooling_layer=False):
    method get_input_embeddings (line 637) | def get_input_embeddings(self):
    method set_input_embeddings (line 640) | def set_input_embeddings(self, value):
    method _prune_heads (line 643) | def _prune_heads(self, heads_to_prune):
    method get_extended_attention_mask (line 651) | def get_extended_attention_mask(
    method forward (line 731) | def forward(
  class BertLMHeadModel (line 865) | class BertLMHeadModel(BertPreTrainedModel):
    method __init__ (line 870) | def __init__(self, config):
    method get_output_embeddings (line 878) | def get_output_embeddings(self):
    method set_output_embeddings (line 881) | def set_output_embeddings(self, new_embeddings):
    method forward (line 884) | def forward(
    method prepare_inputs_for_generation (line 992) | def prepare_inputs_for_generation(self, input_ids, query_embeds, past=...
    method _reorder_cache (line 1013) | def _reorder_cache(self, past, beam_idx):
  class BertForMaskedLM (line 1020) | class BertForMaskedLM(BertPreTrainedModel):
    method __init__ (line 1025) | def __init__(self, config):
    method get_output_embeddings (line 1033) | def get_output_embeddings(self):
    method set_output_embeddings (line 1036) | def set_output_embeddings(self, new_embeddings):
    method forward (line 1039) | def forward(
  class Qformer (line 1102) | class Qformer(nn.Module):
    method __init__ (line 1103) | def __init__(self, model_args, vision_tower):
    method build_Qformer (line 1117) | def build_Qformer(self, vision_width, cross_attention_freq, num_query_...
    method forward (line 1135) | def forward(self, image_features, *args, **kwargs):
    method hidden_size (line 1150) | def hidden_size(self):
    method config (line 1154) | def config(self):

FILE: xtuner-eval_niah/longva/model/multimodal_resampler/spatial_pool.py
  class SpatialPool (line 6) | class SpatialPool(nn.Module):
    method __init__ (line 7) | def __init__(self, model_args, vision_tower):
    method forward (line 23) | def forward(self, image_features, images, *args, **kwargs):
    method config (line 35) | def config(self):
    method hidden_size (line 44) | def hidden_size(self):

FILE: xtuner-eval_niah/longva/model/utils.py
  function auto_upgrade (line 4) | def auto_upgrade(config):

FILE: xtuner-eval_niah/longva/train/llama_flash_attn_monkey_patch.py
  function forward (line 16) | def forward(
  function _prepare_decoder_attention_mask (line 77) | def _prepare_decoder_attention_mask(self, attention_mask, input_shape, i...
  function replace_llama_attn_with_flash_attn (line 82) | def replace_llama_attn_with_flash_attn():

FILE: xtuner-eval_niah/longva/train/llava_trainer.py
  function maybe_zero_3 (line 30) | def maybe_zero_3(param, ignore_status=False, name=None):
  function get_mm_adapter_state_maybe_zero_3 (line 45) | def get_mm_adapter_state_maybe_zero_3(named_params, keys_to_match):
  function split_to_even_chunks (line 51) | def split_to_even_chunks(indices, lengths, num_chunks):
  function get_variable_length_grouped_indices (line 73) | def get_variable_length_grouped_indices(lengths, batch_size, world_size,...
  function get_modality_length_grouped_indices (line 89) | def get_modality_length_grouped_indices(lengths, batch_size, world_size,...
  function get_length_grouped_indices (line 129) | def get_length_grouped_indices(lengths, batch_size, world_size, generato...
  function get_length_grouped_indices_auto_single (line 152) | def get_length_grouped_indices_auto_single(lengths, batch_size, world_si...
  function get_modality_length_grouped_indices_auto (line 167) | def get_modality_length_grouped_indices_auto(lengths, batch_size, world_...
  class LengthGroupedSampler (line 196) | class LengthGroupedSampler(Sampler):
    method __init__ (line 202) | def __init__(
    method __len__ (line 223) | def __len__(self):
    method __iter__ (line 226) | def __iter__(self):
  class LLaVATrainer (line 240) | class LLaVATrainer(Trainer):
    method create_accelerator_and_postprocess (line 242) | def create_accelerator_and_postprocess(self):
    method _get_train_sampler (line 273) | def _get_train_sampler(self) -> Optional[torch.utils.data.Sampler]:
    method get_train_dataloader (line 319) | def get_train_dataloader(self) -> DataLoader:
    method create_optimizer (line 356) | def create_optimizer(self):
    method _save_checkpoint (line 435) | def _save_checkpoint(self, model, trial, metrics=None):
    method _save (line 459) | def _save(self, output_dir: Optional[str] = None, state_dict=None):
  class LLaVADPOTrainer (line 466) | class LLaVADPOTrainer(DPOTrainer):
    method _get_train_sampler (line 467) | def _get_train_sampler(self) -> Optional[torch.utils.data.Sampler]:
    method _save_checkpoint (line 483) | def _save_checkpoint(self, model, trial, metrics=None):
    method _save (line 523) | def _save(self, output_dir: Optional[str] = None, state_dict=None):

FILE: xtuner-eval_niah/longva/train/train.py
  class ModelArguments (line 59) | class ModelArguments:
  class DataArguments (line 112) | class DataArguments:
  class TrainingArguments (line 128) | class TrainingArguments(transformers.TrainingArguments):
  function maybe_zero_3 (line 177) | def maybe_zero_3(param, ignore_status=False, name=None):
  function get_peft_state_maybe_zero_3 (line 193) | def get_peft_state_maybe_zero_3(named_params, bias):
  function get_peft_state_non_lora_maybe_zero_3 (line 218) | def get_peft_state_non_lora_maybe_zero_3(named_params, require_grad_only...
  function get_mm_adapter_state_maybe_zero_3 (line 226) | def get_mm_adapter_state_maybe_zero_3(named_params, keys_to_match):
  function find_all_linear_names (line 232) | def find_all_linear_names(model):
  function safe_save_model_for_hf_trainer (line 248) | def safe_save_model_for_hf_trainer(trainer: transformers.Trainer, output...
  function smart_tokenizer_and_embedding_resize (line 292) | def smart_tokenizer_and_embedding_resize(
  function _tokenize_fn (line 315) | def _tokenize_fn(strings: Sequence[str], tokenizer: transformers.PreTrai...
  function _mask_targets (line 337) | def _mask_targets(target, tokenized_lens, speakers):
  function _add_speaker_and_signal (line 348) | def _add_speaker_and_signal(header, source, get_conversation=True):
  function preprocess_multimodal (line 368) | def preprocess_multimodal(sources: Sequence[str], data_args: DataArgumen...
  function preprocess_llama_2 (line 396) | def preprocess_llama_2(sources, tokenizer: transformers.PreTrainedTokeni...
  function preprocess_gemma (line 471) | def preprocess_gemma(sources: List[List[Dict[str, str]]], tokenizer: tra...
  function preprocess_qwen (line 550) | def preprocess_qwen(sources, tokenizer: transformers.PreTrainedTokenizer...
  function preprocess_llama3 (line 605) | def preprocess_llama3(
  function preprocess_v1 (line 687) | def preprocess_v1(sources, tokenizer: transformers.PreTrainedTokenizer, ...
  function preprocess_mpt (line 766) | def preprocess_mpt(sources, tokenizer: transformers.PreTrainedTokenizer,...
  function preprocess_plain (line 847) | def preprocess_plain(
  function preprocess (line 869) | def preprocess(sources: Sequence[str], tokenizer: transformers.PreTraine...
  class LazySupervisedDataset (line 920) | class LazySupervisedDataset(Dataset):
    method __init__ (line 921) | def __init__(self, data_path: str, tokenizer: transformers.PreTrainedT...
    method __len__ (line 992) | def __len__(self):
    method lengths (line 996) | def lengths(self):
    method modality_lengths (line 1004) | def modality_lengths(self):
    method process_image (line 1012) | def process_image(self, image_file):
    method __getitem__ (line 1050) | def __getitem__(self, i) -> Dict[str, torch.Tensor]:
    method _get_item (line 1083) | def _get_item(self, i) -> Dict[str, torch.Tensor]:
  class DataCollatorForSupervisedDataset (line 1184) | class DataCollatorForSupervisedDataset(object):
    method pad_sequence (line 1189) | def pad_sequence(self, input_ids, batch_first, padding_value):
    method __call__ (line 1197) | def __call__(self, instances: Sequence[Dict]) -> Dict[str, torch.Tensor]:
  function make_supervised_data_module (line 1230) | def make_supervised_data_module(tokenizer: transformers.PreTrainedTokeni...
  function get_model (line 1237) | def get_model(model_args, training_args, bnb_model_from_pretrained_args):
  function train (line 1387) | def train(attn_implementation=None):

FILE: xtuner-eval_niah/longva/train/train_dpo.py
  class ModelArguments (line 67) | class ModelArguments:
  class DataArguments (line 117) | class DataArguments:
  class TrainingArguments (line 135) | class TrainingArguments(transformers.TrainingArguments):
  function maybe_zero_3 (line 171) | def maybe_zero_3(param, ignore_status=False, name=None):
  function get_peft_state_maybe_zero_3 (line 187) | def get_peft_state_maybe_zero_3(named_params, bias):
  function get_peft_state_non_lora_maybe_zero_3 (line 212) | def get_peft_state_non_lora_maybe_zero_3(named_params, require_grad_only...
  function get_mm_adapter_state_maybe_zero_3 (line 220) | def get_mm_adapter_state_maybe_zero_3(named_params, keys_to_match):
  function find_all_linear_names (line 226) | def find_all_linear_names(model):
  function safe_save_model_for_hf_trainer (line 242) | def safe_save_model_for_hf_trainer(trainer: transformers.Trainer, output...
  function smart_tokenizer_and_embedding_resize (line 286) | def smart_tokenizer_and_embedding_resize(
  function _tokenize_fn (line 309) | def _tokenize_fn(strings: Sequence[str], tokenizer: transformers.PreTrai...
  function _mask_targets (line 331) | def _mask_targets(target, tokenized_lens, speakers):
  function _add_speaker_and_signal (line 342) | def _add_speaker_and_signal(header, source, get_conversation=True):
  function preprocess_multimodal (line 362) | def preprocess_multimodal(sources: Sequence[str], data_args: DataArgumen...
  function preprocess_multimodal_movie (line 383) | def preprocess_multimodal_movie(sources: Sequence[str], data_args: DataA...
  function preprocess_llama_2 (line 400) | def preprocess_llama_2(sources, tokenizer: transformers.PreTrainedTokeni...
  function make_conv (line 475) | def make_conv(prompt, answer):
  function preprocess_gemma (line 488) | def preprocess_gemma(sources: List[List[Dict[str, str]]], tokenizer: tra...
  function preprocess_qwen (line 567) | def preprocess_qwen(sources, tokenizer: transformers.PreTrainedTokenizer...
  function preprocess_llama3 (line 618) | def preprocess_llama3(
  function preprocess_v1 (line 667) | def preprocess_v1(sources, tokenizer: transformers.PreTrainedTokenizer, ...
  function preprocess_mpt (line 746) | def preprocess_mpt(sources, tokenizer: transformers.PreTrainedTokenizer,...
  function preprocess_plain (line 827) | def preprocess_plain(
  function preprocess (line 849) | def preprocess(sources: Sequence[str], tokenizer: transformers.PreTraine...
  function load_data (line 900) | def load_data(data_path):
  class DPODataset (line 908) | class DPODataset(Dataset):
    method __init__ (line 911) | def __init__(self, data_path: str, tokenizer: transformers.PreTrainedT...
    method __len__ (line 978) | def __len__(self):
    method lengths (line 982) | def lengths(self):
    method modality_lengths (line 993) | def modality_lengths(self):
    method process_image (line 1003) | def process_image(self, image_file):
    method __getitem__ (line 1041) | def __getitem__(self, i) -> Dict[str, torch.Tensor]:
    method _get_item (line 1081) | def _get_item(self, i) -> Dict[str, torch.Tensor]:
  class DPODataCollator (line 1187) | class DPODataCollator(DPODataCollatorWithPadding):
    method collate (line 1192) | def collate(self, batch):
    method tokenize_batch_element (line 1239) | def tokenize_batch_element(self, prompt: str, chosen: str, rejected: s...
    method __call__ (line 1274) | def __call__(self, features: List[Dict[str, Any]]) -> Dict[str, Any]:
  function make_dpo_data_module (line 1308) | def make_dpo_data_module(tokenizer: transformers.PreTrainedTokenizer, da...
  function get_model (line 1314) | def get_model(model_args, training_args, bnb_model_from_pretrained_args):
  function train (line 1484) | def train(attn_implementation=None):

FILE: xtuner-eval_niah/longva/utils.py
  function process_video_with_pyav (line 25) | def process_video_with_pyav(video_file, data_args):
  function rank0_print (line 49) | def rank0_print(*args):
  function build_logger (line 57) | def build_logger(logger_name, logger_filename):
  class StreamToLogger (line 99) | class StreamToLogger(object):
    method __init__ (line 104) | def __init__(self, logger, log_level=logging.INFO):
    method __getattr__ (line 110) | def __getattr__(self, attr):
    method write (line 113) | def write(self, buf):
    method flush (line 127) | def flush(self):
  function disable_torch_init (line 133) | def disable_torch_init():
  function violates_moderation (line 143) | def violates_moderation(text):
  function pretty_print_semaphore (line 165) | def pretty_print_semaphore(semaphore):

FILE: xtuner-eval_niah/vision_niah/multi_eval_vision_niah.py
  function safe_tokenize (line 67) | def safe_tokenize(tokenizer, text):
  function eval_forward (line 74) | def eval_forward(model, input_embeds, answer_embeds, pad_id, answer_ids,...
  function load_haystack (line 134) | def load_haystack(args):
  function load_text_embeddings (line 143) | def load_text_embeddings(str, tokenizer, model, replace_double_newline=F...
  function inference (line 164) | def inference(args):
  function plot (line 318) | def plot(args,  all_accuries):
  function main (line 380) | def main(args):

FILE: xtuner-eval_niah/vision_niah/multi_produce_needle_embedding.py
  function main (line 19) | def main(args):

FILE: xtuner-eval_niah/vision_niah/produce_haystack_embedding.py
  function load_video_batches (line 12) | def load_video_batches(video_path, batch_size):
  function main (line 23) | def main(args):

FILE: xtuner-eval_niah/vision_niah/single_eval_vision_niah.py
  function safe_tokenize (line 67) | def safe_tokenize(tokenizer, text):
  function eval_forward (line 74) | def eval_forward( model, input_embeds, answer_embeds, pad_id, answer_ids...
  function load_haystack (line 132) | def load_haystack(args):
  function load_text_embeddings (line 141) | def load_text_embeddings(str, tokenizer, model, replace_double_newline=F...
  function inference (line 162) | def inference(args):
  function plot (line 275) | def plot(args,  all_accuries):
  function main (line 337) | def main(args):

FILE: xtuner-eval_niah/vision_niah/single_produce_needle_embedding.py
  function main (line 18) | def main(args):

FILE: xtuner-eval_niah/xtuner/_lite/__init__.py
  function get_logger (line 11) | def get_logger():

FILE: xtuner-eval_niah/xtuner/_lite/accelerate/dispatches/__init__.py
  function _dispatch_forward_fn (line 9) | def _dispatch_forward_fn(module, dispatch_fn):
  function dispatch_internlm_varlen_attn_forward (line 13) | def dispatch_internlm_varlen_attn_forward(module):
  function dispatch_llama_varlen_attn_forward (line 20) | def dispatch_llama_varlen_attn_forward(module):
  function dispatch_qwen2_varlen_attn_forward (line 27) | def dispatch_qwen2_varlen_attn_forward(module):
  function dispatch_clip_attn_forward (line 34) | def dispatch_clip_attn_forward(module):
  function dispatch_rms_norm_forward (line 41) | def dispatch_rms_norm_forward(module):
  function dispatch_internlm_mla_varlen_attn_forward (line 47) | def dispatch_internlm_mla_varlen_attn_forward(module):
  function dispatch_modules (line 66) | def dispatch_modules(model, use_varlen_attn=False):

FILE: xtuner-eval_niah/xtuner/_lite/accelerate/dispatches/_attention.py
  function _get_unpad_data (line 17) | def _get_unpad_data(attention_mask):
  function upad_qkv (line 30) | def upad_qkv(query_layer, key_layer, value_layer, attention_mask,
  function flash_attn_wo_mask (line 75) | def flash_attn_wo_mask(
  function flash_attn_w_mask (line 96) | def flash_attn_w_mask(
  function varlen_flash_attn (line 128) | def varlen_flash_attn(

FILE: xtuner-eval_niah/xtuner/_lite/accelerate/dispatches/_fused/layer_norm.py
  function layer_norm_forward (line 6) | def layer_norm_forward(self, hidden_states):

FILE: xtuner-eval_niah/xtuner/_lite/accelerate/dispatches/_fused/rms_norm.py
  function rms_norm_forward (line 14) | def rms_norm_forward(self, hidden_states):

FILE: xtuner-eval_niah/xtuner/_lite/accelerate/dispatches/_fused/rotary.py
  function rotary_kernel (line 11) | def rotary_kernel(
  function apply_rotary (line 163) | def apply_rotary(
  class ApplyRotaryEmb (line 265) | class ApplyRotaryEmb(torch.autograd.Function):
    method forward (line 268) | def forward(
    method backward (line 302) | def backward(ctx, do):

FILE: xtuner-eval_niah/xtuner/_lite/accelerate/dispatches/clip.py
  function clip_flash_attn_forward (line 11) | def clip_flash_attn_forward(

FILE: xtuner-eval_niah/xtuner/_lite/accelerate/dispatches/internlm2.py
  class InternLM2RotaryEmbedding (line 22) | class InternLM2RotaryEmbedding(torch.nn.Module):
    method __init__ (line 24) | def __init__(self,
    method forward (line 47) | def forward(self, x, seq_len):
  function rotate_half (line 68) | def rotate_half(x):
  function apply_rotary_pos_emb (line 75) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  function apply_rotary_pos_emb_mla (line 83) | def apply_rotary_pos_emb_mla(q, k, cos, sin, position_ids=None, unsqueez...
  function repeat_kv (line 113) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  function repeat_kv_bshd (line 131) | def repeat_kv_bshd(hidden_states: torch.Tensor, n_rep: int) -> torch.Ten...
  function _internlm2_varlen_attn_forward (line 145) | def _internlm2_varlen_attn_forward(
  function internlm2_attn_forward_inference (line 324) | def internlm2_attn_forward_inference(
  function internlm2_varlen_attn_forward (line 451) | def internlm2_varlen_attn_forward(
  function _internlm2_mla_varlen_attn_forward (line 478) | def _internlm2_mla_varlen_attn_forward(
  function internlm2_mla_varlen_attn_forward (line 596) | def internlm2_mla_varlen_attn_forward(

FILE: xtuner-eval_niah/xtuner/_lite/accelerate/dispatches/llama.py
  class Cache (line 15) | class Cache:
  function repeat_kv_bshd (line 19) | def repeat_kv_bshd(hidden_states: torch.Tensor, n_rep: int) -> torch.Ten...
  function llama_varlen_attn_forward (line 33) | def llama_varlen_attn_forward(

FILE: xtuner-eval_niah/xtuner/_lite/accelerate/dispatches/qwen2.py
  function qwen2_attn_forward_inference (line 32) | def qwen2_attn_forward_inference(
  function qwen2_varlen_attn_forward (line 148) | def qwen2_varlen_attn_forward(

FILE: xtuner-eval_niah/xtuner/_lite/accelerate/packed.py
  function packed_sequence (line 9) | def packed_sequence(num_tokens, enable=False, sp_size=1):

FILE: xtuner-eval_niah/xtuner/_lite/auto.py
  function download_model_from_hub (line 20) | def download_model_from_hub(
  class AutoModelForCausalLM (line 64) | class AutoModelForCausalLM:
    method from_config (line 91) | def from_config(cls,
    method from_pretrained (line 102) | def from_pretrained(
    method _flash_attn_kwargs (line 139) | def _flash_attn_kwargs(config: PretrainedConfig) -> dict:
    method _long_ctx_kwargs (line 162) | def _long_ctx_kwargs(config: PretrainedConfig,

FILE: xtuner-eval_niah/xtuner/_lite/chat/messages/base.py
  class BaseMessages (line 10) | class BaseMessages(BaseModel):
    method add (line 13) | def add(self, role: str, content):
    method pop (line 17) | def pop(self):
    method get_prompt (line 21) | def get_prompt(self, chat_template: ChatTemplate) -> str:
    method tokenize (line 25) | def tokenize(self, tokenizer: PreTrainedTokenizer,
    method from_dict (line 30) | def from_dict(cls, item: Dict) -> 'BaseMessages':

FILE: xtuner-eval_niah/xtuner/_lite/chat/messages/chat.py
  class TextContentItem (line 15) | class TextContentItem(BaseModel):
    method apply_chat_template (line 19) | def apply_chat_template(self, chat_template: HybridChatTemplate) -> str:
  class ImageContentItem (line 23) | class ImageContentItem(BaseModel):
    method apply_chat_template (line 27) | def apply_chat_template(self, chat_template: HybridChatTemplate) -> str:
  class ChatMsg (line 35) | class ChatMsg(BaseModel):
    method __init__ (line 41) | def __init__(self, *args, **kwargs):
    method collect_img_urls (line 53) | def collect_img_urls(self) -> List[str]:
    method get_prompt (line 61) | def get_prompt(self, chat_template: ChatTemplate) -> str:
    method tokenize (line 86) | def tokenize(
  class ChatMessages (line 107) | class ChatMessages(BaseMessages):
    method add (line 111) | def add(self, role, content, loss=False):
    method pop (line 114) | def pop(self):
    method get_prompt (line 117) | def get_prompt(self, chat_template: ChatTemplate) -> str:
    method tokenize (line 127) | def tokenize(self, tokenizer: PreTrainedTokenizer,
    method from_str (line 169) | def from_str(cls, prompt: str) -> 'ChatMessages':
    method from_dict (line 175) | def from_dict(cls, item: dict) -> 'ChatMessages':

FILE: xtuner-eval_niah/xtuner/_lite/chat/templates/chat.py
  class ChatTemplate (line 6) | class ChatTemplate(BaseModel):
    method decorate_system (line 18) | def decorate_system(self, text: str) -> str:
    method decorate_assistant (line 22) | def decorate_assistant(self, text: str) -> str:
    method decorate_user (line 26) | def decorate_user(self, text: str) -> str:
    method check_system (line 31) | def check_system(cls, v: str) -> str:
    method check_user (line 41) | def check_user(cls, v: str) -> str:
    method check_assistant (line 51) | def check_assistant(cls, v: str) -> str:

FILE: xtuner-eval_niah/xtuner/_lite/chat/templates/hybrid.py
  class HybridChatTemplate (line 6) | class HybridChatTemplate(BaseModel):
    method mm_token_maps (line 42) | def mm_token_maps(self) -> Dict[str, int]:
    method decorate_system (line 47) | def decorate_system(self, text: str) -> str:
    method decorate_assistant (line 51) | def decorate_assistant(self, text: str) -> str:
    method decorate_user (line 55) | def decorate_user(self, text: str) -> str:
    method decorate_files (line 59) | def decorate_files(self, text: str) -> str:
    method decorate_functions (line 63) | def decorate_functions(self, text: str) -> str:
    method decorate_function_call (line 67) | def decorate_function_call(self, text: str, func: str) -> str:
    method decorate_function_result (line 71) | def decorate_function_result(self, text: str) -> str:
    method decorate_code_interpreter (line 75) | def decorate_code_interpreter(self, text: str) -> str:
    method decorate_code_interpreter_call (line 79) | def decorate_code_interpreter_call(self, text: str, func: str) -> str:
    method decorate_code_interpreter_result (line 84) | def decorate_code_interpreter_result(self, text: str) -> str:
    method check_system (line 90) | def check_system(cls, v: str) -> str:
    method check_user (line 100) | def check_user(cls, v: str) -> str:
    method check_assistant (line 110) | def check_assistant(cls, v: str) -> str:
    method check_function_call (line 121) | def check_function_call(cls, v: str) -> str:
    method check_function_result (line 137) | def check_function_result(cls, v: str) -> str:
    method check_functions (line 148) | def check_functions(cls, v: str) -> str:
    method check_code_interpreter (line 159) | def check_code_interpreter(cls, v: str) -> str:
    method check_code_interpreter_call (line 170) | def check_code_interpreter_call(cls, v: str) -> str:
    method check_code_interpreter_result (line 186) | def check_code_interpreter_result(cls, v: str) -> str:

FILE: xtuner-eval_niah/xtuner/_lite/datasets/cache.py
  class CacheDataset (line 4) | class CacheDataset(torch.utils.data.Dataset):
    method cached_dir (line 7) | def cached_dir(self):
    method cached (line 11) | def cached(self):
    method cache (line 14) | def cache(self, cache_dir):
    method load_cache (line 17) | def load_cache(self):
    method from_cache (line 21) | def from_cache(self, cache_dir):

FILE: xtuner-eval_niah/xtuner/_lite/datasets/format.py
  class Alpaca2Openai (line 4) | class Alpaca2Openai():
    method source_format (line 7) | def source_format(cls):
    method target_format (line 16) | def target_format(cls):
    method convert (line 32) | def convert(data):
  function llava_to_openai (line 50) | def llava_to_openai(data):
  function llava_to_openai_interleave (line 101) | def llava_to_openai_interleave(data):

FILE: xtuner-eval_niah/xtuner/_lite/datasets/llava.py
  class LlavaTokenizeFunction (line 16) | class LlavaTokenizeFunction():
    method __init__ (line 18) | def __init__(self,
    method __call__ (line 31) | def __call__(self, item):
  class LlavaTokenizedDataset (line 58) | class LlavaTokenizedDataset(TextTokenizedDataset):
    method __init__ (line 60) | def __init__(self, dataset, image_processor, max_length):
    method process_tokenized_data (line 64) | def process_tokenized_data(self, tokenized_data):
    method from_cache (line 87) | def from_cache(cls, cache_dir, image_processor, max_length):
    method __getitem__ (line 93) | def __getitem__(self, item):
  class LlavaRawDataset (line 112) | class LlavaRawDataset(torch.utils.data.Dataset):
    method __init__ (line 114) | def __init__(self, dataset, image_processor, max_length, tokenize_fn):
    method process_tokenized_data (line 122) | def process_tokenized_data(self, tokenized_data):
    method __getitem__ (line 144) | def __getitem__(self, item):
  class SoftPackerForLlava (line 151) | class SoftPackerForLlava(SoftPackerForText):
    method __init__ (line 153) | def __init__(self,
    method __getitem__ (line 161) | def __getitem__(self, item):
    method from_cache (line 230) | def from_cache(cls, cache_dir, image_processor, max_length):
  class LlavaCollator (line 245) | class LlavaCollator():
    method __init__ (line 247) | def __init__(self, pack_batch=False):
    method __call__ (line 250) | def __call__(self, instances):

FILE: xtuner-eval_niah/xtuner/_lite/datasets/load.py
  function load_json (line 22) | def load_json(file):
  function load_jsonl (line 28) | def load_jsonl(file):
  function load_bin (line 39) | def load_bin(file):
  function master_only_load (line 46) | def master_only_load(load_fn):
  function multi_thread_map (line 75) | def multi_thread_map(map_fns, dataset, desc, num_proc=8):
  function openai_format (line 95) | def openai_format(item):
  function load_hf_dataset (line 102) | def load_hf_dataset(path,
  function load_from_cache (line 133) | def load_from_cache(cache_dir, init_fn):
  function _gpu_parallel_load_local_datasets (line 174) | def _gpu_parallel_load_local_datasets(paths,
  function _cpu_parallel_load_local_datasets (line 361) | def _cpu_parallel_load_local_datasets(paths,
  function load_local_datasets (line 579) | def load_local_datasets(paths,
  function load_datasets (line 601) | def load_datasets(paths,
  function load_ms_dataset (line 721) | def load_ms_dataset():

FILE: xtuner-eval_niah/xtuner/_lite/datasets/pretrain.py
  class SoftPackerForPretrain (line 7) | class SoftPackerForPretrain(SoftPackerForText):
    method __getitem__ (line 9) | def __getitem__(self, item):

FILE: xtuner-eval_niah/xtuner/_lite/datasets/text.py
  function sort_and_return_indices (line 21) | def sort_and_return_indices(lst):
  class TextTokenizeFunction (line 25) | class TextTokenizeFunction():
    method __init__ (line 27) | def __init__(self, tokenizer, chat_template, raw_format='openai'):
    method __call__ (line 33) | def __call__(self, item):
  class TextTokenizedDataset (line 41) | class TextTokenizedDataset(CacheDataset):
    method __init__ (line 43) | def __init__(self, dataset, max_length):
    method cached (line 58) | def cached(self):
    method cached_dir (line 62) | def cached_dir(self):
    method cache (line 65) | def cache(self, cache_dir):
    method load_cache (line 76) | def load_cache(self):
    method from_cache (line 82) | def from_cache(cls, cache_dir, max_length):
    method _free (line 88) | def _free(self):
    method __len__ (line 91) | def __len__(self):
    method __getitem__ (line 94) | def __getitem__(self, item):
  class TextOnlineTokenizeDataset (line 119) | class TextOnlineTokenizeDataset(torch.utils.data.Dataset):
    method __init__ (line 121) | def __init__(self, dataset, tokenize_fn):
    method __len__ (line 127) | def __len__(self):
    method __getitem__ (line 130) | def __getitem__(self, item):
  class SoftPackerForText (line 151) | class SoftPackerForText(CacheDataset):
    method __init__ (line 153) | def __init__(self, dataset, max_length=2048, pack_info=None, seed=None):
    method max_length_per_pack (line 176) | def max_length_per_pack(self):
    method cached (line 184) | def cached(self):
    method cached_dir (line 188) | def cached_dir(self):
    method cache (line 191) | def cache(self, cache_dir):
    method load_cache (line 208) | def load_cache(self):
    method _free (line 216) | def _free(self):
    method __len__ (line 220) | def __len__(self):
    method __getitem__ (line 223) | def __getitem__(self, item):
    method get_pack_info (line 286) | def get_pack_info(cls, dataset, max_length, seed=None):
    method from_cache (line 326) | def from_cache(cls, cache_dir, max_length, seed=None):
  class HardPackerForText (line 342) | class HardPackerForText(SoftPackerForText):
    method __init__ (line 361) | def __init__(self, dataset, max_length=2048, pack_info=None):
    method _cal_max_length (line 367) | def _cal_max_length(cls, begin, end, shfl_item_rngs_left,
    method get_pack_info (line 382) | def get_pack_info(cls, dataset, max_length, seed=None):
    method _pack_ids_and_labels_in_range (line 426) | def _pack_ids_and_labels_in_range(self, begin: int, end: int):
    method __len__ (line 471) | def __len__(self):
    method __getitem__ (line 474) | def __getitem__(self, item):
  class TextCollator (line 502) | class TextCollator:
    method __init__ (line 504) | def __init__(self, pack_batch=False, force_div_ring=False, ring_size=1...
    method __call__ (line 512) | def __call__(self, instances):

FILE: xtuner-eval_niah/xtuner/_lite/modelings/__init__.py
  function register_remote_code (line 6) | def register_remote_code():

FILE: xtuner-eval_niah/xtuner/_lite/modelings/internlm2/configuration_internlm2.py
  class InternLM2Config (line 27) | class InternLM2Config(PretrainedConfig):
    method __init__ (line 94) | def __init__(  # pylint: disable=W0102
    method _rope_scaling_validation (line 150) | def _rope_scaling_validation(self):

FILE: xtuner-eval_niah/xtuner/_lite/modelings/internlm2/modeling_internlm2.py
  function _get_unpad_data (line 62) | def _get_unpad_data(attention_mask):
  class InternLM2RMSNorm (line 75) | class InternLM2RMSNorm(nn.Module):
    method __init__ (line 78) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 83) | def forward(self, hidden_states):
  class InternLM2RotaryEmbedding (line 95) | class InternLM2RotaryEmbedding(nn.Module):
    method __init__ (line 98) | def __init__(self,
    method forward (line 118) | def forward(self, x, position_ids):
  class InternLM2LinearScalingRotaryEmbedding (line 137) | class InternLM2LinearScalingRotaryEmbedding(InternLM2RotaryEmbedding):
    method forward (line 140) | def forward(self, x, position_ids):
  class InternLM2DynamicNTKScalingRotaryEmbedding (line 147) | class InternLM2DynamicNTKScalingRotaryEmbedding(InternLM2RotaryEmbedding):
    method forward (line 151) | def forward(self, x, position_ids):
  function rotate_half (line 171) | def rotate_half(x):
  function apply_rotary_pos_emb (line 178) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  class InternLM2MLP (line 204) | class InternLM2MLP(nn.Module):
    method __init__ (line 207) | def __init__(self, config):
    method forward (line 220) | def forward(self, x):
  function repeat_kv (line 226) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class InternLM2Attention (line 242) | class InternLM2Attention(nn.Module):
    method __init__ (line 245) | def __init__(self,
    method _init_rope (line 281) | def _init_rope(self):
    method forward (line 308) | def forward(
  class InternLM2FlashAttention2 (line 405) | class InternLM2FlashAttention2(InternLM2Attention):
    method __init__ (line 412) | def __init__(self, *args, **kwargs):
    method forward (line 424) | def forward(
    method _flash_attention_forward (line 529) | def _flash_attention_forward(self,
    method _upad_input (line 598) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class InternLM2SdpaAttention (line 641) | class InternLM2SdpaAttention(InternLM2Attention):
    method forward (line 649) | def forward(
  class InternLM2DecoderLayer (line 757) | class InternLM2DecoderLayer(nn.Module):
    method __init__ (line 760) | def __init__(self, config: InternLM2Config, layer_idx: int):
    method forward (line 775) | def forward(
  class InternLM2PreTrainedModel (line 853) | class InternLM2PreTrainedModel(PreTrainedModel):
    method _init_weights (line 869) | def _init_weights(self, module):
  class InternLM2Model (line 948) | class InternLM2Model(InternLM2PreTrainedModel):
    method __init__ (line 957) | def __init__(self, config: InternLM2Config):
    method get_input_embeddings (line 978) | def get_input_embeddings(self):
    method set_input_embeddings (line 981) | def set_input_embeddings(self, value):
    method forward (line 985) | def forward(
    method _update_causal_mask (line 1106) | def _update_causal_mask(
  class InternLM2ForCausalLM (line 1199) | class InternLM2ForCausalLM(InternLM2PreTrainedModel):
    method __init__ (line 1205) | def __init__(self, config):
    method get_input_embeddings (line 1215) | def get_input_embeddings(self):
    method set_input_embeddings (line 1218) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1221) | def get_output_embeddings(self):
    method set_output_embeddings (line 1224) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1227) | def set_decoder(self, decoder):
    method get_decoder (line 1230) | def get_decoder(self):
    method forward (line 1236) | def forward(
    method prepare_inputs_for_generation (line 1329) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1413) | def _reorder_cache(past_key_values, beam_idx):
    method build_inputs (line 1421) | def build_inputs(self,
    method chat (line 1440) | def chat(
    method stream_chat (line 1487) | def stream_chat(
  class InternLM2ForSequenceClassification (line 1594) | class InternLM2ForSequenceClassification(InternLM2PreTrainedModel):
    method __init__ (line 1597) | def __init__(self, config):
    method get_input_embeddings (line 1606) | def get_input_embeddings(self):
    method set_input_embeddings (line 1609) | def set_input_embeddings(self, value):
    method forward (line 1613) | def forward(
  class InternLM2ForQuestionAnswering (line 1719) | class InternLM2ForQuestionAnswering(InternLM2PreTrainedModel):
    method __init__ (line 1724) | def __init__(self, config):
    method get_input_embeddings (line 1732) | def get_input_embeddings(self):
    method set_input_embeddings (line 1735) | def set_input_embeddings(self, value):
    method forward (line 1739) | def forward(
  class InternLM2ForTokenClassification (line 1823) | class InternLM2ForTokenClassification(InternLM2PreTrainedModel):
    method __init__ (line 1826) | def __init__(self, config):
    method get_input_embeddings (line 1842) | def get_input_embeddings(self):
    method set_input_embeddings (line 1845) | def set_input_embeddings(self, value):
    method forward (line 1849) | def forward(

FILE: xtuner-eval_niah/xtuner/_lite/modelings/llava/configuration_internlm2.py
  class InternLM2Config (line 27) | class InternLM2Config(PretrainedConfig):
    method __init__ (line 94) | def __init__(  # pylint: disable=W0102
    method _rope_scaling_validation (line 150) | def _rope_scaling_validation(self):

FILE: xtuner-eval_niah/xtuner/_lite/modelings/llava/configuration_llava.py
  class EnhancedLlavaConfig (line 23) | class EnhancedLlavaConfig(PretrainedConfig):
    method __init__ (line 76) | def __init__(
    method save_pretrained (line 143) | def save_pretrained(self, save_directory: Union[str, os.PathLike], pus...

FILE: xtuner-eval_niah/xtuner/_lite/modelings/llava/modeling_internlm2.py
  function _get_unpad_data (line 62) | def _get_unpad_data(attention_mask):
  class InternLM2RMSNorm (line 75) | class InternLM2RMSNorm(nn.Module):
    method __init__ (line 78) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 83) | def forward(self, hidden_states):
  class InternLM2RotaryEmbedding (line 95) | class InternLM2RotaryEmbedding(nn.Module):
    method __init__ (line 98) | def __init__(self,
    method forward (line 118) | def forward(self, x, position_ids):
  class InternLM2LinearScalingRotaryEmbedding (line 137) | class InternLM2LinearScalingRotaryEmbedding(InternLM2RotaryEmbedding):
    method forward (line 140) | def forward(self, x, position_ids):
  class InternLM2DynamicNTKScalingRotaryEmbedding (line 147) | class InternLM2DynamicNTKScalingRotaryEmbedding(InternLM2RotaryEmbedding):
    method forward (line 151) | def forward(self, x, position_ids):
  function rotate_half (line 171) | def rotate_half(x):
  function apply_rotary_pos_emb (line 178) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  class InternLM2MLP (line 204) | class InternLM2MLP(nn.Module):
    method __init__ (line 207) | def __init__(self, config):
    method forward (line 220) | def forward(self, x):
  function repeat_kv (line 226) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class InternLM2Attention (line 242) | class InternLM2Attention(nn.Module):
    method __init__ (line 245) | def __init__(self,
    method _init_rope (line 281) | def _init_rope(self):
    method forward (line 308) | def forward(
  class InternLM2FlashAttention2 (line 405) | class InternLM2FlashAttention2(InternLM2Attention):
    method __init__ (line 412) | def __init__(self, *args, **kwargs):
    method forward (line 424) | def forward(
    method _flash_attention_forward (line 529) | def _flash_attention_forward(self,
    method _upad_input (line 598) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class InternLM2SdpaAttention (line 641) | class InternLM2SdpaAttention(InternLM2Attention):
    method forward (line 649) | def forward(
  class InternLM2DecoderLayer (line 757) | class InternLM2DecoderLayer(nn.Module):
    method __init__ (line 760) | def __init__(self, config: InternLM2Config, layer_idx: int):
    method forward (line 775) | def forward(
  class InternLM2PreTrainedModel (line 853) | class InternLM2PreTrainedModel(PreTrainedModel):
    method _init_weights (line 869) | def _init_weights(self, module):
  class InternLM2Model (line 948) | class InternLM2Model(InternLM2PreTrainedModel):
    method __init__ (line 957) | def __init__(self, config: InternLM2Config):
    method get_input_embeddings (line 978) | def get_input_embeddings(self):
    method set_input_embeddings (line 981) | def set_input_embeddings(self, value):
    method forward (line 985) | def forward(
    method _update_causal_mask (line 1106) | def _update_causal_mask(
  class InternLM2ForCausalLM (line 1199) | class InternLM2ForCausalLM(InternLM2PreTrainedModel):
    method __init__ (line 1205) | def __init__(self, config):
    method get_input_embeddings (line 1215) | def get_input_embeddings(self):
    method set_input_embeddings (line 1218) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1221) | def get_output_embeddings(self):
    method set_output_embeddings (line 1224) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1227) | def set_decoder(self, decoder):
    method get_decoder (line 1230) | def get_decoder(self):
    method forward (line 1236) | def forward(
    method prepare_inputs_for_generation (line 1329) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1413) | def _reorder_cache(past_key_values, beam_idx):
    method build_inputs (line 1421) | def build_inputs(self,
    method chat (line 1440) | def chat(
    method stream_chat (line 1487) | def stream_chat(
  class InternLM2ForSequenceClassification (line 1594) | class InternLM2ForSequenceClassification(InternLM2PreTrainedModel):
    method __init__ (line 1597) | def __init__(self, config):
    method get_input_embeddings (line 1606) | def get_input_embeddings(self):
    method set_input_embeddings (line 1609) | def set_input_embeddings(self, value):
    method forward (line 1613) | def forward(
  class InternLM2ForQuestionAnswering (line 1719) | class InternLM2ForQuestionAnswering(InternLM2PreTrainedModel):
    method __init__ (line 1724) | def __init__(self, config):
    method get_input_embeddings (line 1732) | def get_input_embeddings(self):
    method set_input_embeddings (line 1735) | def set_input_embeddings(self, value):
    method forward (line 1739) | def forward(
  class InternLM2ForTokenClassification (line 1823) | class InternLM2ForTokenClassification(InternLM2PreTrainedModel):
    method __init__ (line 1826) | def __init__(self, config):
    method get_input_embeddings (line 1842) | def get_input_embeddings(self):
    method set_input_embeddings (line 1845) | def set_input_embeddings(self, value):
    method forward (line 1849) | def forward(

FILE: xtuner-eval_niah/xtuner/_lite/modelings/llava/modeling_llava.py
  class LlavaCausalLMOutputWithPast (line 46) | class LlavaCausalLMOutputWithPast(ModelOutput):
  class LlavaMultiModalProjector (line 87) | class LlavaMultiModalProjector(nn.Module):
    method __init__ (line 88) | def __init__(self, config: EnhancedLlavaConfig):
    method forward (line 95) | def forward(self, image_features):
  class LlavaPreTrainedModel (line 123) | class LlavaPreTrainedModel(PreTrainedModel):
    method _init_weights (line 131) | def _init_weights(self, module):
    method _supports_sdpa (line 154) | def _supports_sdpa(self):
  class LlavaForConditionalGeneration (line 237) | class LlavaForConditionalGeneration(LlavaPreTrainedModel):
    method __init__ (line 241) | def __init__(self, config: EnhancedLlavaConfig):
    method get_input_embeddings (line 253) | def get_input_embeddings(self):
    method set_input_embeddings (line 256) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 259) | def get_output_embeddings(self):
    method set_output_embeddings (line 262) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 265) | def set_decoder(self, decoder):
    method get_decoder (line 268) | def get_decoder(self):
    method tie_weights (line 271) | def tie_weights(self):
    method resize_token_embeddings (line 274) | def resize_token_embeddings(self, new_num_tokens: Optional[int] = None...
    method _merge_input_ids_with_image_features (line 281) | def _merge_input_ids_with_image_features(self, image_features, inputs_...
    method forward (line 360) | def forward(
    method prepare_inputs_for_generation (line 516) | def prepare_inputs_for_generation(
    method _reorder_cache (line 569) | def _reorder_cache(self, *args, **kwargs):

FILE: xtuner-eval_niah/xtuner/_lite/modelings/llava/processing_llava.py
  class LlavaProcessor (line 28) | class LlavaProcessor(ProcessorMixin):
    method __init__ (line 49) | def __init__(self, image_processor=None, tokenizer=None, chat_template...
    method __call__ (line 52) | def __call__(
    method batch_decode (line 117) | def batch_decode(self, *args, **kwargs):
    method decode (line 125) | def decode(self, *args, **kwargs):
    method model_input_names (line 134) | def model_input_names(self):

FILE: xtuner-eval_niah/xtuner/_lite/parallel/comm.py
  function _all_to_all (line 13) | def _all_to_all(
  class _AllToAll (line 29) | class _AllToAll(torch.autograd.Function):
    method forward (line 40) | def forward(ctx: Any, input: Tensor, sp_group: dist.ProcessGroup,
    method backward (line 51) | def backward(ctx: Any, grad_output: Tensor) -> Tuple:
  function all_to_all (line 67) | def all_to_all(
  function all_to_all_list (line 94) | def all_to_all_list(object_list, group=None):

FILE: xtuner-eval_niah/xtuner/_lite/parallel/fsdp/checkpointing.py
  function checkpoint_check_fn (line 7) | def checkpoint_check_fn(submodule, target=RECOMPUTE_MODULES, selective=1...

FILE: xtuner-eval_niah/xtuner/_lite/parallel/fsdp/lazy.py
  function dp_lazy_init (line 7) | def dp_lazy_init(module, module_map, dp_mesh):
  function dp_sp_lazy_init (line 35) | def dp_sp_lazy_init(module, module_map, dp_mesh, sp_mesh):
  function dp_tp_lazy_init (line 61) | def dp_tp_lazy_init(module, module_map, dp_mesh, tp_mesh):
  class LoadWoInit (line 139) | class LoadWoInit:
    method __init__ (line 142) | def __init__(self):
    method __enter__ (line 151) | def __enter__(self, *args, **kwargs):
    method __exit__ (line 160) | def __exit__(self, *args, **kwargs):

FILE: xtuner-eval_niah/xtuner/_lite/parallel/fsdp/precision.py
  function set_require_grad_param_to_fp32 (line 5) | def set_require_grad_param_to_fp32(model: nn.Module):

FILE: xtuner-eval_niah/xtuner/_lite/parallel/fsdp/wrap.py
  function layer_auto_wrap_policy (line 12) | def layer_auto_wrap_policy(
  function layer_and_emb_wrap_policy (line 27) | def layer_and_emb_wrap_policy(
  function token_embedding_wrap_policy (line 49) | def token_embedding_wrap_policy(
  function all_required_grad_wrap_policy (line 66) | def all_required_grad_wrap_policy(

FILE: xtuner-eval_niah/xtuner/_lite/parallel/sampler.py
  class ParallelSampler (line 13) | class ParallelSampler(Sampler):
    method __init__ (line 42) | def __init__(
    method __iter__ (line 78) | def __iter__(self) -> Iterator[int]:
    method __len__ (line 99) | def __len__(self) -> int:
    method set_epoch (line 103) | def set_epoch(self, epoch: int, step=0) -> None:
  function get_length_grouped_indices (line 117) | def get_length_grouped_indices(max_lengths,
  class LengthGroupedSampler (line 146) | class LengthGroupedSampler(Sampler):
    method __init__ (line 148) | def __init__(self,
    method __iter__ (line 200) | def __iter__(self) -> Iterator[int]:
    method __len__ (line 222) | def __len__(self) -> int:
    method set_epoch (line 226) | def set_epoch(self, epoch: int, step=0) -> None:

FILE: xtuner-eval_niah/xtuner/_lite/parallel/sequence/attention.py
  function pre_process_for_sequence_parallel_attn (line 8) | def pre_process_for_sequence_parallel_attn(query_states,
  function post_process_for_sequence_parallel_attn (line 41) | def post_process_for_sequence_parallel_attn(attn_output,
  function sequence_parallel_wrapper (line 54) | def sequence_parallel_wrapper(local_attn):

FILE: xtuner-eval_niah/xtuner/_lite/parallel/sequence/data_collate.py
  function pad_for_sequence_parallel (line 7) | def pad_for_sequence_parallel(tensor, padding_value, dim=-1):
  function pad_cumulative_len_for_sequence_parallel (line 26) | def pad_cumulative_len_for_sequence_parallel(cumulative_len):

FILE: xtuner-eval_niah/xtuner/_lite/parallel/sequence/ops.py
  function split_for_sequence_parallel (line 6) | def split_for_sequence_parallel(input, dim: int, sp_group: dist.ProcessG...
  function gather_for_sequence_parallel (line 33) | def gather_for_sequence_parallel(input, dim: int, sp_group: dist.Process...
  class _GatherForwardSplitBackward (line 60) | class _GatherForwardSplitBackward(torch.autograd.Function):
    method forward (line 68) | def forward(ctx, input, dim, sp_group, grad_scale):
    method backward (line 75) | def backward(ctx, grad_output):
  class _SplitForwardGatherBackward (line 85) | class _SplitForwardGatherBackward(torch.autograd.Function):
    method forward (line 93) | def forward(ctx, input, dim, sp_group, grad_scale):
    method backward (line 100) | def backward(ctx, grad_output):
  function split_forward_gather_backward (line 109) | def split_forward_gather_backward(input, dim, sp_group, grad_scale=None):
  function gather_forward_split_backward (line 123) | def gather_forward_split_backward(input, dim, sp_group, grad_scale=None):

FILE: xtuner-eval_niah/xtuner/_lite/parallel/sequence/reduce_loss.py
  class _ReduceLoss (line 7) | class _ReduceLoss(torch.autograd.Function):
    method forward (line 10) | def forward(ctx, mean_loss, loss_scale, process_group):
    method backward (line 22) | def backward(ctx, grad_output):
  function reduce_sequence_parallel_loss (line 26) | def reduce_sequence_parallel_loss(mean_loss,

FILE: xtuner-eval_niah/xtuner/_lite/parallel/setup.py
  function setup_sp (line 15) | def setup_sp(sp_size):
  function setup_tp (line 31) | def setup_tp(tp_size):
  function setup_dp (line 47) | def setup_dp():
  function set_seq_parallel_pg (line 67) | def set_seq_parallel_pg(sp_ulysses_degree, sp_ring_degree):
  function setup_parallel (line 93) | def setup_parallel(sp_size=1, tp_size=1, ring_size=1):
  function get_ulysess_mesh (line 108) | def get_ulysess_mesh():
  function get_ring_mesh (line 112) | def get_ring_mesh():
  function get_ulysess_group (line 116) | def get_ulysess_group():
  function get_ring_group (line 120) | def get_ring_group():
  function get_ulysess_world_size (line 124) | def get_ulysess_world_size():
  function get_ring_world_size (line 135) | def get_ring_world_size():
  function get_dp_mesh (line 146) | def get_dp_mesh():
  function get_dp_group (line 150) | def get_dp_group():
  function get_dp_world_size (line 154) | def get_dp_world_size():
  function get_sp_mesh (line 165) | def get_sp_mesh():
  function get_sp_group (line 169) | def get_sp_group():
  function get_sp_world_size (line 173) | def get_sp_world_size():
  function get_tp_mesh (line 184) | def get_tp_mesh():
  function get_tp_group (line 188) | def get_tp_group():
  function get_tp_world_size (line 192) | def get_tp_world_size():

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/comm/all_to_all.py
  function all_to_all_4D (line 15) | def all_to_all_4D(
  class SeqAllToAll4D (line 96) | class SeqAllToAll4D(torch.autograd.Function):
    method forward (line 98) | def forward(
    method backward (line 113) | def backward(ctx: Any, *grad_output: Tensor) -> Tuple[None, Tensor, No...
  function all_to_all_5D (line 124) | def all_to_all_5D(
  class SeqAllToAll5D (line 207) | class SeqAllToAll5D(torch.autograd.Function):
    method forward (line 209) | def forward(
    method backward (line 224) | def backward(ctx: Any, *grad_output: Tensor) -> Tuple[None, Tensor, No...

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/comm/extract_local.py
  function stripe_extract_local (line 7) | def stripe_extract_local(value, rank, world_size, rd, ud, *args, **kwargs):
  function basic_extract_local (line 29) | def basic_extract_local(value, rank, world_size, *args, **kwargs):
  function zigzag_extract_local (line 33) | def zigzag_extract_local(value, rank, world_size, rd, ud, dim=1, *args, ...

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/globals.py
  class Singleton (line 4) | class Singleton:
    method __new__ (line 7) | def __new__(cls, *args, **kwargs):
  class ProcessGroupSingleton (line 13) | class ProcessGroupSingleton(Singleton):
    method __init__ (line 14) | def __init__(self):
  function set_seq_parallel_pg (line 22) | def set_seq_parallel_pg(

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/hybrid/async_attn_layer.py
  class AsyncLongContextAttention (line 11) | class AsyncLongContextAttention(torch.nn.Module):
    method __init__ (line 21) | def __init__(
    method forward (line 42) | def forward(
    method backward (line 195) | def backward(self, *args, **kwargs):

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/hybrid/attn_layer.py
  class LongContextAttention (line 15) | class LongContextAttention(torch.nn.Module):
    method __init__ (line 25) | def __init__(
    method forward (line 45) | def forward(
  class LongContextAttentionQKVPacked (line 134) | class LongContextAttentionQKVPacked(torch.nn.Module):
    method __init__ (line 144) | def __init__(
    method forward (line 164) | def forward(
  function llama3_varlen_attention_sp_ulysses_ring (line 225) | def llama3_varlen_attention_sp_ulysses_ring(
  function attention_sp_ulysses_ring (line 299) | def attention_sp_ulysses_ring(
  class LongContextVarLenAttentionForLlaMa3 (line 347) | class LongContextVarLenAttentionForLlaMa3(torch.nn.Module):
    method __init__ (line 357) | def __init__(
    method forward (line 372) | def forward(

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/ring/llama3_flash_attn_varlen.py
  class AsyncHandles (line 10) | class AsyncHandles:
    method __init__ (line 12) | def __init__(self) -> None:
    method register (line 15) | def register(self, handle):
    method wait (line 18) | def wait(self):
  function llama3_flash_attn_prepare_cu_seqlens (line 24) | def llama3_flash_attn_prepare_cu_seqlens(cu_seqlens, causal, rank, world...
  function llama3_flash_attn_varlen_forward (line 63) | def llama3_flash_attn_varlen_forward(
  function llama3_flash_attn_varlen_backward (line 164) | def llama3_flash_attn_varlen_backward(
  class Llama3FlashAttnVarlenFunc (line 313) | class Llama3FlashAttnVarlenFunc(torch.autograd.Function):
    method forward (line 315) | def forward(
    method backward (line 375) | def backward(ctx, dout, *args):
  function llama3_flash_attn_varlen_qkvpacked_func (line 401) | def llama3_flash_attn_varlen_qkvpacked_func(
  function llama3_flash_attn_varlen_kvpacked_func (line 439) | def llama3_flash_attn_varlen_kvpacked_func(
  function llama3_flash_attn_varlen_func (line 478) | def llama3_flash_attn_varlen_func(

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/ring/ring_flash_attn.py
  function ring_flash_attn_forward (line 8) | def ring_flash_attn_forward(
  function ring_flash_attn_backward (line 61) | def ring_flash_attn_backward(
  class RingFlashAttnFunc (line 145) | class RingFlashAttnFunc(torch.autograd.Function):
    method forward (line 147) | def forward(
    method backward (line 191) | def backward(ctx, dout, *args):
  function ring_flash_attn_qkvpacked_func (line 211) | def ring_flash_attn_qkvpacked_func(
  function ring_flash_attn_kvpacked_func (line 237) | def ring_flash_attn_kvpacked_func(
  function ring_flash_attn_func (line 264) | def ring_flash_attn_func(
  function ring_flash_attn_inference_func (line 293) | def ring_flash_attn_inference_func(

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/ring/ring_flash_attn_varlen.py
  function ring_flash_attn_varlen_forward (line 25) | def ring_flash_attn_varlen_forward(
  function ring_flash_attn_varlen_backward (line 95) | def ring_flash_attn_varlen_backward(
  class RingFlashAttnVarlenFunc (line 184) | class RingFlashAttnVarlenFunc(torch.autograd.Function):
    method forward (line 186) | def forward(
    method backward (line 235) | def backward(ctx, dout, *args):
  function ring_flash_attn_varlen_qkvpacked_func (line 257) | def ring_flash_attn_varlen_qkvpacked_func(
  function ring_flash_attn_varlen_kvpacked_func (line 287) | def ring_flash_attn_varlen_kvpacked_func(
  function ring_flash_attn_varlen_func (line 318) | def ring_flash_attn_varlen_func(

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/ring/stripe_flash_attn.py
  function stripe_flash_attn_forward (line 7) | def stripe_flash_attn_forward(
  function stripe_flash_attn_backward (line 81) | def stripe_flash_attn_backward(
  class StripeFlashAttnFunc (line 199) | class StripeFlashAttnFunc(torch.autograd.Function):
    method forward (line 201) | def forward(
    method backward (line 245) | def backward(ctx, dout, *args):
  function stripe_flash_attn_qkvpacked_func (line 265) | def stripe_flash_attn_qkvpacked_func(
  function stripe_flash_attn_kvpacked_func (line 291) | def stripe_flash_attn_kvpacked_func(
  function stripe_flash_attn_func (line 318) | def stripe_flash_attn_func(

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/ring/triton_utils.py
  function flatten_kernel (line 7) | def flatten_kernel(
  function flatten_varlen_lse (line 39) | def flatten_varlen_lse(lse, cu_seqlens):
  function unflatten_kernel (line 71) | def unflatten_kernel(
  function unflatten_varlen_lse (line 103) | def unflatten_varlen_lse(lse, cu_seqlens, max_seqlen: int):

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/ring/utils.py
  function get_default_args (line 14) | def get_default_args(func):
  function _update_out_and_lse (line 25) | def _update_out_and_lse(
  function update_out_and_lse (line 45) | def update_out_and_lse(
  function flatten_varlen_lse (line 69) | def flatten_varlen_lse(lse, cu_seqlens):
  function unflatten_varlen_lse (line 78) | def unflatten_varlen_lse(lse, cu_seqlens, max_seqlen: int):
  class RingComm (line 90) | class RingComm:
    method __init__ (line 91) | def __init__(self, process_group: dist.ProcessGroup):
    method send_recv (line 105) | def send_recv(
    method commit (line 121) | def commit(self):
    method wait (line 126) | def wait(self):

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/ring/zigzag_ring_flash_attn.py
  function zigzag_ring_flash_attn_forward (line 7) | def zigzag_ring_flash_attn_forward(
  function zigzag_ring_flash_attn_backward (line 81) | def zigzag_ring_flash_attn_backward(
  class ZigZagRingFlashAttnFunc (line 187) | class ZigZagRingFlashAttnFunc(torch.autograd.Function):
    method forward (line 189) | def forward(
    method backward (line 233) | def backward(ctx, dout, *args):
  function zigzag_ring_flash_attn_qkvpacked_func (line 253) | def zigzag_ring_flash_attn_qkvpacked_func(
  function zigzag_ring_flash_attn_kvpacked_func (line 279) | def zigzag_ring_flash_attn_kvpacked_func(
  function zigzag_ring_flash_attn_func (line 306) | def zigzag_ring_flash_attn_func(

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/ring/zigzag_ring_flash_attn_varlen.py
  function get_half_index (line 24) | def get_half_index(cu_seqlens, *, front: bool):
  function get_half_lse (line 43) | def get_half_lse(lse, cu_seqlens, *, front: bool):
  function zigzag_ring_flash_attn_varlen_forward (line 74) | def zigzag_ring_flash_attn_varlen_forward(
  function zigzag_ring_flash_attn_varlen_backward (line 184) | def zigzag_ring_flash_attn_varlen_backward(
  class ZigZagRingFlashAttnVarlenFunc (line 305) | class ZigZagRingFlashAttnVarlenFunc(torch.autograd.Function):
    method forward (line 307) | def forward(
    method backward (line 369) | def backward(ctx, dout, *args):
  function zigzag_ring_flash_attn_varlen_qkvpacked_func (line 400) | def zigzag_ring_flash_attn_varlen_qkvpacked_func(
  function zigzag_ring_flash_attn_varlen_kvpacked_func (line 430) | def zigzag_ring_flash_attn_varlen_kvpacked_func(
  function zigzag_ring_flash_attn_varlen_func (line 461) | def zigzag_ring_flash_attn_varlen_func(

FILE: xtuner-eval_niah/xtuner/_lite/yunchang/ulysses/attn_layer.py
  function torch_attn (line 16) | def torch_attn(query,
  class UlyssesAttention (line 42) | class UlyssesAttention(torch.nn.Module):
    method __init__ (line 52) | def __init__(
    method forward (line 70) | def forward(

FILE: xtuner-eval_niah/xtuner/apis/datasets/alpaca.py
  function alpaca_enzh_dataset (line 14) | def alpaca_enzh_dataset(tokenizer,
  function alpaca_enzh_data_collator (line 41) | def alpaca_enzh_data_collator(return_hf_format=False):
  function alpaca_zh_dataset (line 45) | def alpaca_zh_dataset(tokenizer,
  function alpaca_zh_data_collator (line 66) | def alpaca_zh_data_collator(return_hf_format=False):
  function alpaca_dataset (line 70) | def alpaca_dataset(tokenizer,
  function alpaca_data_collator (line 91) | def alpaca_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/datasets/arxiv.py
  function arxiv_dataset (line 12) | def arxiv_dataset(tokenizer,
  function arxiv_data_collator (line 37) | def arxiv_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/datasets/code_alpaca.py
  function code_alpaca_dataset (line 12) | def code_alpaca_dataset(tokenizer,
  function code_alpaca_data_collator (line 33) | def code_alpaca_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/datasets/colorist.py
  function colorist_dataset (line 12) | def colorist_dataset(tokenizer,
  function colorist_data_collator (line 33) | def colorist_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/datasets/lawyer.py
  function lawyer_dataset (line 15) | def lawyer_dataset(tokenizer,
  function lawyer_data_collator (line 40) | def lawyer_data_collator(return_hf_format=False):
  function lawyer_crime_dataset (line 44) | def lawyer_crime_dataset(tokenizer,
  function lawyer_crime_data_collator (line 68) | def lawyer_crime_data_collator(return_hf_format=False):
  function lawyer_reference_dataset (line 72) | def lawyer_reference_dataset(tokenizer,
  function lawyer_reference_data_collator (line 96) | def lawyer_reference_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/datasets/medical.py
  function medical_dataset (line 12) | def medical_dataset(tokenizer,
  function medical_data_collator (line 33) | def medical_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/datasets/moss_003_sft.py
  function moss_003_sft_dataset (line 10) | def moss_003_sft_dataset(tokenizer,
  function moss_003_sft_data_collator (line 29) | def moss_003_sft_data_collator(return_hf_format=False):
  function moss_003_sft_no_plugins_dataset (line 33) | def moss_003_sft_no_plugins_dataset(tokenizer,
  function moss_003_sft_no_plugins_data_collator (line 50) | def moss_003_sft_no_plugins_data_collator(return_hf_format=False):
  function moss_003_sft_plugins_dataset (line 54) | def moss_003_sft_plugins_dataset(tokenizer,
  function moss_003_sft_plugins_data_collator (line 71) | def moss_003_sft_plugins_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/datasets/oasst1.py
  function oasst1_dataset (line 12) | def oasst1_dataset(tokenizer,
  function oasst1_data_collator (line 33) | def oasst1_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/datasets/open_orca.py
  function openorca_dataset (line 12) | def openorca_dataset(tokenizer,
  function openorca_data_collator (line 33) | def openorca_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/datasets/sql.py
  function sql_dataset (line 12) | def sql_dataset(tokenizer,
  function sql_data_collator (line 33) | def sql_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/datasets/tiny_codes.py
  function tiny_codes_dataset (line 12) | def tiny_codes_dataset(tokenizer,
  function tiny_codes_data_collator (line 33) | def tiny_codes_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/datasets/wizardlm.py
  function wizardlm_dataset (line 12) | def wizardlm_dataset(tokenizer,
  function wizardlm_data_collator (line 33) | def wizardlm_data_collator(return_hf_format=False):

FILE: xtuner-eval_niah/xtuner/apis/model.py
  function build_qlora_model (line 12) | def build_qlora_model(model_name_or_path,
  function build_lora_model (line 52) | def build_lora_model(model_name_or_path,
  function build_model (line 78) | def build_model(model_name_or_path, return_tokenizer=True):

FILE: xtuner-eval_niah/xtuner/apis/training_args.py
  class DefaultTrainingArguments (line 12) | class DefaultTrainingArguments(TrainingArguments):

FILE: xtuner-eval_niah/xtuner/configs/__init__.py
  function get_cfgs_name_path (line 5) | def get_cfgs_name_path():

FILE: xtuner-eval_niah/xtuner/configs/llava/llama3_8b_instruct_clip_vit_large_p14_336/convert_xtuner_weights_to_hf.py
  function convert_state_dict_to_hf (line 25) | def convert_state_dict_to_hf(state_dict, mapping):
  function convert_to_hf (line 38) | def convert_to_hf(text_model_id, vision_model_id, projector_weight, save...
  function main (line 131) | def main():

FILE: xtuner-eval_niah/xtuner/configs/llava/llama3_8b_instruct_clip_vit_large_p14_336/convert_xtuner_weights_to_llava.py
  function convert_state_dict_to_hf (line 27) | def convert_state_dict_to_hf(state_dict, mapping):
  function convert_to_llava (line 39) | def convert_to_llava(text_model_id, vision_model_id, projector_weight,
  function main (line 94) | def main():

FILE: xtuner-eval_niah/xtuner/configs/llava/phi3_mini_4k_instruct_clip_vit_large_p14_336/convert_phi_to_llama.py
  function convert_phi_to_llama (line 13) | def convert_phi_to_llama(phi_path, save_path):
  function main (line 91) | def main():

FILE: xtuner-eval_niah/xtuner/configs/llava/phi3_mini_4k_instruct_clip_vit_large_p14_336/convert_xtuner_weights_to_hf.py
  function convert_state_dict_to_hf (line 25) | def convert_state_dict_to_hf(state_dict, mapping):
  function convert_to_hf (line 38) | def convert_to_hf(text_model_id, vision_model_id, projector_weight, save...
  function main (line 128) | def main():

FILE: xtuner-eval_niah/xtuner/configs/llava/phi3_mini_4k_instruct_clip_vit_large_p14_336/convert_xtuner_weights_to_llava.py
  function convert_state_dict_to_hf (line 27) | def convert_state_dict_to_hf(state_dict, mapping):
  function convert_to_llava (line 39) | def convert_to_llava(text_model_id, vision_model_id, projector_weight,
  function main (line 94) | def main():

FILE: xtuner-eval_niah/xtuner/dataset/collate_fns/default_collate_fn.py
  function default_collate_fn (line 12) | def default_collate_fn(instances: Sequence[Dict],

FILE: xtuner-eval_niah/xtuner/dataset/collate_fns/mmlu_collate_fn.py
  function mmlu_collate_fn (line 10) | def mmlu_collate_fn(instances: Sequence[Dict],

FILE: xtuner-eval_niah/xtuner/dataset/collate_fns/preference_collate_fn.py
  function preference_collate_fn (line 13) | def preference_collate_fn(instances: Sequence[Dict],

FILE: xtuner-eval_niah/xtuner/dataset/concat_dataset.py
  class ConcatDataset (line 7) | class ConcatDataset(TorchConcatDataset):
    method __init__ (line 9) | def __init__(self, datasets):
    method __repr__ (line 15) | def __repr__(self):

FILE: xtuner-eval_niah/xtuner/dataset/huggingface.py
  function get_lengths (line 18) | def get_lengths(example):
  function build_origin_dataset (line 22) | def build_origin_dataset(dataset, split):
  function map_dataset (line 39) | def map_dataset(dataset, dataset_map_fn, map_num_proc):
  function add_template_to_dataset (line 54) | def add_template_to_dataset(dataset, template_map_fn, map_num_proc):
  function tokenize_dataset (line 67) | def tokenize_dataset(dataset, tokenizer, max_length, with_image_token,
  function pack_dataset (line 88) | def pack_dataset(dataset, max_length, use_varlen_attn, shuffle_before_pack,
  function process (line 100) | def process(dataset,
  function process_hf_dataset (line 222) | def process_hf_dataset(dataset,

FILE: xtuner-eval_niah/xtuner/dataset/intern_repo.py
  class JsonlDataset (line 22) | class JsonlDataset(torch.utils.data.Dataset):
    method __init__ (line 36) | def __init__(self, path: str, min_length=50):
    method __getitem__ (line 62) | def __getitem__(self, idx):
    method get_dataset_name (line 84) | def get_dataset_name(self):
    method _get_mmap (line 87) | def _get_mmap(self):
    method __setstate__ (line 100) | def __setstate__(self, state):
    method __getstate__ (line 104) | def __getstate__(self):
    method __del__ (line 111) | def __del__(self):
    method exists (line 118) | def exists(path):
    method __len__ (line 121) | def __len__(self):
  class PackedDataset (line 128) | class PackedDataset(torch.utils.data.Dataset):
    method __init__ (line 137) | def __init__(self, dataset, packed_length: int = 8192, seed: int = 1024):
    method __len__ (line 162) | def __len__(self):
    method search_sample_index (line 165) | def search_sample_index(self, pack_idx: int = 0):
    method mapping (line 172) | def mapping(self, pack_idx: int = 0):
    method build_pack (line 192) | def build_pack(self, begin_sample_idx: int, begin_token_id: int,
    method __getitem__ (line 228) | def __getitem__(self, item: int):
  function load_intern_repo_tokenized_dataset (line 235) | def load_intern_repo_tokenized_dataset(folder,
  function load_intern_repo_untokenized_dataset (line 268) | def load_intern_repo_untokenized_dataset(processed_dataset_dict_path=None,
  function build_packed_dataset_rank0 (line 332) | def build_packed_dataset_rank0(dataset_cfg, packed_length=8192, seed=1024):
  function build_packed_dataset (line 352) | def build_packed_dataset(*args, **kwargs):

FILE: xtuner-eval_niah/xtuner/dataset/json_dataset.py
  function load_json_file (line 7) | def load_json_file(data_files=None, data_dir=None, suffix=None):

FILE: xtuner-eval_niah/xtuner/dataset/llava.py
  function load_jsonl (line 19) | def load_jsonl(json_file):
  class LLaVADataset (line 28) | class LLaVADataset(Dataset):
    method __init__ (line 30) | def __init__(self,
    method modality_length (line 89) | def modality_length(self):
    method __len__ (line 98) | def __len__(self):
    method __getitem__ (line 101) | def __getitem__(self, index):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/alpaca_map_fn.py
  function alpaca_map_fn (line 4) | def alpaca_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/alpaca_zh_map_fn.py
  function alpaca_zh_map_fn (line 4) | def alpaca_zh_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/arxiv_map_fn.py
  function arxiv_map_fn (line 5) | def arxiv_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/code_alpaca_map_fn.py
  function code_alpaca_map_fn (line 5) | def code_alpaca_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/colors_map_fn.py
  function colors_map_fn (line 5) | def colors_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/crime_kg_assitant_map_fn.py
  function crime_kg_assitant_map_fn (line 5) | def crime_kg_assitant_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/default_map_fn.py
  function default_map_fn (line 2) | def default_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/law_reference_map_fn.py
  function law_reference_map_fn (line 5) | def law_reference_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/llava_map_fn.py
  function llava_image_only_map_fn (line 5) | def llava_image_only_map_fn(example):
  function llava_map_fn (line 25) | def llava_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/medical_map_fn.py
  function medical_map_fn (line 5) | def medical_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/msagent_map_fn.py
  function replace_think (line 9) | def replace_think(match):
  function replace_exec (line 22) | def replace_exec(match):
  function extract_json_objects (line 34) | def extract_json_objects(text, decoder=json.JSONDecoder()):
  function msagent_react_map_fn (line 53) | def msagent_react_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/oasst1_map_fn.py
  function oasst1_map_fn (line 2) | def oasst1_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/openai_map_fn.py
  function openai_map_fn (line 2) | def openai_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/openorca_map_fn.py
  function openorca_map_fn (line 2) | def openorca_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/pretrain_map_fn.py
  function pretrain_map_fn (line 2) | def pretrain_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/sql_map_fn.py
  function sql_map_fn (line 5) | def sql_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/stack_exchange_map_fn.py
  function stack_exchange_map_fn (line 2) | def stack_exchange_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/tiny_codes_map_fn.py
  function tiny_codes_map_fn (line 5) | def tiny_codes_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/dataset_map_fns/wizardlm_map_fn.py
  function wizardlm_map_fn (line 2) | def wizardlm_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/map_fns/template_map_fn.py
  function template_map_fn (line 7) | def template_map_fn(example, template):
  function template_map_fn_factory (line 33) | def template_map_fn_factory(template):

FILE: xtuner-eval_niah/xtuner/dataset/modelscope.py
  function process_ms_dataset (line 8) | def process_ms_dataset(dataset, split='train', *args, **kwargs):

FILE: xtuner-eval_niah/xtuner/dataset/moss_sft.py
  class MOSSSFTDataset (line 15) | class MOSSSFTDataset(Dataset):
    method __init__ (line 17) | def __init__(self, data_file, tokenizer, max_length=2048, bot_name=None):
    method load_data (line 43) | def load_data(self):
    method process_data (line 118) | def process_data(self):
    method __len__ (line 125) | def __len__(self):
    method __getitem__ (line 128) | def __getitem__(self, index):

FILE: xtuner-eval_niah/xtuner/dataset/preference_dataset.py
  function _worker (line 24) | def _worker(
  function _chunk_data_to_queue (line 41) | def _chunk_data_to_queue(data_queue: Queue, data: List[Dict], chunk_size...
  function _multi_progress (line 61) | def _multi_progress(tokenize_fun_p, dataset, nproc, task_num, chunksize,
  function load_jsonl_dataset (line 90) | def load_jsonl_dataset(data_files=None, data_dir=None, suffix=None):
  function tokenize (line 110) | def tokenize(pair: str,
  class PreferenceDataset (line 156) | class PreferenceDataset(Dataset):
    method __init__ (line 158) | def __init__(
    method __len__ (line 194) | def __len__(self):
    method __getitem__ (line 197) | def __getitem__(self, idx):
  class PackedDatasetWrapper (line 201) | class PackedDatasetWrapper(Dataset):
    method __init__ (line 203) | def __init__(self,
    method __len__ (line 252) | def __len__(self):
    method __getitem__ (line 255) | def __getitem__(self, index):
  function unpack_seq (line 280) | def unpack_seq(seq, cu_seqlens):
  function broad_cast_dataset (line 288) | def broad_cast_dataset(dataset):
  function map_dataset (line 309) | def map_dataset(dataset, dataset_map_fn, map_num_proc):
  function build_preference_dataset (line 324) | def build_preference_dataset(
  function intel_orca_dpo_map_fn (line 368) | def intel_orca_dpo_map_fn(example):
  function orpo_dpo_mix_40k_map_fn (line 381) | def orpo_dpo_mix_40k_map_fn(example):

FILE: xtuner-eval_niah/xtuner/dataset/refcoco_json.py
  class RefCOCOJsonDataset (line 26) | class RefCOCOJsonDataset(LLaVADataset):
    method __init__ (line 37) | def __init__(
    method reformat_data (line 81) | def reformat_data(self, json_data):
    method gen_refcoco_conversations (line 91) | def gen_refcoco_conversations(cls, data, instruction_template='{}'):
    method get_data_json (line 127) | def get_data_json(
    method normalize_bbox (line 168) | def normalize_bbox(cls, bbox, height, width):
  class RefCOCOJsonEvalDataset (line 176) | class RefCOCOJsonEvalDataset(RefCOCOJsonDataset):
    method reformat_data (line 179) | def reformat_data(self, json_data):
  class InvRefCOCOJsonDataset (line 188) | class InvRefCOCOJsonDataset(RefCOCOJsonDataset):
    method gen_refcoco_conversations (line 200) | def gen_refcoco_conversations(cls, data, instruction_template='{}'):
  class REFER (line 241) | class REFER:
    method __init__ (line 243) | def __init__(self, data_root, vis_root, dataset='refcoco', splitBy='un...
    method createIndex (line 276) | def createIndex(self):
    method getRefIds (line 340) | def getRefIds(self, image_ids=[], cat_ids=[], ref_ids=[], split=''):
    method getAnnIds (line 372) | def getAnnIds(self, image_ids=[], cat_ids=[], ref_ids=[]):
    method getImgIds (line 397) | def getImgIds(self, ref_ids=[]):
    method getCatIds (line 408) | def getCatIds(self):
    method loadRefs (line 411) | def loadRefs(self, ref_ids=[]):
    method loadAnns (line 417) | def loadAnns(self, ann_ids=[]):
    method loadImgs (line 423) | def loadImgs(self, image_ids=[]):
    method loadCats (line 429) | def loadCats(self, cat_ids=[]):
    method getRefBox (line 435) | def getRefBox(self, ref_id):
    method showRef (line 440) | def showRef(self, ref, seg_box='box'):

FILE: xtuner-eval_niah/xtuner/dataset/samplers/intern_repo.py
  class InternRepoSampler (line 13) | class InternRepoSampler(Sampler):
    method __init__ (line 15) | def __init__(self,
    method __iter__ (line 35) | def __iter__(self) -> Iterator[int]:
    method __len__ (line 54) | def __len__(self) -> int:
    method set_epoch (line 58) | def set_epoch(self, epoch: int) -> None:
  class InternlmRepoSampler (line 71) | class InternlmRepoSampler(InternRepoSampler):
    method __init__ (line 73) | def __init__(self,

FILE: xtuner-eval_niah/xtuner/dataset/samplers/length_grouped.py
  function get_length_grouped_indices (line 11) | def get_length_grouped_indices(lengths, group_batch_size, generator=None):
  class LengthGroupedSampler (line 72) | class LengthGroupedSampler(Sampler):
    method __init__ (line 74) | def __init__(self,
    method __iter__ (line 124) | def __iter__(self) -> Iterator[int]:
    method __len__ (line 144) | def __len__(self) -> int:
    method set_epoch (line 148) | def set_epoch(self, epoch: int) -> None:

FILE: xtuner-eval_niah/xtuner/dataset/utils.py
  function get_bos_eos_token_ids (line 15) | def get_bos_eos_token_ids(tokenizer):
  function encode_fn (line 36) | def encode_fn(example,
  class Packer (line 132) | class Packer:
    method __init__ (line 135) | def __init__(self,
    method get_cumulative_len (line 146) | def get_cumulative_len(self, chunk_num):
    method get_position_ids (line 179) | def get_position_ids(self, cumulative_len):
    method __call__ (line 191) | def __call__(self, batch):
  function expand2square (line 245) | def expand2square(pil_img, background_color):
  function load_image (line 259) | def load_image(image_file):
  function decode_base64_to_image (line 268) | def decode_base64_to_image(base64_string):

FILE: xtuner-eval_niah/xtuner/engine/_strategy/deepspeed.py
  class DeepSpeedStrategy (line 11) | class DeepSpeedStrategy(MMEngineDeepSpeedStrategy):
    method __init__ (line 13) | def __init__(self, *args, **kwargs):
    method _wrap_model (line 24) | def _wrap_model(self, model):
    method save_checkpoint (line 33) | def save_checkpoint(self, *args, **kwargs) -> None:
    method load_checkpoint (line 45) | def load_checkpoint(self, *args, **kwargs) -> None:
    method resume (line 54) | def resume(self, *args, **kwargs) -> None:
    method _setup_distributed (line 63) | def _setup_distributed(  # type: ignore

FILE: xtuner-eval_niah/xtuner/engine/hooks/dataset_info_hook.py
  function split_list (line 8) | def split_list(lst, value):
  class DatasetInfoHook (line 21) | class DatasetInfoHook(Hook):
    method __init__ (line 23) | def __init__(self, tokenizer, is_intern_repo_dataset=False):
    method log (line 27) | def log(self, runner, dataset, mode='train'):
    method before_train (line 42) | def before_train(self, runner) -> None:
    method before_val (line 52) | def before_val(self, runner) -> None:
    method before_test (line 56) | def before_test(self, runner) -> None:

FILE: xtuner-eval_niah/xtuner/engine/hooks/evaluate_chat_hook.py
  class EvaluateChatHook (line 20) | class EvaluateChatHook(Hook):
    method __init__ (line 24) | def __init__(self,
    method _save_eval_output (line 98) | def _save_eval_output(self, runner, eval_outputs):
    method _eval_images (line 106) | def _eval_images(self,
    method _eval_language (line 167) | def _eval_language(self,
    method _generate_samples (line 194) | def _generate_samples(self,
    method before_train (line 232) | def before_train(self, runner):
    method _is_save_checkpoint (line 236) | def _is_save_checkpoint(self, runner):
    method after_train_iter (line 254) | def after_train_iter(self,
    method after_train (line 273) | def after_train(self, runner):
    method after_val (line 277) | def after_val(self, runner) -> None:

FILE: xtuner-eval_niah/xtuner/engine/hooks/hf_checkpoint_hook.py
  class HFCheckpointHook (line 19) | class HFCheckpointHook(Hook):
    method __init__ (line 23) | def __init__(self, out_dir: Optional[Union[str, Path]] = None) -> None:
    method _use_shard_moe (line 27) | def _use_shard_moe(llm):
    method after_run (line 32) | def after_run(self, runner) -> None:

FILE: xtuner-eval_niah/xtuner/engine/hooks/throughput_hook.py
  class ThroughputHook (line 16) | class ThroughputHook(Hook):
    method __init__ (line 22) | def __init__(self,
    method _guess_is_casual_attn (line 37) | def _guess_is_casual_attn(model):
    method _get_batch_size_and_sequence_len (line 47) | def _get_batch_size_and_sequence_len(data_batch):
    method _guess_use_activation_checkpointing (line 55) | def _guess_use_activation_checkpointing(model):
    method before_run (line 61) | def before_run(self, runner) -> None:
    method after_train_iter (line 88) | def after_train_iter(self,

FILE: xtuner-eval_niah/xtuner/engine/hooks/varlen_attn_args_to_messagehub_hook.py
  class VarlenAttnArgsToMessageHubHook (line 11) | class VarlenAttnArgsToMessageHubHook(Hook):
    method before_train_iter (line 13) | def before_train_iter(self,
    method after_train_iter (line 31) | def after_train_iter(self,
    method before_val_iter (line 41) | def before_val_iter(self,
    method after_val_iter (line 68) | def after_val_iter(self,

FILE: xtuner-eval_niah/xtuner/engine/runner/loops.py
  class TrainLoop (line 8) | class TrainLoop(IterBasedTrainLoop):
    method __init__ (line 10) | def __init__(self,

FILE: xtuner-eval_niah/xtuner/entry_point.py
  function list_dataset_format (line 127) | def list_dataset_format():
  function list_cfg (line 132) | def list_cfg():
  function copy_cfg (line 137) | def copy_cfg():
  function log_dataset (line 142) | def log_dataset():
  function check_custom_dataset (line 147) | def check_custom_dataset():
  function train (line 152) | def train():
  function test (line 157) | def test():
  function chat (line 162) | def chat():
  function mmbench (line 167) | def mmbench():
  function pth_to_hf (line 172) | def pth_to_hf():
  function merge (line 177) | def merge():
  function split (line 182) | def split():
  function arxiv_preprocess (line 187) | def arxiv_preprocess():
  function convert_refcoco (line 192) | def convert_refcoco():
  function convert_help_msg (line 197) | def convert_help_msg():
  function preprocess_help_msg (line 201) | def preprocess_help_msg():
  function eval_refcoco (line 205) | def eval_refcoco():
  function cli (line 244) | def cli():

FILE: xtuner-eval_niah/xtuner/evaluation/metrics/mmlu_metric.py
  class MMLUMetric (line 14) | class MMLUMetric(BaseMetric):
    method __init__ (line 91) | def __init__(self, tokenizer, *args, **kwargs):
    method ABCD_to_0123 (line 102) | def ABCD_to_0123(abcd):
    method find_first_zero_index (line 106) | def find_first_zero_index(tensor):
    method accuracy (line 114) | def accuracy(preds, gts):
    method process (line 120) | def process(self, data_batch: Any, data_samples: Sequence[dict]) -> None:
    method compute_metrics (line 147) | def compute_metrics(self, results: list) -> dict:
    method _print_results (line 236) | def _print_results(self, table_metrics: dict) -> None:

FILE: xtuner-eval_niah/xtuner/evaluation/metrics/reward_metric.py
  class RewardMetric (line 12) | class RewardMetric(BaseMetric):
    method __init__ (line 17) | def __init__(self,
    method process (line 22) | def process(self, data_batch, data_samples: Sequence[dict]):
    method compute_metrics (line 49) | def compute_metrics(self, results: List):

FILE: xtuner-eval_niah/xtuner/model/dpo.py
  function create_reference_model (line 19) | def create_reference_model(model):
  class DPO (line 36) | class DPO(SupervisedFinetune):
    method __init__ (line 39) | def __init__(self,
    method _gather_masked_logits (line 55) | def _gather_masked_logits(self, logits, labels, mask):
    method get_logps (line 61) | def get_logps(
    method get_var_len_atten_logps (line 89) | def get_var_len_atten_logps(self, all_logits, all_ref_logits, labels,
    method _split_for_sequence_parallel (line 141) | def _split_for_sequence_parallel(data):
    method compute_loss (line 153) | def compute_loss(self, data, data_samples=None):

FILE: xtuner-eval_niah/xtuner/model/llava.py
  function convert_state_dict_to_hf (line 30) | def convert_state_dict_to_hf(state_dict, mapping):
  class LLaVAModel (line 42) | class LLaVAModel(BaseModel):
    method __init__ (line 44) | def __init__(self,
    method _parse_lora_config (line 121) | def _parse_lora_config(self, lora_config):
    method _prepare_llm_for_lora (line 127) | def _prepare_llm_for_lora(self,
    method _prepare_visual_encoder_for_lora (line 138) | def _prepare_visual_encoder_for_lora(self,
    method gradient_checkpointing_enable (line 147) | def gradient_checkpointing_enable(self):
    method activation_checkpointing_enable (line 150) | def activation_checkpointing_enable(self):
    method gradient_checkpointing_disable (line 155) | def gradient_checkpointing_disable(self):
    method activation_checkpointing_disable (line 158) | def activation_checkpointing_disable(self):
    method init_weights (line 163) | def init_weights(self):
    method state_dict (line 166) | def state_dict(self, *args, **kwargs):
    method _prepare_for_long_context_training (line 194) | def _prepare_for_long_context_training(cfg, llm_cfg,
    method _prepare_for_flash_attn (line 221) | def _prepare_for_flash_attn(cfg, llm_cfg):
    method _prepare_for_qlora_zero3 (line 250) | def _prepare_for_qlora_zero3(cfg):
    method _dispatch_lm_model_cfg (line 266) | def _dispatch_lm_model_cfg(self, cfg, max_position_embeddings=None):
    method _build_from_cfg_or_module (line 277) | def _build_from_cfg_or_module(self, cfg_or_mod):
    method forward (line 286) | def forward(self, data, data_samples=None, mode='loss'):
    method _forward (line 313) | def _forward(self, data, data_samples=None):
    method predict (line 319) | def predict(self, data, data_samples=None):
    method compute_loss (line 324) | def compute_loss(self, data, data_samples=None):
    method __getattr__ (line 329) | def __getattr__(self, name: str):
    method to_hf (line 335) | def to_hf(self,
    method to_xtuner_llava (line 352) | def to_xtuner_llava(self,
    method to_huggingface_llava (line 402) | def to_huggingface_llava(self,
    method to_official_llava (line 536) | def to_official_llava(self,

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/__init__.py
  function log_once (line 137) | def log_once(func):
  function dispatch_attn_forward (line 150) | def dispatch_attn_forward(model):
  function dispatch_varlen_attn_forward (line 178) | def dispatch_varlen_attn_forward(model):
  function dispatch_rmsnorm_forward (line 207) | def dispatch_rmsnorm_forward(model):
  function replace_rote (line 226) | def replace_rote(model):
  function dispatch_modules (line 254) | def dispatch_modules(model, use_varlen_attn=False):

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/attention.py
  function flash_attn_wo_mask (line 15) | def flash_attn_wo_mask(
  function flash_attn_w_mask (line 36) | def flash_attn_w_mask(
  function varlen_flash_attn (line 68) | def varlen_flash_attn(

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/baichuan.py
  function baichuan2_norm_head_forward (line 9) | def baichuan2_norm_head_forward(self, hidden_states):
  function rotate_half (line 14) | def rotate_half(x):
  function apply_rotary_pos_emb (line 21) | def apply_rotary_pos_emb(q, k, cos_, sin_, position_ids):
  function baichuan_7b_attn_forward (line 31) | def baichuan_7b_attn_forward(
  function baichuan_13b_attn_forward (line 75) | def baichuan_13b_attn_forward(

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/cohere.py
  class Cache (line 17) | class Cache:
  function cohere_attn_forward (line 21) | def cohere_attn_forward(

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/deepseek_v2.py
  function deepseek_attn_forward (line 19) | def deepseek_attn_forward(
  function deepseek_varlen_attn_forward (line 158) | def deepseek_varlen_attn_forward(

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/internlm.py
  class InternLMRotaryEmbedding (line 21) | class InternLMRotaryEmbedding(torch.nn.Module):
    method __init__ (line 23) | def __init__(self,
    method forward (line 43) | def forward(self, x, seq_len):
  function rotate_half (line 64) | def rotate_half(x):
  function apply_rotary_pos_emb (line 71) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids):
  function internlm_attn_forward (line 79) | def internlm_attn_forward(
  function internlm_varlen_attn_forward (line 140) | def internlm_varlen_attn_forward(

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/internlm2.py
  function rotate_half (line 16) | def rotate_half(x):
  function apply_rotary_pos_emb (line 23) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  function repeat_kv (line 31) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  function repeat_kv_bshd (line 49) | def repeat_kv_bshd(hidden_states: torch.Tensor, n_rep: int) -> torch.Ten...
  function internlm2_attn_forward (line 63) | def internlm2_attn_forward(
  function internlm2_varlen_attn_forward (line 174) | def internlm2_varlen_attn_forward(

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/llama.py
  class Cache (line 20) | class Cache:
  function repeat_kv_bshd (line 24) | def repeat_kv_bshd(hidden_states: torch.Tensor, n_rep: int) -> torch.Ten...
  function llama_attn_forward (line 38) | def llama_attn_forward(
  function llama_attn_forward_legacy (line 160) | def llama_attn_forward_legacy(
  function llama_varlen_attn_forward (line 290) | def llama_varlen_attn_forward(
  function llama_varlen_attn_forward_legacy (line 399) | def llama_varlen_attn_forward_legacy(

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/mistral.py
  class MistralRotaryEmbedding (line 32) | class MistralRotaryEmbedding(nn.Module):
    method __init__ (line 34) | def __init__(self,
    method _set_cos_sin_cache (line 53) | def _set_cos_sin_cache(self, seq_len, device, dtype):
    method forward (line 64) | def forward(self, x, seq_len=None):
  function repeat_kv_bshd (line 78) | def repeat_kv_bshd(hidden_states: torch.Tensor, n_rep: int) -> torch.Ten...
  function mistral_attn_forward (line 92) | def mistral_attn_forward(
  function mistral_varlen_attn_forward (line 241) | def mistral_varlen_attn_forward(

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/phi3.py
  class Cache (line 18) | class Cache:
  function repeat_kv (line 40) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  function rotate_half (line 59) | def rotate_half(x):
  function apply_rotary_pos_emb (line 67) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  function phi3_attn_forward (line 94) | def phi3_attn_forward(
  function phi3_varlen_attn_forward (line 261) | def phi3_varlen_attn_forward(

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/qwen2.py
  function qwen2_attn_forward (line 30) | def qwen2_attn_forward(
  function qwen2_varlen_attn_forward (line 177) | def qwen2_varlen_attn_forward(

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/triton_kernels/layer_norm.py
  function layer_norm_forward (line 6) | def layer_norm_forward(self, hidden_states):

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/triton_kernels/rms_norm.py
  function _rms_norm_fwd_fused (line 8) | def _rms_norm_fwd_fused(
  function _rms_norm_bwd_dx_fused (line 45) | def _rms_norm_bwd_dx_fused(
  function _rms_norm_bwd_dwdb (line 100) | def _rms_norm_bwd_dwdb(
  class RMSNorm (line 122) | class RMSNorm(torch.autograd.Function):
    method forward (line 125) | def forward(ctx, x, weight, eps):
    method backward (line 159) | def backward(ctx, dy):
  function rms_norm_forward (line 214) | def rms_norm_forward(self, hidden_states):

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/triton_kernels/rotary.py
  function rotary_kernel (line 11) | def rotary_kernel(
  function apply_rotary (line 163) | def apply_rotary(
  class ApplyRotaryEmb (line 265) | class ApplyRotaryEmb(torch.autograd.Function):
    method forward (line 268) | def forward(
    method backward (line 302) | def backward(ctx, do):

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/utils.py
  function _get_unpad_data (line 10) | def _get_unpad_data(attention_mask):
  function upad_qkv (line 23) | def upad_qkv(query_layer, key_layer, value_layer, attention_mask,

FILE: xtuner-eval_niah/xtuner/model/modules/dispatch/yi.py
  function rotate_half (line 8) | def rotate_half(x):
  function apply_rotary_pos_emb (line 15) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids):
  function repeat_kv (line 27) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  function yi_attn_forward (line 45) | def yi_attn_forward(

FILE: xtuner-eval_niah/xtuner/model/modules/projector/configuration_projector.py
  class ProjectorConfig (line 5) | class ProjectorConfig(PretrainedConfig):
    method __init__ (line 9) | def __init__(

FILE: xtuner-eval_niah/xtuner/model/modules/projector/modeling_projector.py
  class ProjectorModel (line 10) | class ProjectorModel(PreTrainedModel):
    method __init__ (line 16) | def __init__(self, config: ProjectorConfig) -> None:
    method enable_input_require_grads (line 35) | def enable_input_require_grads(self):
    method _set_gradient_checkpointing (line 42) | def _set_gradient_checkpointing(self, module, value=False):
    method forward (line 46) | def forward(self, x):

FILE: xtuner-eval_niah/xtuner/model/orpo.py
  class ORPO (line 17) | class ORPO(SupervisedFinetune):
    method __init__ (line 25) | def __init__(self, *args, beta=0.1, **kwargs):
    method _gather_masked_logits (line 29) | def _gather_masked_logits(self, logits, labels, mask):
    method get_logps (line 35) | def get_logps(
    method get_var_len_atten_logps (line 56) | def get_var_len_atten_logps(self, all_logits, average_log_prob, labels,
    method cross_entropy_loss (line 94) | def cross_entropy_loss(self, logits, labels):
    method odds_ratio_loss (line 106) | def odds_ratio_loss(
    method _split_for_sequence_parallel (line 127) | def _split_for_sequence_parallel(data):
    method compute_loss (line 139) | def compute_loss(self, data, data_samples=None):

FILE: xtuner-eval_niah/xtuner/model/reward.py
  function reduce_mean (line 35) | def reduce_mean(tensor):
  function smart_tokenizer_and_embedding_resize (line 44) | def smart_tokenizer_and_embedding_resize(
  class RewardModel (line 87) | class RewardModel(BaseModel):
    method __init__ (line 89) | def __init__(
    method gradient_checkpointing_enable (line 157) | def gradient_checkpointing_enable(self):
    method activation_checkpointing_enable (line 160) | def activation_checkpointing_enable(self):
    method gradient_checkpointing_disable (line 163) | def gradient_checkpointing_disable(self):
    method activation_checkpointing_disable (line 166) | def activation_checkpointing_disable(self):
    method _prepare_for_lora (line 169) | def _prepare_for_lora(self,
    method init_weights (line 182) | def init_weights(self):
    method _prepare_for_long_context_training (line 186) | def _prepare_for_long_context_training(cfg, llm_cfg,
    method _prepare_for_flash_attn (line 211) | def _prepare_for_flash_attn(cfg, llm_cfg):
    method _prepare_for_qlora_zero3 (line 240) | def _prepare_for_qlora_zero3(cfg):
    method _dispatch_lm_model_cfg (line 256) | def _dispatch_lm_model_cfg(self, cfg, max_position_embeddings=None):
    method _build_from_cfg_or_module (line 267) | def _build_from_cfg_or_module(self, cfg_or_mod):
    method forward (line 276) | def forward(self, data, data_samples=None, mode='loss'):
    method _forward (line 287) | def _forward(self, data, data_samples=None):
    method predict (line 292) | def predict(self, data, data_samples=None):
    method _split_for_sequence_parallel (line 299) | def _split_for_sequence_parallel(data):
    method compute_loss (line 311) | def compute_loss(self, data, labels=None):
    method ranking_loss (line 381) | def ranking_loss(self, chosen_logits, rejected_logits, avg_factor):
    method focal_loss (line 385) | def focal_loss(self, chosen_logits, rejected_logits, avg_factor):
    method log_barrier_penalty (line 394) | def log_barrier_penalty(self,
    method l2_penalty (line 408) | def l2_penalty(self, logits, avg_factor=1):
    method state_dict (line 411) | def state_dict(self, *args, **kwargs):
    method __getattr__ (line 418) | def __getattr__(self, name: str):
    method to_hf (line 424) | def to_hf(self,

FILE: xtuner-eval_niah/xtuner/model/sft.py
  function smart_tokenizer_and_embedding_resize (line 28) | def smart_tokenizer_and_embedding_resize(
  class SupervisedFinetune (line 71) | class SupervisedFinetune(BaseModel):
    method __init__ (line 73) | def __init__(self,
    method gradient_checkpointing_enable (line 122) | def gradient_checkpointing_enable(self):
    method activation_checkpointing_enable (line 125) | def activation_checkpointing_enable(self):
    method gradient_checkpointing_disable (line 128) | def gradient_checkpointing_disable(self):
    method activation_checkpointing_disable (line 131) | def activation_checkpointing_disable(self):
    method _prepare_for_lora (line 134) | def _prepare_for_lora(self,
    method init_weights (line 147) | def init_weights(self):
    method _prepare_for_long_context_training (line 151) | def _prepare_for_long_context_training(cfg, llm_cfg,
    method _prepare_for_flash_attn (line 176) | def _prepare_for_flash_attn(cfg, llm_cfg):
    method _prepare_for_qlora_zero3 (line 206) | def _prepare_for_qlora_zero3(cfg):
    method _dispatch_lm_model_cfg (line 222) | def _dispatch_lm_model_cfg(self, cfg, max_position_embeddings=None):
    method _build_from_cfg_or_module (line 233) | def _build_from_cfg_or_module(self, cfg_or_mod):
    method forward (line 242) | def forward(self, data, data_samples=None, mode='loss'):
    method _forward (line 253) | def _forward(self, data, data_samples=None):
    method predict (line 259) | def predict(self, data, data_samples=None):
    method _split_for_sequence_parallel (line 265) | def _split_for_sequence_parallel(data):
    method _compute_sequence_parallel_loss (line 277) | def _compute_sequence_parallel_loss(self, data):
    method compute_loss (line 287) | def compute_loss(self, data, data_samples=None):
    method state_dict (line 295) | def state_dict(self, *args, **kwargs):
    method __getattr__ (line 302) | def __getattr__(self, name: str):
    method to_hf (line 308) | def to_hf(self,

FILE: xtuner-eval_niah/xtuner/model/transformers_models/deepseek_v2/configuration_deepseek.py
  class DeepseekV2Config (line 11) | class DeepseekV2Config(PretrainedConfig):
    method __init__ (line 124) | def __init__(

FILE: xtuner-eval_niah/xtuner/model/transformers_models/deepseek_v2/modeling_deepseek.py
  function _get_unpad_data (line 76) | def _get_unpad_data(attention_mask):
  class DeepseekV2RMSNorm (line 89) | class DeepseekV2RMSNorm(nn.Module):
    method __init__ (line 91) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 97) | def forward(self, hidden_states):
  class DeepseekV2RotaryEmbedding (line 109) | class DeepseekV2RotaryEmbedding(nn.Module):
    method __init__ (line 111) | def __init__(self,
    method _set_cos_sin_cache (line 134) | def _set_cos_sin_cache(self, seq_len, device, dtype):
    method forward (line 147) | def forward(self, x, seq_len=None):
  class DeepseekV2LinearScalingRotaryEmbedding (line 160) | class DeepseekV2LinearScalingRotaryEmbedding(DeepseekV2RotaryEmbedding):
    method __init__ (line 166) | def __init__(
    method _set_cos_sin_cache (line 177) | def _set_cos_sin_cache(self, seq_len, device, dtype):
  class DeepseekV2DynamicNTKScalingRotaryEmbedding (line 193) | class DeepseekV2DynamicNTKScalingRotaryEmbedding(DeepseekV2RotaryEmbeddi...
    method __init__ (line 199) | def __init__(
    method _set_cos_sin_cache (line 210) | def _set_cos_sin_cache(self, seq_len, device, dtype):
  function yarn_find_correction_dim (line 236) | def yarn_find_correction_dim(num_rotations,
  function yarn_find_correction_range (line 246) | def yarn_find_correction_range(low_rot,
  function yarn_get_mscale (line 258) | def yarn_get_mscale(scale=1, mscale=1):
  function yarn_linear_ramp_mask (line 264) | def yarn_linear_ramp_mask(min, max, dim):
  class DeepseekV2YarnRotaryEmbedding (line 273) | class DeepseekV2YarnRotaryEmbedding(DeepseekV2RotaryEmbedding):
    method __init__ (line 275) | def __init__(
    method _set_cos_sin_cache (line 296) | def _set_cos_sin_cache(self, seq_len, device, dtype):
  function rotate_half (line 336) | def rotate_half(x):
  function apply_rotary_pos_emb (line 344) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
  class DeepseekV2MLP (line 379) | class DeepseekV2MLP(nn.Module):
    method __init__ (line 381) | def __init__(self, config, hidden_size=None, intermediate_size=None):
    method forward (line 397) | def forward(self, x):
  class MoEGate (line 403) | class MoEGate(nn.Module):
    method __init__ (line 405) | def __init__(self, config):
    method reset_parameters (line 425) | def reset_parameters(self) -> None:
    method forward (line 430) | def forward(self, hidden_states):
  class AddAuxiliaryLoss (line 502) | class AddAuxiliaryLoss(torch.autograd.Function):
    method forward (line 507) | def forward(ctx, x, loss):
    method backward (line 514) | def backward(ctx, grad_output):
  class ExpertShard (line 522) | class ExpertShard(nn.Module):
    method __init__ (line 524) | def __init__(self, config, shard_idx, expert_in_one_shard=10):
    method reset_parameters (line 539) | def reset_parameters(self) -> None:
    method expert_forward (line 545) | def expert_forward(self, current_state, expert_idx):
    method forward (line 555) | def forward(self, hidden_states, flat_topk_idx, y):
  class DeepseekV2MoEShard (line 563) | class DeepseekV2MoEShard(nn.Module):
    method __init__ (line 566) | def __init__(self, config):
    method forward (line 597) | def forward(self, hidden_states):
  class DeepseekV2MoE (line 623) | class DeepseekV2MoE(nn.Module):
    method __init__ (line 626) | def __init__(self, config):
    method forward (line 658) | def forward(self, hidden_states):
    method moe_infer (line 684) | def moe_infer(self, x, topk_ids, topk_weight):
  function repeat_kv (line 754) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class DeepseekV2Attention (line 773) | class DeepseekV2Attention(nn.Module):
    method __init__ (line 776) | def __init__(self,
    method _init_rope (line 845) | def _init_rope(self):
    method _shape (line 890) | def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
    method forward (line 894) | def forward(
  class DeepseekV2FlashAttention2 (line 999) | class DeepseekV2FlashAttention2(DeepseekV2Attention):
    method __init__ (line 1008) | def __init__(self, *args, **kwargs):
    method forward (line 1017) | def forward(
    method _flash_attention_forward (line 1146) | def _flash_attention_forward(
    method _upad_input (line 1224) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class DeepseekV2DecoderLayer (line 1278) | class DeepseekV2DecoderLayer(nn.Module):
    method __init__ (line 1280) | def __init__(self, config: DeepseekV2Config, layer_idx: int):
    method forward (line 1305) | def forward(
  function _load_pretrained_model (line 1367) | def _load_pretrained_model(
  class DeepseekV2PreTrainedModel (line 1421) | class DeepseekV2PreTrainedModel(PreTrainedModel):
    method _init_weights (line 1431) | def _init_weights(self, module):
    method from_pretrained (line 1443) | def from_pretrained(cls, pretrained_model_name_or_path, *args, **kwargs):
  class DeepseekV2Model (line 1529) | class DeepseekV2Model(DeepseekV2PreTrainedModel):
    method __init__ (line 1537) | def __init__(self, config: DeepseekV2Config):
    method get_input_embeddings (line 1557) | def get_input_embeddings(self):
    method set_input_embeddings (line 1560) | def set_input_embeddings(self, value):
    method forward (line 1564) | def forward(
  class DeepseekV2ForCausalLM (line 1718) | class DeepseekV2ForCausalLM(DeepseekV2PreTrainedModel):
    method __init__ (line 1721) | def __init__(self, config):
    method get_input_embeddings (line 1731) | def get_input_embeddings(self):
    method set_input_embeddings (line 1734) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1737) | def get_output_embeddings(self):
    method set_output_embeddings (line 1740) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1743) | def set_decoder(self, decoder):
    method get_decoder (line 1746) | def get_decoder(self):
    method forward (line 1752) | def forward(
    method prepare_inputs_for_generation (line 1842) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1901) | def _reorder_cache(past_key_values, beam_idx):
  class DeepseekV2ForSequenceClassification (line 1925) | class DeepseekV2ForSequenceClassification(DeepseekV2PreTrainedModel):
    method __init__ (line 1927) | def __init__(self, config):
    method get_input_embeddings (line 1936) | def get_input_embeddings(self):
    method set_input_embeddings (line 1939) | def set_input_embeddings(self, value):
    method forward (line 1943) | def forward(

FILE: xtuner-eval_niah/xtuner/model/transformers_models/deepseek_v2/tokenization_deepseek_fast.py
  class DeepseekTokenizerFast (line 6) | class DeepseekTokenizerFast(LlamaTokenizerFast):
    method convert_ids_to_tokens (line 8) | def convert_ids_to_tokens(
    method _convert_id_to_token (line 35) | def _convert_id_to_token(self, index: int) -> Optional[str]:

FILE: xtuner-eval_niah/xtuner/model/transformers_models/mixtral/configuration_mixtral.py
  class MixtralConfig (line 22) | class MixtralConfig(PretrainedConfig):
    method __init__ (line 114) | def __init__(

FILE: xtuner-eval_niah/xtuner/model/transformers_models/mixtral/modeling_mixtral.py
  function load_balancing_loss_func (line 56) | def load_balancing_loss_func(
  function _get_unpad_data (line 138) | def _get_unpad_data(attention_mask):
  class MixtralRMSNorm (line 152) | class MixtralRMSNorm(nn.Module):
    method __init__ (line 154) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 160) | def forward(self, hidden_states):
  class MixtralRotaryEmbedding (line 170) | class MixtralRotaryEmbedding(nn.Module):
    method __init__ (line 172) | def __init__(self,
    method _set_cos_sin_cache (line 194) | def _set_cos_sin_cache(self, seq_len, device, dtype):
    method forward (line 208) | def forward(self, x, seq_len=None):
  function rotate_half (line 221) | def rotate_half(x):
  function apply_rotary_pos_emb (line 229) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
  function repeat_kv (line 258) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class MixtralAttention (line 277) | class MixtralAttention(nn.Module):
    method __init__ (line 284) | def __init__(self, config: MixtralConfig, layer_idx: Optional[int] = N...
    method _shape (line 327) | def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
    method forward (line 331) | def forward(
  class MixtralFlashAttention2 (line 416) | class MixtralFlashAttention2(MixtralAttention):
    method __init__ (line 426) | def __init__(self, *args, **kwargs):
    method forward (line 435) | def forward(
    method _flash_attention_forward (line 567) | def _flash_attention_forward(
    method _upad_input (line 670) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class MixtralSdpaAttention (line 722) | class MixtralSdpaAttention(MixtralAttention):
    method forward (line 732) | def forward(
  class MixtralBlockSparseTop2MLP (line 825) | class MixtralBlockSparseTop2MLP(nn.Module):
    method __init__ (line 827) | def __init__(self, config: MixtralConfig):
    method forward (line 838) | def forward(self, hidden_states):
  class MixtralSparseMoeBlock (line 845) | class MixtralSparseMoeBlock(nn.Module):
    method __init__ (line 856) | def __init__(self, config):
    method forward (line 873) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
  class ExpertShard (line 921) | class ExpertShard(nn.Module):
    method __init__ (line 923) | def __init__(self, config, expert_in_one_shard=1):
    method forward (line 934) | def forward(self, hidden_states, expert_mask, routing_weights,
  class MixtralSparseShardMoeBlock (line 955) | class MixtralSparseShardMoeBlock(nn.Module):
    method __init__ (line 957) | def __init__(self, config):
    method forward (line 981) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
  class MixtralDecoderLayer (line 1021) | class MixtralDecoderLayer(nn.Module):
    method __init__ (line 1023) | def __init__(self, config: MixtralConfig, layer_idx: int):
    method forward (line 1044) | def forward(
  function _load_pretrained_model (line 1107) | def _load_pretrained_model(
  class MixtralPreTrainedModel (line 1162) | class MixtralPreTrainedModel(PreTrainedModel):
    method _init_weights (line 1172) | def _init_weights(self, module):
    method from_pretrained (line 1184) | def from_pretrained(cls, pretrained_model_name_or_path, *args, **kwargs):
  class MixtralModel (line 1268) | class MixtralModel(MixtralPreTrainedModel):
    method __init__ (line 1276) | def __init__(self, config: MixtralConfig):
    method get_input_embeddings (line 1294) | def get_input_embeddings(self):
    method set_input_embeddings (line 1297) | def set_input_embeddings(self, value):
    method forward (line 1302) | def forward(
  class MixtralForCausalLM (line 1475) | class MixtralForCausalLM(MixtralPreTrainedModel):
    method __init__ (line 1478) | def __init__(self, config):
    method get_input_embeddings (line 1490) | def get_input_embeddings(self):
    method set_input_embeddings (line 1493) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1496) | def get_output_embeddings(self):
    method set_output_embeddings (line 1499) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1502) | def set_decoder(self, decoder):
    method get_decoder (line 1505) | def get_decoder(self):
    method forward (line 1512) | def forward(
    method prepare_inputs_for_generation (line 1621) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1683) | def _reorder_cache(past_key_values, beam_idx):
  class MixtralForSequenceClassification (line 1708) | class MixtralForSequenceClassification(MixtralPreTrainedModel):
    method __init__ (line 1710) | def __init__(self, config):
    method get_input_embeddings (line 1719) | def get_input_embeddings(self):
    method set_input_embeddings (line 1722) | def set_input_embeddings(self, value):
    method forward (line 1726) | def forward(

FILE: xtuner-eval_niah/xtuner/model/utils.py
  function set_obj_dtype (line 14) | def set_obj_dtype(d):
  function try_build_module (line 20) | def try_build_module(cfg):
  function traverse_dict (line 33) | def traverse_dict(d):
  function find_all_linear_names (line 47) | def find_all_linear_names(model):
  class LoadWoInit (line 61) | class LoadWoInit:
    method __init__ (line 64) | def __init__(self):
    method __enter__ (line 73) | def __enter__(self, *args, **kwargs):
    method __exit__ (line 82) | def __exit__(self, *args, **kwargs):
  function get_peft_model_state_dict (line 92) | def get_peft_model_state_dict(model, state_dict=None, adapter_name='defa...
  function prepare_inputs_labels_for_multimodal (line 138) | def prepare_inputs_labels_for_multimodal(
  function make_inputs_require_grad (line 293) | def make_inputs_require_grad(module, input, output):
  function guess_load_checkpoint (line 297) | def guess_load_checkpoint(pth_model):

FILE: xtuner-eval_niah/xtuner/parallel/sequence/attention.py
  function pre_process_for_sequence_parallel_attn (line 16) | def pre_process_for_sequence_parallel_attn(query_states,
  function post_process_for_sequence_parallel_attn (line 96) | def post_process_for_sequence_parallel_attn(attn_output,
  function sequence_parallel_wrapper (line 130) | def sequence_parallel_wrapper(local_attn):

FILE: xtuner-eval_niah/xtuner/parallel/sequence/comm.py
  function _all_to_all (line 9) | def _all_to_all(
  class _AllToAll (line 25) | class _AllToAll(torch.autograd.Function):
    method forward (line 36) | def forward(ctx: Any, input: Tensor, sp_group: dist.ProcessGroup,
    method backward (line 47) | def backward(ctx: Any, grad_output: Tensor) -> Tuple:
  function all_to_all (line 63) | def all_to_all(
  function split_for_sequence_parallel (line 90) | def split_for_sequence_parallel(input, dim: int, sp_group: dist.ProcessG...
  function gather_for_sequence_parallel (line 117) | def gather_for_sequence_parallel(input, dim: int, sp_group: dist.Process...
  class _GatherForwardSplitBackward (line 144) | class _GatherForwardSplitBackward(torch.autograd.Function):
    method forward (line 152) | def forward(ctx, input, dim, sp_group, grad_scale):
    method backward (line 159) | def backward(ctx, grad_output):
  class _SplitForwardGatherBackward (line 169) | class _SplitForwardGatherBackward(torch.autograd.Function):
    method forward (line 177) | def forward(ctx, input, dim, sp_group, grad_scale):
    method backward (line 184) | def backward(ctx, grad_output):
  function split_forward_gather_backward (line 193) | def split_forward_gather_backward(input, dim, sp_group, grad_scale=None):
  function gather_forward_split_backward (line 207) | def gather_forward_split_backward(input, dim, sp_group, grad_scale=None):

FILE: xtuner-eval_niah/xtuner/parallel/sequence/data_collate.py
  function pad_for_sequence_parallel (line 7) | def pad_for_sequence_parallel(tensor, padding_value, dim=-1):
  function pad_cumulative_len_for_sequence_parallel (line 26) | def pad_cumulative_len_for_sequence_parallel(cumulative_len):

FILE: xtuner-eval_niah/xtuner/parallel/sequence/reduce_loss.py
  class _ReduceLoss (line 7) | class _ReduceLoss(torch.autograd.Function):
    method forward (line 10) | def forward(ctx, mean_loss, loss_scale, process_group):
    method backward (line 22) | def backward(ctx, grad_output):
  function reduce_sequence_parallel_loss (line 26) | def reduce_sequence_parallel_loss(mean_loss,

FILE: xtuner-eval_niah/xtuner/parallel/sequence/sampler.py
  class SequenceParallelSampler (line 12) | class SequenceParallelSampler(DefaultSampler):
    method __init__ (line 14) | def __init__(self,

FILE: xtuner-eval_niah/xtuner/parallel/sequence/setup_distributed.py
  function init_sequence_parallel (line 17) | def init_sequence_parallel(sequence_parallel_size: int = 1):
  function init_inner_sequence_parallel (line 56) | def init_inner_sequence_parallel(inner_sequence_parallel_size: int = 1):
  function is_inner_sequence_parallel_initialized (line 82) | def is_inner_sequence_parallel_initialized():
  function get_inner_sequence_parallel_group (line 86) | def get_inner_sequence_parallel_group():
  function get_inner_sequence_parallel_world_size (line 90) | def get_inner_sequence_parallel_world_size():
  function get_inner_sequence_parallel_rank (line 102) | def get_inner_sequence_parallel_rank():
  function get_sequence_parallel_group (line 114) | def get_sequence_parallel_group():
  function get_sequence_parallel_world_size (line 119) | def get_sequence_parallel_world_size():
  function get_sequence_parallel_rank (line 132) | def get_sequence_parallel_rank():
  function get_data_parallel_group (line 145) | def get_data_parallel_group():
  function get_data_parallel_world_size (line 152) | def get_data_parallel_world_size():
  function get_data_parallel_rank (line 165) | def get_data_parallel_rank():

FILE: xtuner-eval_niah/xtuner/tools/chat.py
  function remove_prefix (line 26) | def remove_prefix(state_dict, prefix):
  function parse_args (line 37) | def parse_args():
  function get_input (line 131) | def get_input():
  function main (line 146) | def main():

FILE: xtuner-eval_niah/xtuner/tools/check_custom_dataset.py
  function parse_args (line 13) | def parse_args():
  function is_standard_format (line 22) | def is_standard_format(dataset):
  function main (line 40) | def main():

FILE: xtuner-eval_niah/xtuner/tools/copy_cfg.py
  function parse_args (line 11) | def parse_args():
  function add_copy_suffix (line 19) | def add_copy_suffix(string):
  function main (line 24) | def main():

FILE: xtuner-eval_niah/xtuner/tools/data_preprocess/arxiv.py
  function parse_args (line 7) | def parse_args():
  function has_intersection (line 25) | def has_intersection(list1, list2):
  function read_json_file (line 31) | def read_json_file(file_path):
  function main (line 43) | def main():

FILE: xtuner-eval_niah/xtuner/tools/data_preprocess/convert_refcoco.py
  function parse_args (line 8) | def parse_args():

FILE: xtuner-eval_niah/xtuner/tools/eval_refcoco.py
  function merge_outputs (line 31) | def merge_outputs(otuputs):
  function master_print (line 44) | def master_print(msg):
  function parse_args (line 48) | def parse_args():
  function eval_iou (line 110) | def eval_iou(answers):
  function build_model (line 143) | def build_model(args):
  function generate (line 233) | def generate(
  function main (line 299) | def main():

FILE: xtuner-eval_niah/xtuner/tools/get_data_order.py
  function parse_args (line 6) | def parse_args():
  function save_data_order (line 18) | def save_data_order(data_folder, save_folder, file_type='.bin'):

FILE: xtuner-eval_niah/xtuner/tools/list_cfg.py
  function parse_args (line 7) | def parse_args():
  function main (line 15) | def main(pattern=None):

FILE: xtuner-eval_niah/xtuner/tools/list_dataset_format.py
  function main (line 5) | def main():

FILE: xtuner-eval_niah/xtuner/tools/log_dataset.py
  function parse_args (line 9) | def parse_args():
  function main (line 22) | def main():

FILE: xtuner-eval_niah/xtuner/tools/mmbench.py
  function parse_args (line 38) | def parse_args():
  function master_print (line 95) | def master_print(msg):
  class MMBenchDataset (line 99) | class MMBenchDataset(Dataset):
    method __init__ (line 116) | def __init__(self, data_file):
    method get_image (line 122) | def get_image(self, image):
    method __len__ (line 130) | def __len__(self):
    method __getitem__ (line 133) | def __getitem__(self, idx):
    method load_from_df (line 166) | def load_from_df(self, idx, key):
    method eval_result (line 173) | def eval_result(self, result_df, show=True):
  function main (line 271) | def main():

FILE: xtuner-eval_niah/xtuner/tools/model_converters/merge.py
  function parse_args (line 12) | def parse_args():
  function main (line 43) | def main():

FILE: xtuner-eval_niah/xtuner/tools/model_converters/modeling_internlm2_reward/configuration_internlm2.py
  class InternLM2Config (line 28) | class InternLM2Config(PretrainedConfig):
    method __init__ (line 78) | def __init__(  # pylint: disable=W0102
    method _rope_scaling_validation (line 135) | def _rope_scaling_validation(self):

FILE: xtuner-eval_niah/xtuner/tools/model_converters/modeling_internlm2_reward/modeling_internlm2.py
  function _import_flash_attn (line 56) | def _import_flash_attn():
  function _get_unpad_data (line 68) | def _get_unpad_data(attention_mask):
  function _make_causal_mask (line 81) | def _make_causal_mask(
  function _expand_mask (line 99) | def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Option...
  class InternLM2RMSNorm (line 114) | class InternLM2RMSNorm(nn.Module):
    method __init__ (line 115) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 123) | def forward(self, hidden_states):
  class InternLM2RotaryEmbedding (line 132) | class InternLM2RotaryEmbedding(nn.Module):
    method __init__ (line 133) | def __init__(self, dim, max_position_embeddings=2048, base=10000, devi...
    method _set_cos_sin_cache (line 147) | def _set_cos_sin_cache(self, seq_len, device, dtype):
    method forward (line 157) | def forward(self, x, seq_len=None):
  class InternLM2LinearScalingRotaryEmbedding (line 169) | class InternLM2LinearScalingRotaryEmbedding(InternLM2RotaryEmbedding):
    method __init__ (line 172) | def __init__(self, dim, max_position_embeddings=2048, base=10000, devi...
    method _set_cos_sin_cache (line 176) | def _set_cos_sin_cache(self, seq_len, device, dtype):
  class InternLM2DynamicNTKScalingRotaryEmbedding (line 189) | class InternLM2DynamicNTKScalingRotaryEmbedding(InternLM2RotaryEmbedding):
    method __init__ (line 194) | def __init__(self, dim, max_position_embeddings=2048, base=10000, devi...
    method _set_cos_sin_cache (line 198) | def _set_cos_sin_cache(self, seq_len, device, dtype):
  function rotate_half (line 218) | def rotate_half(x):
  function apply_rotary_pos_emb (line 226) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
  class InternLM2MLP (line 235) | class InternLM2MLP(nn.Module):
    method __init__ (line 236) | def __init__(self, config):
    method forward (line 246) | def forward(self, x):
  function repeat_kv (line 253) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class InternLM2Attention (line 266) | class InternLM2Attention(nn.Module):
    method __init__ (line 269) | def __init__(self, config: InternLM2Config):
    method _init_rope (line 295) | def _init_rope(self):
    method _shape (line 323) | def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
    method forward (line 326) | def forward(
  class InternLM2FlashAttention2 (line 415) | class InternLM2FlashAttention2(InternLM2Attention):
    method forward (line 422) | def forward(
    method _flash_attention_forward (line 494) | def _flash_attention_forward(
    method _unpad_input (line 548) | def _unpad_input(self, query_layer, key_layer, value_layer, attention_...
  class InternLM2DecoderLayer (line 593) | class InternLM2DecoderLayer(nn.Module):
    method __init__ (line 594) | def __init__(self, config: InternLM2Config):
    method forward (line 604) | def forward(
  class InternLM2PreTrainedModel (line 689) | class InternLM2PreTrainedModel(PreTrainedModel):
    method _init_weights (line 696) | def _init_weights(self, module):
  class InternLM2Model (line 778) | class InternLM2Model(InternLM2PreTrainedModel):
    method __init__ (line 788) | def __init__(self, config: InternLM2Config):
    method get_input_embeddings (line 803) | def get_input_embeddings(self):
    method set_input_embeddings (line 806) | def set_input_embeddings(self, value):
    method _prepare_decoder_attention_mask (line 809) | def _prepare_decoder_attention_mask(self, attention_mask, input_shape,...
    method forward (line 833) | def forward(
  class InternLM2ForCausalLM (line 967) | class InternLM2ForCausalLM(InternLM2PreTrainedModel):
    method __init__ (line 972) | def __init__(self, config):
    method get_input_embeddings (line 981) | def get_input_embeddings(self):
    method set_input_embeddings (line 984) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 987) | def get_output_embeddings(self):
    method set_output_embeddings (line 990) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 993) | def set_decoder(self, decoder):
    method get_decoder (line 996) | def get_decoder(self):
    method forward (line 1001) | def forward(
    method prepare_inputs_for_generation (line 1088) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1128) | def _reorder_cache(past_key_values, beam_idx):
    method build_inputs (line 1136) | def build_inputs(self, tokenizer, query: str, history: List[Tuple[str,...
    method chat (line 1149) | def chat(
    method stream_chat (line 1185) | def stream_chat(
  class InternLM2ForRewardModel (line 1271) | class InternLM2ForRewardModel(InternLM2PreTrainedModel):
    method __init__ (line 1276) | def __init__(self, config):
    method get_input_embeddings (line 1286) | def get_input_embeddings(self):
    method set_input_embeddings (line 1289) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1292) | def get_output_embeddings(self):
    method set_output_embeddings (line 1295) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1298) | def set_decoder(self, decoder):
    method get_decoder (line 1301) | def get_decoder(self):
    method forward (line 1306) | def forward(
    method get_score (line 1386) | def get_score(
    method get_scores (line 1403) | def get_scores(
    method compare (line 1428) | def compare(
    method rank (line 1444) | def rank(
  class InternLM2ForSequenceClassification (line 1474) | class InternLM2ForSequenceClassification(InternLM2PreTrainedModel):
    method __init__ (line 1475) | def __init__(self, config):
    method get_input_embeddings (line 1484) | def get_input_embeddings(self):
    method set_input_embeddings (line 1487) | def set_input_embeddings(self, value):
    method forward (line 1491) | def forward(

FILE: xtuner-eval_niah/xtuner/tools/model_converters/pth_to_hf.py
  function parse_args (line 20) | def parse_args():
  function main (line 60) | def main():

FILE: xtuner-eval_niah/xtuner/tools/model_converters/split.py
  function parse_args (line 13) | def parse_args():
  function main (line 22) | def main():

FILE: xtuner-eval_niah/xtuner/tools/plugins/api.py
  function plugins_api (line 5) | def plugins_api(input_str,

FILE: xtuner-eval_niah/xtuner/tools/plugins/calculate.py
  function Calculate (line 5) | def Calculate(expression):

FILE: xtuner-eval_niah/xtuner/tools/plugins/search.py
  function parse_results (line 15) | def parse_results(results, k=10):
  function search (line 26) | def search(api_key, search_term, **kwargs):
  function Search (line 47) | def Search(q, k=10):

FILE: xtuner-eval_niah/xtuner/tools/plugins/solve.py
  function Solve (line 11) | def Solve(equations_str):

FILE: xtuner-eval_niah/xtuner/tools/process_untokenized_datasets.py
  function parse_args (line 15) | def parse_args():
  function modify_config (line 23) | def modify_config(config, dataset_save_folder):
  function process_untokenized_dataset (line 45) | def process_untokenized_dataset(config):

FILE: xtuner-eval_niah/xtuner/tools/process_untokenized_datasets_legacy.py
  function parse_args (line 42) | def parse_args():
  function process_one (line 78) | def process_one(fp,
  function process_untokenized_dataset (line 109) | def process_untokenized_dataset(folder,

FILE: xtuner-eval_niah/xtuner/tools/process_untokenized_llava_data.py
  function parse_args (line 13) | def parse_args():
  function build_llava_dataset (line 21) | def build_llava_dataset(config):

FILE: xtuner-eval_niah/xtuner/tools/test.py
  function parse_args (line 16) | def parse_args():
  function register_function (line 45) | def register_function(cfg_dict):
  function main (line 60) | def main():

FILE: xtuner-eval_niah/xtuner/tools/tokenize_ftdp_datasets.py
  function chatml_format (line 95) | def chatml_format(
  function write_bin_meta_bin (line 271) | def write_bin_meta_bin(path, dataset_name, filename, samples):
  function tokenize_and_save (line 320) | def tokenize_and_save(tokenizer, processed_dir, tokenized_dir):
  function tokenizer_add_special_tokens (line 389) | def tokenizer_add_special_tokens(tokenizer):
  function save_new_tokenizer (line 397) | def save_new_tokenizer(tokenizer, save_dir):
  function parse_args (line 402) | def parse_args():
  function main (line 419) | def main():

FILE: xtuner-eval_niah/xtuner/tools/train.py
  function parse_args (line 29) | def parse_args():
  function register_function (line 65) | def register_function(cfg_dict):
  function check_cfg (line 80) | def check_cfg(cfg, args):
  function main (line 127) | def main():

FILE: xtuner-eval_niah/xtuner/tools/utils.py
  function get_base_model (line 13) | def get_base_model(model):
  function get_streamer (line 21) | def get_streamer(model):
  class DecodeOutputStreamer (line 41) | class DecodeOutputStreamer(BaseStreamer):
    method __init__ (line 44) | def __init__(self, tokenizer, skip_prompt=True) -> None:
    method _decode_with_raw_id (line 59) | def _decode_with_raw_id(self, value):
    method _decode_fallback (line 74) | def _decode_fallback(self, value):
    method put (line 83) | def put(self, value):
    method end (line 94) | def end(self):
  class NoSpaceStreamer (line 100) | class NoSpaceStreamer(DecodeOutputStreamer):
    method __init__ (line 102) | def __init__(self, tokenizer, skip_prompt=True) -> None:
    method decode (line 113) | def decode(self, value):
  function get_stop_criteria (line 123) | def get_stop_criteria(
  function auto_dtype_of_deepspeed_config (line 133) | def auto_dtype_of_deepspeed_config(ds_config):
  function is_cn_string (line 152) | def is_cn_string(s):
  function get_seed_from_checkpoint (line 158) | def get_seed_from_checkpoint(pth_model):

FILE: xtuner-eval_niah/xtuner/utils/fileio.py
  function patch_func (line 8) | def patch_func(module, fn_name_to_wrap):
  function patch_fileio (line 23) | def patch_fileio(global_vars=None):
  function patch_hf_auto_from_pretrained (line 231) | def patch_hf_auto_from_pretrained(petrel_hub):
  function patch_hf_save_pretrained (line 279) | def patch_hf_save_pretrained():
  function patch_deepspeed_engine (line 320) | def patch_deepspeed_engine():

FILE: xtuner-eval_niah/xtuner/utils/handle_moe_load_and_save.py
  function print_on_rank0 (line 32) | def print_on_rank0(info):
  function get_expert_num_per_shard (line 37) | def get_expert_num_per_shard(model):
  function mix_sort (line 43) | def mix_sort(expert_name):
  function _get_merged_param_name (line 49) | def _get_merged_param_name(origin_param_name, expert_num_per_shard):
  function _merge_experts_weight (line 60) | def _merge_experts_weight(state_dict, expert_num_per_shard, order_mapping):
  function load_state_dict_into_model (line 99) | def load_state_dict_into_model(model_to_load, pretrained_model_path):
  function _get_origin_param_name (line 172) | def _get_origin_param_name(merged_param_name, expert_num_per_shard, is_w...
  function _split_param (line 197) | def _split_param(merged_param, is_w1w3):
  function get_origin_state_dict (line 207) | def get_origin_state_dict(state_dict, model):

FILE: xtuner-eval_niah/xtuner/utils/stop_criteria.py
  class StopWordStoppingCriteria (line 5) | class StopWordStoppingCriteria(StoppingCriteria):
    method __init__ (line 8) | def __init__(self, tokenizer, stop_word):
    method __call__ (line 13) | def __call__(self, input_ids, *args, **kwargs) -> bool:

FILE: xtuner-eval_niah/xtuner/utils/zero_to_any_dtype.py
  class zero_model_state (line 43) | class zero_model_state:
  function atoi (line 60) | def atoi(text):
  function natural_keys (line 64) | def natural_keys(text):
  function get_model_state_file (line 71) | def get_model_state_file(checkpoint_dir, zero_stage):
  function get_checkpoint_files (line 88) | def get_checkpoint_files(checkpoint_dir, glob_pattern):
  function get_optim_files (line 102) | def get_optim_files(checkpoint_dir):
  function get_model_state_files (line 106) | def get_model_state_files(checkpoint_dir):
  function parse_model_states (line 110) | def parse_model_states(files, dtype=DEFAULT_DTYPE):
  function parse_optim_states (line 161) | def parse_optim_states(files, ds_checkpoint_dir, dtype=DEFAULT_DTYPE):
  function _get_state_dict_from_zero_checkpoint (line 217) | def _get_state_dict_from_zero_checkpoint(ds_checkpoint_dir,
  function _zero2_merge_frozen_params (line 250) | def _zero2_merge_frozen_params(state_dict, zero_model_states):
  function _has_callable (line 286) | def _has_callable(obj, fn):
  function _zero2_merge_trainable_params (line 291) | def _zero2_merge_trainable_params(state_dict, world_size, flat_groups,
  function _get_state_dict_from_zero2_checkpoint (line 378) | def _get_state_dict_from_zero2_checkpoint(world_size, flat_groups,
  function zero3_partitioned_param_info (line 403) | def zero3_partitioned_param_info(unpartitioned_numel, world_size):
  function _zero3_merge_frozen_params (line 410) | def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
  function _zero3_merge_trainable_params (line 457) | def _zero3_merge_trainable_params(state_dict, world_size, flat_groups,
  function _get_state_dict_from_zero3_checkpoint (line 536) | def _get_state_dict_from_zero3_checkpoint(world_size, flat_groups,
  function get_state_dict_from_zero_checkpoint (line 561) | def get_state_dict_from_zero_checkpoint(checkpoint_dir,
  function convert_zero_checkpoint_to_state_dict (line 622) | def convert_zero_checkpoint_to_state_dict(checkpoint_dir,
  function load_state_dict_from_zero_checkpoint (line 648) | def load_state_dict_from_zero_checkpoint(model,

FILE: xtuner-eval_niah/xtuner/version.py
  function parse_version_info (line 6) | def parse_version_info(version_str):

FILE: xtuner-train_internvideo2_5/setup.py
  function readme (line 6) | def readme():
  function get_version (line 15) | def get_version():
  function parse_requirements (line 21) | def parse_requirements(fname='requirements.txt', with_version=True):

FILE: xtuner-train_internvideo2_5/unify_internvl2_train_r16.py
  function parse_args (line 48) | def parse_args():
  class LazyInternVL2Dataset (line 219) | class LazyInternVL2Dataset(BaseOrigDataset):
    method __init__ (line 220) | def __init__(self, data_name, data, model_name,
    method calc_group_len (line 283) | def calc_group_len(self):
    method pre_tokenize_fn_for_pack (line 324) | def pre_tokenize_fn_for_pack(self, data_item):
    method _get_preprocess_function (line 338) | def _get_preprocess_function(self):
    method _get_transform (line 354) | def _get_transform(self):
    method load_image (line 360) | def load_image(self, image_path):
    method multi_modal_get_item (line 366) | def multi_modal_get_item(self, data_item, pack_data=False):
    method multi_modal_multi_image_get_item (line 369) | def multi_modal_multi_image_get_item(self, data_item, pack_data=False):
    method _get_num_frames_by_duration (line 373) | def _get_num_frames_by_duration(self, duration):
    method video_get_item (line 388) | def video_get_item(self, data_item, pack_data=False):
    method pure_text_get_item (line 474) | def pure_text_get_item(self, data_item, pack_data=False):
    method __getitem__ (line 521) | def __getitem__(self, i):
  function packing_collate (line 544) | def packing_collate(features, pack_batch=True, pad_id=0, sp_size=1):
  function build_llava_model (line 614) | def build_llava_model(args, dtype=torch.float32, device='cpu'):
  function build_fsdp_model (line 648) | def build_fsdp_model(rank0_model, meta_model, dp_mesh, tp_mesh, dtype, a...
  function llava_train (line 716) | def llava_train(args):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/__init__.py
  function get_logger (line 11) | def get_logger():

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/__init__.py
  function _dispatch_forward_fn (line 9) | def _dispatch_forward_fn(module, dispatch_fn):
  function dispatch_internlm_varlen_attn_forward (line 13) | def dispatch_internlm_varlen_attn_forward(module):
  function dispatch_clip_attn_forward (line 20) | def dispatch_clip_attn_forward(module):
  function dispatch_rms_norm_forward (line 27) | def dispatch_rms_norm_forward(module):
  function dispatch_phi3_varlen_attn_forward (line 33) | def dispatch_phi3_varlen_attn_forward(module):
  function dispatch_llama3_varlen_attn_forward (line 40) | def dispatch_llama3_varlen_attn_forward(module):
  function dispatch_qwen2vl_varlen_attn_forward (line 47) | def dispatch_qwen2vl_varlen_attn_forward(module):
  function dispatch_internvl2_forward (line 54) | def dispatch_internvl2_forward(module):
  function dispatch_qwen2_varlen_attn_forward (line 61) | def dispatch_qwen2_varlen_attn_forward(module):
  function dispatch_internvl2_hico_forward (line 67) | def dispatch_internvl2_hico_forward(module):
  function dispatch_modules (line 89) | def dispatch_modules(model, exclude_cls=[]):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/_attention.py
  function _get_unpad_data (line 17) | def _get_unpad_data(attention_mask):
  function upad_qkv (line 30) | def upad_qkv(query_layer, key_layer, value_layer, attention_mask,
  function flash_attn_wo_mask (line 75) | def flash_attn_wo_mask(
  function flash_attn_w_mask (line 96) | def flash_attn_w_mask(
  function varlen_flash_attn (line 128) | def varlen_flash_attn(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/_fused/layer_norm.py
  function layer_norm_forward (line 6) | def layer_norm_forward(self, hidden_states):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/_fused/rms_norm.py
  function rms_norm_forward (line 14) | def rms_norm_forward(self, hidden_states):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/_fused/rotary.py
  function rotary_kernel (line 11) | def rotary_kernel(
  function apply_rotary (line 163) | def apply_rotary(
  class ApplyRotaryEmb (line 265) | class ApplyRotaryEmb(torch.autograd.Function):
    method forward (line 268) | def forward(
    method backward (line 302) | def backward(ctx, do):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/clip.py
  function clip_flash_attn_forward (line 11) | def clip_flash_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/internlm2.py
  class InternLM2RotaryEmbedding (line 15) | class InternLM2RotaryEmbedding(torch.nn.Module):
    method __init__ (line 17) | def __init__(self,
    method forward (line 40) | def forward(self, x, seq_len):
  function rotate_half (line 61) | def rotate_half(x):
  function apply_rotary_pos_emb (line 68) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  function apply_rotary_pos_emb_old (line 75) | def apply_rotary_pos_emb_old(q, k, cos, sin, position_ids, unsqueeze_dim...
  function repeat_kv (line 84) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  function repeat_kv_bshd (line 102) | def repeat_kv_bshd(hidden_states: torch.Tensor, n_rep: int) -> torch.Ten...
  function _internlm2_varlen_attn_forward (line 116) | def _internlm2_varlen_attn_forward(
  function internlm2_varlen_attn_forward (line 256) | def internlm2_varlen_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/internvl2.py
  function rescale_sp_loss (line 16) | def rescale_sp_loss(loss_per_sp_rank,
  function internvl2_forward (line 40) | def internvl2_forward(
  function internvl2_hico_forward (line 212) | def internvl2_hico_forward(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/llama3.py
  function rotate_half (line 9) | def rotate_half(x):
  function apply_rotary_pos_emb (line 16) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  function _llama3_varlen_attn_forward (line 43) | def _llama3_varlen_attn_forward(
  function llama3_varlen_attn_forward (line 128) | def llama3_varlen_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/new.py
  function _dispatch_forward_fn (line 9) | def _dispatch_forward_fn(module, dispatch_fn):
  function dispatch_internlm_varlen_attn_forward (line 13) | def dispatch_internlm_varlen_attn_forward(module):
  function dispatch_clip_attn_forward (line 20) | def dispatch_clip_attn_forward(module):
  function dispatch_rms_norm_forward (line 27) | def dispatch_rms_norm_forward(module):
  function dispatch_phi3_varlen_attn_forward (line 33) | def dispatch_phi3_varlen_attn_forward(module):
  function dispatch_llama3_varlen_attn_forward (line 40) | def dispatch_llama3_varlen_attn_forward(module):
  function dispatch_qwen2vl_varlen_attn_forward (line 47) | def dispatch_qwen2vl_varlen_attn_forward(module):
  function dispatch_internvl2_forward (line 54) | def dispatch_internvl2_forward(module):
  function dispatch_qwen2_varlen_attn_forward (line 61) | def dispatch_qwen2_varlen_attn_forward(module):
  function dispatch_modules (line 82) | def dispatch_modules(model, exclude_cls=[]):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/phi3.py
  function rotate_half (line 14) | def rotate_half(x):
  function apply_rotary_pos_emb (line 21) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  function repeat_kv (line 47) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  function _phi3_varlen_attn_forward (line 59) | def _phi3_varlen_attn_forward(
  function ph3_varlen_attn_forward (line 224) | def ph3_varlen_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/qwen2.py
  function rotate_half (line 12) | def rotate_half(x):
  function apply_rotary_pos_emb (line 19) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  function qwen2_varlen_attn_forward (line 46) | def qwen2_varlen_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/dispatches/qwen_vl2.py
  function qwen2_vl_varlen_attn_forward (line 26) | def qwen2_vl_varlen_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/fsdp/checkpointing.py
  function checkpoint_check_fn (line 8) | def checkpoint_check_fn(submodule, target=RECOMPUTE_MODULES, selective=1...

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/fsdp/clip_grad.py
  function clip_grad_norm_ (line 14) | def clip_grad_norm_(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/fsdp/lazy.py
  function dp_lazy_init (line 7) | def dp_lazy_init(module, module_map, dp_mesh):
  function dp_sp_lazy_init (line 35) | def dp_sp_lazy_init(module, module_map, dp_mesh, sp_mesh):
  function dp_tp_lazy_init (line 61) | def dp_tp_lazy_init(module, module_map, dp_mesh, tp_mesh):
  class LoadWoInit (line 139) | class LoadWoInit:
    method __init__ (line 142) | def __init__(self):
    method __enter__ (line 151) | def __enter__(self, *args, **kwargs):
    method __exit__ (line 160) | def __exit__(self, *args, **kwargs):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/fsdp/precision.py
  function set_require_grad_param_to_fp32 (line 5) | def set_require_grad_param_to_fp32(model: nn.Module):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/fsdp/wrap.py
  function layer_auto_wrap_policy (line 14) | def layer_auto_wrap_policy(
  function token_embedding_wrap_policy (line 29) | def token_embedding_wrap_policy(
  function all_required_grad_wrap_policy (line 46) | def all_required_grad_wrap_policy(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/accelerate/packed.py
  function packed_sequence (line 6) | def packed_sequence(num_tokens, enable=False, sp_size=1, skip_position_i...

FILE: xtuner-train_internvideo2_5/xtuner/_lite/auto.py
  function download_model_from_hub (line 20) | def download_model_from_hub(
  class AutoModelForCausalLM (line 64) | class AutoModelForCausalLM:
    method from_config (line 91) | def from_config(cls,
    method from_pretrained (line 102) | def from_pretrained(
    method _flash_attn_kwargs (line 139) | def _flash_attn_kwargs(config: PretrainedConfig) -> dict:
    method _long_ctx_kwargs (line 162) | def _long_ctx_kwargs(config: PretrainedConfig,

FILE: xtuner-train_internvideo2_5/xtuner/_lite/chat/messages/base.py
  class BaseMessages (line 10) | class BaseMessages(BaseModel):
    method add (line 13) | def add(self, role: str, content):
    method pop (line 17) | def pop(self):
    method get_prompt (line 21) | def get_prompt(self, chat_template: ChatTemplate) -> str:
    method tokenize (line 25) | def tokenize(self, tokenizer: PreTrainedTokenizer,
    method from_dict (line 30) | def from_dict(cls, item: Dict) -> 'BaseMessages':

FILE: xtuner-train_internvideo2_5/xtuner/_lite/chat/messages/chat.py
  class TextContentItem (line 15) | class TextContentItem(BaseModel):
    method apply_chat_template (line 19) | def apply_chat_template(self, chat_template: HybridChatTemplate) -> str:
  class ImageContentItem (line 23) | class ImageContentItem(BaseModel):
    method apply_chat_template (line 27) | def apply_chat_template(self, chat_template: HybridChatTemplate) -> str:
  class ChatMsg (line 35) | class ChatMsg(BaseModel):
    method __init__ (line 41) | def __init__(self, *args, **kwargs):
    method collect_img_urls (line 53) | def collect_img_urls(self) -> List[str]:
    method get_prompt (line 61) | def get_prompt(self, chat_template: ChatTemplate) -> str:
    method tokenize (line 86) | def tokenize(
  class ChatMessages (line 107) | class ChatMessages(BaseMessages):
    method add (line 111) | def add(self, role, content, loss=False):
    method pop (line 114) | def pop(self):
    method get_prompt (line 117) | def get_prompt(self, chat_template: ChatTemplate) -> str:
    method tokenize (line 127) | def tokenize(self, tokenizer: PreTrainedTokenizer,
    method from_str (line 169) | def from_str(cls, prompt: str) -> 'ChatMessages':
    method from_dict (line 175) | def from_dict(cls, item: dict) -> 'ChatMessages':

FILE: xtuner-train_internvideo2_5/xtuner/_lite/chat/templates/chat.py
  class ChatTemplate (line 6) | class ChatTemplate(BaseModel):
    method decorate_system (line 18) | def decorate_system(self, text: str) -> str:
    method decorate_assistant (line 22) | def decorate_assistant(self, text: str) -> str:
    method decorate_user (line 26) | def decorate_user(self, text: str) -> str:
    method check_system (line 31) | def check_system(cls, v: str) -> str:
    method check_user (line 41) | def check_user(cls, v: str) -> str:
    method check_assistant (line 51) | def check_assistant(cls, v: str) -> str:

FILE: xtuner-train_internvideo2_5/xtuner/_lite/chat/templates/hybrid.py
  class HybridChatTemplate (line 6) | class HybridChatTemplate(BaseModel):
    method mm_token_maps (line 42) | def mm_token_maps(self) -> Dict[str, int]:
    method decorate_system (line 47) | def decorate_system(self, text: str) -> str:
    method decorate_assistant (line 51) | def decorate_assistant(self, text: str) -> str:
    method decorate_user (line 55) | def decorate_user(self, text: str) -> str:
    method decorate_files (line 59) | def decorate_files(self, text: str) -> str:
    method decorate_functions (line 63) | def decorate_functions(self, text: str) -> str:
    method decorate_function_call (line 67) | def decorate_function_call(self, text: str, func: str) -> str:
    method decorate_function_result (line 71) | def decorate_function_result(self, text: str) -> str:
    method decorate_code_interpreter (line 75) | def decorate_code_interpreter(self, text: str) -> str:
    method decorate_code_interpreter_call (line 79) | def decorate_code_interpreter_call(self, text: str, func: str) -> str:
    method decorate_code_interpreter_result (line 84) | def decorate_code_interpreter_result(self, text: str) -> str:
    method check_system (line 90) | def check_system(cls, v: str) -> str:
    method check_user (line 100) | def check_user(cls, v: str) -> str:
    method check_assistant (line 110) | def check_assistant(cls, v: str) -> str:
    method check_function_call (line 121) | def check_function_call(cls, v: str) -> str:
    method check_function_result (line 137) | def check_function_result(cls, v: str) -> str:
    method check_functions (line 148) | def check_functions(cls, v: str) -> str:
    method check_code_interpreter (line 159) | def check_code_interpreter(cls, v: str) -> str:
    method check_code_interpreter_call (line 170) | def check_code_interpreter_call(cls, v: str) -> str:
    method check_code_interpreter_result (line 186) | def check_code_interpreter_result(cls, v: str) -> str:

FILE: xtuner-train_internvideo2_5/xtuner/_lite/checkpoint.py
  function _no_hook (line 13) | def _no_hook(module: nn.Module):
  function checkpoint (line 28) | def checkpoint(module: nn.Module) -> nn.Module:
  class MyModel (line 66) | class MyModel(nn.Module):
    method __init__ (line 68) | def __init__(self):
    method forward (line 73) | def forward(self, x, b, a=4, c=4):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/datasets/dataset_fn.py
  function _apply_exif_orientation (line 30) | def _apply_exif_orientation(image):
  function log_format (line 77) | def log_format(rank, debug=False):
  function check_args (line 90) | def check_args(args):
  function set_logger_envs (line 124) | def set_logger_envs(args):
  class SoftPackDataset (line 166) | class SoftPackDataset(Dataset):
    method __init__ (line 168) | def __init__(self, datasets, pack_max_length=32768, concat_before_pack...
    method get_pack_infos (line 195) | def get_pack_infos(self, dataset, dataset_id, num_tokens):
    method __len__ (line 238) | def __len__(self):
    method __getitem__ (line 241) | def __getitem__(self, item):
  function load_json_or_jsonl (line 247) | def load_json_or_jsonl(json_path):
  function calculate_jsonl_sha256 (line 260) | def calculate_jsonl_sha256(path):
  function calculate_json_sha256 (line 268) | def calculate_json_sha256(file_path):
  function expand2square (line 277) | def expand2square(pil_img, background_color):
  class BaseOrigDataset (line 291) | class BaseOrigDataset(Dataset):
    method __init__ (line 292) | def __init__(self,
    method __len__ (line 363) | def __len__(self):
    method calc_group_len (line 366) | def calc_group_len(self):
    method calc_packing_info (line 369) | def calc_packing_info(self):
    method count_tokens_for_pack (line 392) | def count_tokens_for_pack(self, cache_dir=None):
    method pre_tokenize_fn_for_pack (line 432) | def pre_tokenize_fn_for_pack(self, data):
    method process_text (line 435) | def process_text(self, conversations, media_type='image', image_grids=...
    method _process_media_format_first_round (line 488) | def _process_media_format_first_round(self, input_, media_type, image_...
    method modality_length (line 492) | def modality_length(self):
    method length (line 496) | def length(self):
  function build_dataset (line 502) | def build_dataset(args, datasets):
  function build_train_dataloader (line 515) | def build_train_dataloader(args, train_dataset, collate_fn):
  function _prepare_input (line 554) | def _prepare_input(data, device='cuda'):
  function is_interval (line 567) | def is_interval(step, total_steps, interval):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/datasets/format.py
  class Alpaca2Openai (line 4) | class Alpaca2Openai():
    method source_format (line 7) | def source_format(cls):
    method target_format (line 16) | def target_format(cls):
    method convert (line 32) | def convert(data):
  function llava_to_openai (line 50) | def llava_to_openai(data):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/datasets/llava.py
  function expand2square (line 14) | def expand2square(pil_img, background_color):
  class LlavaTokenizeFunction (line 28) | class LlavaTokenizeFunction():
    method __init__ (line 30) | def __init__(self,
    method __call__ (line 45) | def __call__(self, item):
  class LlavaTokenizedDataset (line 80) | class LlavaTokenizedDataset(torch.utils.data.Dataset):
    method __init__ (line 82) | def __init__(self, dataset, image_processor, pad_image_to_square=False):
    method process_tokenized_data (line 88) | def process_tokenized_data(self, tokenized_data):
    method __getitem__ (line 117) | def __getitem__(self, item):
  class LlavaRawDataset (line 131) | class LlavaRawDataset(LlavaTokenizedDataset):
    method __init__ (line 133) | def __init__(self, dataset, image_processor, tokenize_fn, pad_image_to...
    method modality_length (line 165) | def modality_length(self):
    method length (line 169) | def length(self):
    method __getitem__ (line 174) | def __getitem__(self, item):
    method __len__ (line 179) | def __len__(self):
  class SoftPackerForLlava (line 183) | class SoftPackerForLlava(SoftPackerForText):
    method __init__ (line 185) | def __init__(self,
    method __getitem__ (line 196) | def __getitem__(self, item):
  class LlavaCollator (line 273) | class LlavaCollator():
    method __init__ (line 275) | def __init__(self, pack_batch=False):
    method __call__ (line 278) | def __call__(self, instances):
  class LlavaDatasetForNonPack (line 343) | class LlavaDatasetForNonPack(torch.utils.data.Dataset):
    method __init__ (line 345) | def __init__(self,
    method modality_length (line 385) | def modality_length(self):
    method length (line 389) | def length(self):
    method __getitem__ (line 394) | def __getitem__(self, item):
    method __len__ (line 398) | def __len__(self):
  class NewLlavaTokenizeFunction (line 402) | class NewLlavaTokenizeFunction():
    method __init__ (line 404) | def __init__(self,
    method __call__ (line 423) | def __call__(self, item):
  class NewLlavaCollator (line 468) | class NewLlavaCollator():
    method __init__ (line 470) | def __init__(self, pad_token_id=0, ignore_id=-100, pack_batch=False):
    method __call__ (line 475) | def __call__(self, instances):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/datasets/load.py
  function load_json (line 20) | def load_json(file):
  function load_jsonl (line 26) | def load_jsonl(file):
  function load_bin (line 34) | def load_bin(file):
  function master_only_load (line 41) | def master_only_load(load_fn):
  function multi_thread_map (line 70) | def multi_thread_map(map_fns, dataset, desc, num_proc=8):
  function openai_format (line 90) | def openai_format(item):
  function load_hf_dataset (line 97) | def load_hf_dataset(path,
  function load_from_cache (line 128) | def load_from_cache(cache_dir, keep_cache_path=None):
  function load_local_datasets (line 141) | def load_local_datasets(paths,
  function load_datasets (line 376) | def load_datasets(paths,
  function load_ms_dataset (line 496) | def load_ms_dataset():

FILE: xtuner-train_internvideo2_5/xtuner/_lite/datasets/load_new.py
  function calculate_jsonl_sha256 (line 24) | def calculate_jsonl_sha256(path):
  function calculate_tokenize_fn_sha256 (line 32) | def calculate_tokenize_fn_sha256(tokenize_fn):
  class JsonlDataset (line 39) | class JsonlDataset(torch.utils.data.Dataset):
    method __init__ (line 41) | def __init__(self,
    method count_offsets (line 101) | def count_offsets(self, cache_dir=None):
    method count_tokens (line 119) | def count_tokens(self, cache_dir=None):
    method __len__ (line 165) | def __len__(self):
    method __getitem__ (line 168) | def __getitem__(self, item):
  function calculate_json_sha256 (line 190) | def calculate_json_sha256(file_path):
  class JsonDataset (line 199) | class JsonDataset(torch.utils.data.Dataset):
    method __init__ (line 201) | def __init__(self,
    method count_tokens (line 256) | def count_tokens(self, cache_dir=None):
    method __len__ (line 302) | def __len__(self):
    method __getitem__ (line 305) | def __getitem__(self, item):
  function load_local_datasets (line 330) | def load_local_datasets(paths,
  function load_datasets (line 414) | def load_datasets(paths,
  function load_ms_dataset (line 479) | def load_ms_dataset():
  class SoftPackDataset (line 483) | class SoftPackDataset(torch.utils.data.Dataset):
    method __init__ (line 485) | def __init__(self, datasets, target=2048, blend=False, sort=False):
    method get_pack_infos (line 503) | def get_pack_infos(self, dataset, dataset_id, num_tokens):
    method __len__ (line 545) | def __len__(self):
    method __getitem__ (line 548) | def __getitem__(self, item):
  class SftCollator (line 556) | class SftCollator():
    method __init__ (line 558) | def __init__(self, pad_token_id=0, ignore_id=-100, pack_batch=False):
    method __call__ (line 563) | def __call__(self, instances):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/datasets/text.py
  function sort_and_return_indices (line 23) | def sort_and_return_indices(lst):
  class TextTokenizeFunction (line 27) | class TextTokenizeFunction():
    method __init__ (line 29) | def __init__(self, tokenizer, chat_template, raw_format='openai'):
    method __call__ (line 35) | def __call__(self, item):
  class TextTokenizedDataset (line 43) | class TextTokenizedDataset(torch.utils.data.Dataset):
    method __init__ (line 45) | def __init__(self, dataset):
    method __len__ (line 50) | def __len__(self):
    method __getitem__ (line 53) | def __getitem__(self, item):
  class TextOnlineTokenizeDataset (line 72) | class TextOnlineTokenizeDataset(torch.utils.data.Dataset):
    method __init__ (line 74) | def __init__(self, dataset, tokenize_fn):
    method __len__ (line 80) | def __len__(self):
    method __getitem__ (line 83) | def __getitem__(self, item):
  class SoftPackerForText (line 104) | class SoftPackerForText(torch.utils.data.Dataset):
    method __init__ (line 106) | def __init__(self, dataset, max_length=2048, pack_info=None):
    method cache (line 123) | def cache(self):
    method load_cache (line 141) | def load_cache(self):
    method __len__ (line 149) | def __len__(self):
    method __getitem__ (line 152) | def __getitem__(self, item):
    method get_pack_info (line 205) | def get_pack_info(cls, dataset, max_length):
    method get_pack_infos (line 254) | def get_pack_infos(cls, datasets, max_length):
  class HardPackerForText (line 287) | class HardPackerForText(torch.utils.data.Dataset):
    method __init__ (line 306) | def __init__(self, dataset, max_length=2048, pack_info=None):
    method _cal_max_length (line 323) | def _cal_max_length(begin, end, shfl_item_rngs_left, shfl_item_rngs_ri...
    method get_pack_info (line 337) | def get_pack_info(cls, dataset, max_length):
    method get_pack_infos (line 380) | def get_pack_infos(cls, datasets, max_length):
    method _pack_ids_and_labels_in_range (line 412) | def _pack_ids_and_labels_in_range(self, begin: int, end: int):
    method __len__ (line 457) | def __len__(self):
    method __getitem__ (line 460) | def __getitem__(self, item):
  class TextCollator (line 488) | class TextCollator():
    method __init__ (line 490) | def __init__(self, pack_batch=False):
    method __call__ (line 493) | def __call__(self, instances):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/internvl/conversation.py
  class SeparatorStyle (line 13) | class SeparatorStyle(IntEnum):
  class Conversation (line 37) | class Conversation:
    method get_prompt (line 61) | def get_prompt(self) -> str:
    method set_system_message (line 251) | def set_system_message(self, system_message: str):
    method append_message (line 255) | def append_message(self, role: str, message: str):
    method update_last_message (line 259) | def update_last_message(self, message: str):
    method to_gradio_chatbot (line 267) | def to_gradio_chatbot(self):
    method to_openai_api_messages (line 277) | def to_openai_api_messages(self):
    method copy (line 289) | def copy(self):
    method dict (line 304) | def dict(self):
  function register_conv_template (line 318) | def register_conv_template(template: Conversation, override: bool = False):
  function get_conv_template (line 328) | def get_conv_template(name: str) -> Conversation:

FILE: xtuner-train_internvideo2_5/xtuner/_lite/internvl/dataset.py
  function get_frame_indices (line 37) | def get_frame_indices(num_frames, vlen, sample='rand', fix_start=None, i...
  function read_frames_gif (line 78) | def read_frames_gif(
  function read_frames_decord (line 103) | def read_frames_decord(
  function extract_frame_number (line 136) | def extract_frame_number(filename):
  function sort_frames (line 142) | def sort_frames(frame_paths):
  function read_frames_folder (line 147) | def read_frames_folder(
  class WeightedConcatDataset (line 177) | class WeightedConcatDataset(ConcatDataset):
    method __init__ (line 178) | def __init__(self, datasets, weights):
    method __iter__ (line 184) | def __iter__(self):
    method __len__ (line 187) | def __len__(self):
  function pil_loader (line 191) | def pil_loader(img_str):
  class TCSLoader (line 197) | class TCSLoader(object):
    method __init__ (line 199) | def __init__(self, sc_config_key='sensecore'):
    method __call__ (line 204) | def __call__(self, fn, image_type='image', max_num_frames=-1, min_num_...
  function expand2square (line 229) | def expand2square(pil_img, background_color):
  function simulate_jpeg_degradation (line 243) | def simulate_jpeg_degradation(quality):
  function build_transform (line 258) | def build_transform(is_train, input_size, pad2square=False, normalize_ty...
  function preprocess (line 295) | def preprocess(
  function preprocess_mpt (line 404) | def preprocess_mpt(
  function preprocess_phi3 (line 502) | def preprocess_phi3(
  function preprocess_phi3_fast (line 614) | def preprocess_phi3_fast(
  function preprocess_internlm (line 706) | def preprocess_internlm(
  function find_closest_aspect_ratio (line 804) | def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height...
  function dynamic_preprocess (line 821) | def dynamic_preprocess(image, min_num=1, max_num=6, image_size=448, use_...
  function dynamic_num_patch (line 860) | def dynamic_num_patch(size, min_num=1, max_num=6, image_size=448, use_th...
  function concat_pad_data_collator (line 885) | def concat_pad_data_collator(features, pad_id=0):
  function packing_collate (line 954) | def packing_collate(features, pack_batch=True, pad_id=0):
  function concat_pad_data_collator_dpo (line 1005) | def concat_pad_data_collator_dpo(features, pad_id=0):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/internvl/new_dataset.py
  function get_frame_indices (line 37) | def get_frame_indices(num_frames, vlen, sample='rand', fix_start=None, i...
  function read_frames_gif (line 78) | def read_frames_gif(
  function read_frames_decord (line 103) | def read_frames_decord(
  function extract_frame_number (line 136) | def extract_frame_number(filename):
  function sort_frames (line 142) | def sort_frames(frame_paths):
  function read_frames_folder (line 147) | def read_frames_folder(
  class WeightedConcatDataset (line 177) | class WeightedConcatDataset(ConcatDataset):
    method __init__ (line 178) | def __init__(self, datasets, weights):
    method __iter__ (line 184) | def __iter__(self):
    method __len__ (line 187) | def __len__(self):
  function pil_loader (line 191) | def pil_loader(img_str):
  class TCSLoader (line 197) | class TCSLoader(object):
    method __init__ (line 199) | def __init__(self, conf_path, sc_config_key='sensecore'):
    method __call__ (line 207) | def __call__(self, fn, image_type='image', max_num_frames=-1, min_num_...
  function expand2square (line 245) | def expand2square(pil_img, background_color):
  function simulate_jpeg_degradation (line 259) | def simulate_jpeg_degradation(quality):
  function build_transform (line 274) | def build_transform(is_train, input_size, pad2square=False, normalize_ty...
  function preprocess (line 311) | def preprocess(
  function preprocess_mpt (line 419) | def preprocess_mpt(
  function preprocess_phi3 (line 517) | def preprocess_phi3(
  function preprocess_phi3_fast (line 630) | def preprocess_phi3_fast(
  function preprocess_internlm (line 720) | def preprocess_internlm(
  function find_closest_aspect_ratio (line 819) | def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height...
  function dynamic_preprocess (line 836) | def dynamic_preprocess(image, min_num=1, max_num=6, image_size=448, use_...
  function dynamic_num_patch (line 875) | def dynamic_num_patch(size, min_num=1, max_num=6, image_size=448, use_th...
  function concat_pad_data_collator (line 900) | def concat_pad_data_collator(features, pad_id=0):
  function packing_collate (line 969) | def packing_collate(features, pack_batch=True, pad_id=0):
  function concat_pad_data_collator_dpo (line 1020) | def concat_pad_data_collator_dpo(features, pad_id=0):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/internvl/v1_5/configuration_intern_vit.py
  class InternVisionConfig (line 15) | class InternVisionConfig(PretrainedConfig):
    method __init__ (line 63) | def __init__(
    method from_pretrained (line 107) | def from_pretrained(cls, pretrained_model_name_or_path: Union[str, os....

FILE: xtuner-train_internvideo2_5/xtuner/_lite/internvl/v1_5/configuration_internvl_chat.py
  class InternVLChatConfig (line 19) | class InternVLChatConfig(PretrainedConfig):
    method __init__ (line 23) | def __init__(
    method to_dict (line 73) | def to_dict(self):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/internvl/v1_5/configuration_phi3.py
  class Phi3Config (line 31) | class Phi3Config(PretrainedConfig):
    method __init__ (line 117) | def __init__(
    method _rope_scaling_validation (line 175) | def _rope_scaling_validation(self):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/internvl/v1_5/conversation.py
  class SeparatorStyle (line 13) | class SeparatorStyle(IntEnum):
  class Conversation (line 37) | class Conversation:
    method get_prompt (line 61) | def get_prompt(self) -> str:
    method set_system_message (line 251) | def set_system_message(self, system_message: str):
    method append_message (line 255) | def append_message(self, role: str, message: str):
    method update_last_message (line 259) | def update_last_message(self, message: str):
    method to_gradio_chatbot (line 267) | def to_gradio_chatbot(self):
    method to_openai_api_messages (line 277) | def to_openai_api_messages(self):
    method copy (line 289) | def copy(self):
    method dict (line 304) | def dict(self):
  function register_conv_template (line 318) | def register_conv_template(template: Conversation, override: bool = False):
  function get_conv_template (line 328) | def get_conv_template(name: str) -> Conversation:

FILE: xtuner-train_internvideo2_5/xtuner/_lite/internvl/v1_5/modeling_intern_vit.py
  class FlashAttention (line 33) | class FlashAttention(nn.Module):
    method __init__ (line 44) | def __init__(self, softmax_scale=None, attention_dropout=0.0, device=N...
    method forward (line 49) | def forward(self, qkv, key_padding_mask=None, causal=False, cu_seqlens...
  class InternRMSNorm (line 97) | class InternRMSNorm(nn.Module):
    method __init__ (line 98) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 103) | def forward(self, hidden_states):
  class InternVisionEmbeddings (line 131) | class InternVisionEmbeddings(nn.Module):
    method __init__ (line 132) | def __init__(self, config):
    method _get_pos_embed (line 152) | def _get_pos_embed(self, pos_embed, H, W):
    method forward (line 160) | def forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
  class InternAttention (line 175) | class InternAttention(nn.Module):
    method __init__ (line 178) | def __init__(self, config):
    method _naive_attn (line 208) | def _naive_attn(self, x):
    method _flash_attn (line 227) | def _flash_attn(self, x, key_padding_mask=None, need_weights=False):
    method forward (line 244) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
  class InternMLP (line 249) | class InternMLP(nn.Module):
    method __init__ (line 250) | def __init__(self, config):
    method forward (line 257) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
  class InternVisionEncoderLayer (line 264) | class InternVisionEncoderLayer(nn.Module):
    method __init__ (line 265) | def __init__(self, config, drop_path_rate: float):
    method forward (line 281) | def forward(
  class InternVisionEncoder (line 296) | class InternVisionEncoder(nn.Module):
    method __init__ (line 306) | def __init__(self, config):
    method forward (line 318) | def forward(
  class InternVisionModel (line 365) | class InternVisionModel(PreTrainedModel):
    method __init__ (line 370) | def __init__(self, config):
    method resize_pos_embeddings (line 377) | def resize_pos_embeddings(self, old_size, new_size, patch_size):
    method get_input_embeddings (line 389) | def get_input_embeddings(self):
    method forward (line 392) | def forward(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/internvl/v1_5/modeling_internvl_chat.py
  function version_cmp (line 39) | def version_cmp(v1, v2, op='eq'):
  function rescale_sp_loss (line 47) | def rescale_sp_loss(loss_per_sp_rank,
  class InternVLChatModel (line 71) | class InternVLChatModel(PreTrainedModel):
    method __init__ (line 78) | def __init__(self, config: InternVLChatConfig, vision_model=None, lang...
    method forward (line 125) | def forward(
    method pixel_shuffle (line 363) | def pixel_shuffle(self, x, scale_factor=0.5):
    method extract_feature (line 379) | def extract_feature(self, pixel_values):
    method batch_chat (line 399) | def batch_chat(self, tokenizer, pixel_values, questions, generation_co...
    method chat (line 448) | def chat(self, tokenizer, pixel_values, question, generation_config, h...
    method generate (line 505) | def generate(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/internvl/v1_5/modeling_phi3.py
  class Phi3RMSNorm (line 78) | class Phi3RMSNorm(nn.Module):
    method __init__ (line 79) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 87) | def forward(self, hidden_states):
  function _get_unpad_data (line 96) | def _get_unpad_data(attention_mask):
  class Phi3RotaryEmbedding (line 109) | class Phi3RotaryEmbedding(nn.Module):
    method __init__ (line 110) | def __init__(self, dim, max_position_embeddings=2048, base=10000, devi...
    method forward (line 119) | def forward(self, x, position_ids, seq_len=None):
  class Phi3SuScaledRotaryEmbedding (line 139) | class Phi3SuScaledRotaryEmbedding(Phi3RotaryEmbedding):
    method __init__ (line 140) | def __init__(self, dim, config, device=None):
    method forward (line 148) | def forward(self, x, position_ids, seq_len=None):
  class Phi3YarnScaledRotaryEmbedding (line 180) | class Phi3YarnScaledRotaryEmbedding(Phi3RotaryEmbedding):
    method __init__ (line 181) | def __init__(self, dim, config, device=None):
    method forward (line 189) | def forward(self, x, position_ids, seq_len=None):
  function rotate_half (line 222) | def rotate_half(x):
  function apply_rotary_pos_emb (line 230) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  class Phi3MLP (line 257) | class Phi3MLP(nn.Module):
    method __init__ (line 258) | def __init__(self, config):
    method forward (line 267) | def forward(self, hidden_states: torch.FloatTensor) -> torch.FloatTensor:
  function repeat_kv (line 277) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class Phi3Attention (line 289) | class Phi3Attention(nn.Module):
    method __init__ (line 292) | def __init__(self, config: Phi3Config, layer_idx: Optional[int] = None):
    method _init_rope (line 326) | def _init_rope(self):
    method forward (line 342) | def forward(
  class Phi3FlashAttention2 (line 424) | class Phi3FlashAttention2(Phi3Attention):
    method __init__ (line 432) | def __init__(self, *args, **kwargs):
    method forward (line 440) | def forward(
    method _flash_attention_forward (line 589) | def _flash_attention_forward(
    method _upad_input (line 690) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class Phi3SdpaAttention (line 735) | class Phi3SdpaAttention(Phi3Attention):
    method forward (line 743) | def forward(
  class Phi3DecoderLayer (line 831) | class Phi3DecoderLayer(nn.Module):
    method __init__ (line 832) | def __init__(self, config: Phi3Config, layer_idx: int):
    method forward (line 845) | def forward(
  class Phi3PreTrainedModel (line 930) | class Phi3PreTrainedModel(PreTrainedModel):
    method __init__ (line 942) | def __init__(self, config: Phi3Config):
    method _init_weights (line 948) | def _init_weights(self, module):
  class Phi3Model (line 1034) | class Phi3Model(Phi3PreTrainedModel):
    method __init__ (line 1042) | def __init__(self, config: Phi3Config):
    method get_input_embeddings (line 1060) | def get_input_embeddings(self):
    method set_input_embeddings (line 1063) | def set_input_embeddings(self, value):
    method forward (line 1067) | def forward(
  class Phi3ForCausalLM (line 1204) | class Phi3ForCausalLM(Phi3PreTrainedModel):
    method __init__ (line 1208) | def __init__(self, config):
    method get_input_embeddings (line 1218) | def get_input_embeddings(self):
    method set_input_embeddings (line 1222) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1226) | def get_output_embeddings(self):
    method set_output_embeddings (line 1230) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1234) | def set_decoder(self, decoder):
    method get_decoder (line 1238) | def get_decoder(self):
    method forward (line 1244) | def forward(
    method prepare_inputs_for_generation (line 1359) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1418) | def _reorder_cache(past_key_values, beam_idx):
  class Phi3ForSequenceClassification (line 1443) | class Phi3ForSequenceClassification(Phi3PreTrainedModel):
    method __init__ (line 1444) | def __init__(self, config):
    method get_input_embeddings (line 1453) | def get_input_embeddings(self):
    method set_input_embeddings (line 1456) | def set_input_embeddings(self, value):
    method forward (line 1460) | def forward(
  class Phi3ForTokenClassification (line 1559) | class Phi3ForTokenClassification(Phi3PreTrainedModel):
    method __init__ (line 1560) | def __init__(self, config: Phi3Config):
    method forward (line 1583) | def forward(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/internvl/video_utils.py
  function get_index (line 20) | def get_index(num_frames, num_segments):
  function lazy_load_s3video (line 29) | def lazy_load_s3video(s3path_video, num_frames, video_start, video_end, ...
  function pts_to_secs (line 73) | def pts_to_secs(pts: int, time_base: float, start_pts: int) -> float:
  function get_pyav_video_duration (line 88) | def get_pyav_video_duration(video_reader):
  function get_frame_indices (line 99) | def get_frame_indices(num_frames, vlen, sample='middle', fix_start=None,...
  function read_frames_av (line 169) | def read_frames_av(video_path, num_frames, sample='rand', client=None, f...
  function read_frames_gif (line 198) | def read_frames_gif(
  function read_frames_decord (line 250) | def read_frames_decord(
  function read_frames_img (line 300) | def read_frames_img(
  function read_frames_fake (line 402) | def read_frames_fake(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/modelings/__init__.py
  function register_remote_code (line 31) | def register_remote_code():
  function lce_forward (line 38) | def lce_forward(
  function apply_liger_kernel_to_qwen2_vl (line 188) | def apply_liger_kernel_to_qwen2_vl(
  class LigerSwiGLUMLPForInternlm2 (line 232) | class LigerSwiGLUMLPForInternlm2(nn.Module):
    method __init__ (line 233) | def __init__(self, config):
    method forward (line 244) | def forward(self, x):
  function apply_liger_kernel_to_llava_clip_internlm2 (line 250) | def apply_liger_kernel_to_llava_clip_internlm2(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/modelings/internlm2/configuration_internlm2.py
  class InternLM2Config (line 28) | class InternLM2Config(PretrainedConfig):
    method __init__ (line 95) | def __init__(  # pylint: disable=W0102
    method _rope_scaling_validation (line 151) | def _rope_scaling_validation(self):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/modelings/internlm2/modeling_internlm2.py
  function _get_unpad_data (line 69) | def _get_unpad_data(attention_mask):
  class InternLM2RMSNorm (line 81) | class InternLM2RMSNorm(nn.Module):
    method __init__ (line 84) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 89) | def forward(self, hidden_states):
  class InternLM2RotaryEmbedding (line 100) | class InternLM2RotaryEmbedding(nn.Module):
    method __init__ (line 103) | def __init__(self, dim, max_position_embeddings=2048, base=10000, devi...
    method forward (line 115) | def forward(self, x, position_ids):
  class InternLM2LinearScalingRotaryEmbedding (line 131) | class InternLM2LinearScalingRotaryEmbedding(InternLM2RotaryEmbedding):
    method forward (line 134) | def forward(self, x, position_ids):
  class InternLM2DynamicNTKScalingRotaryEmbedding (line 141) | class InternLM2DynamicNTKScalingRotaryEmbedding(InternLM2RotaryEmbedding):
    method forward (line 145) | def forward(self, x, position_ids):
  function rotate_half (line 159) | def rotate_half(x):
  function apply_rotary_pos_emb (line 166) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  class InternLM2MLP (line 192) | class InternLM2MLP(nn.Module):
    method __init__ (line 195) | def __init__(self, config):
    method forward (line 205) | def forward(self, x):
  function repeat_kv (line 211) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class InternLM2Attention (line 223) | class InternLM2Attention(nn.Module):
    method __init__ (line 226) | def __init__(self, config: InternLM2Config, layer_idx: Optional[int] =...
    method _init_rope (line 261) | def _init_rope(self):
    method forward (line 288) | def forward(
  class InternLM2FlashAttention2 (line 371) | class InternLM2FlashAttention2(InternLM2Attention):
    method __init__ (line 378) | def __init__(self, *args, **kwargs):
    method forward (line 389) | def forward(
    method _flash_attention_forward (line 484) | def _flash_attention_forward(
    method _upad_input (line 543) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class InternLM2SdpaAttention (line 585) | class InternLM2SdpaAttention(InternLM2Attention):
    method forward (line 593) | def forward(
  class InternLM2DecoderLayer (line 694) | class InternLM2DecoderLayer(nn.Module):
    method __init__ (line 697) | def __init__(self, config: InternLM2Config, layer_idx: int):
    method forward (line 708) | def forward(
  class InternLM2PreTrainedModel (line 785) | class InternLM2PreTrainedModel(PreTrainedModel):
    method _init_weights (line 801) | def _init_weights(self, module):
  class InternLM2Model (line 880) | class InternLM2Model(InternLM2PreTrainedModel):
    method __init__ (line 889) | def __init__(self, config: InternLM2Config):
    method get_input_embeddings (line 906) | def get_input_embeddings(self):
    method set_input_embeddings (line 909) | def set_input_embeddings(self, value):
    method forward (line 913) | def forward(
    method _update_causal_mask (line 1025) | def _update_causal_mask(
  class InternLM2ForCausalLM (line 1106) | class InternLM2ForCausalLM(InternLM2PreTrainedModel):
    method __init__ (line 1112) | def __init__(self, config):
    method get_input_embeddings (line 1121) | def get_input_embeddings(self):
    method set_input_embeddings (line 1124) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1127) | def get_output_embeddings(self):
    method set_output_embeddings (line 1130) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1133) | def set_decoder(self, decoder):
    method get_decoder (line 1136) | def get_decoder(self):
    method forward (line 1141) | def forward(
    method prepare_inputs_for_generation (line 1237) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1317) | def _reorder_cache(past_key_values, beam_idx):
    method build_inputs (line 1325) | def build_inputs(self, tokenizer, query: str, history: List[Tuple[str,...
    method chat (line 1340) | def chat(
    method stream_chat (line 1380) | def stream_chat(
  class InternLM2ForSequenceClassification (line 1486) | class InternLM2ForSequenceClassification(InternLM2PreTrainedModel):
    method __init__ (line 1489) | def __init__(self, config):
    method get_input_embeddings (line 1498) | def get_input_embeddings(self):
    method set_input_embeddings (line 1501) | def set_input_embeddings(self, value):
    method forward (line 1505) | def forward(
  class InternLM2ForQuestionAnswering (line 1604) | class InternLM2ForQuestionAnswering(InternLM2PreTrainedModel):
    method __init__ (line 1609) | def __init__(self, config):
    method get_input_embeddings (line 1617) | def get_input_embeddings(self):
    method set_input_embeddings (line 1620) | def set_input_embeddings(self, value):
    method forward (line 1624) | def forward(
  class InternLM2ForTokenClassification (line 1705) | class InternLM2ForTokenClassification(InternLM2PreTrainedModel):
    method __init__ (line 1708) | def __init__(self, config):
    method get_input_embeddings (line 1724) | def get_input_embeddings(self):
    method set_input_embeddings (line 1727) | def set_input_embeddings(self, value):
    method forward (line 1731) | def forward(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/modelings/model_fn.py
  function map_meta_modules (line 18) | def map_meta_modules(model, meta_model):
  class MetaStateful (line 27) | class MetaStateful(Stateful):
    method __init__ (line 28) | def __init__(self, **kwargs):
    method state_dict (line 32) | def state_dict(self):
    method load_state_dict (line 35) | def load_state_dict(self, state_dict) -> None:
    method __getitem__ (line 38) | def __getitem__(self, key):
  function lazy_init_megatron (line 43) | def lazy_init_megatron(module, rank0_map, dp_mesh, tp_mesh=None, pp_mesh...
  function resume (line 98) | def resume(args, fsdp_model, optimizer, warmup_scheduler, cosine_schedul...
  function save_ckpt (line 133) | def save_ckpt(args, step, total_steps, fsdp_model, rank0_model, warmup_s...

FILE: xtuner-train_internvideo2_5/xtuner/_lite/parallel/comm.py
  function _all_to_all (line 13) | def _all_to_all(
  class _AllToAll (line 29) | class _AllToAll(torch.autograd.Function):
    method forward (line 40) | def forward(ctx: Any, input: Tensor, sp_group: dist.ProcessGroup,
    method backward (line 51) | def backward(ctx: Any, grad_output: Tensor) -> Tuple:
  function all_to_all (line 67) | def all_to_all(
  function all_to_all_list (line 94) | def all_to_all_list(object_list, group=None):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/parallel/new_setup.py
  function get_device (line 25) | def get_device():
  function setup_parallel (line 45) | def setup_parallel(sp_size=1, tp_size=1, sp_ring_degree=1):
  function get_sp_world_size (line 127) | def get_sp_world_size():
  function get_ulysess_mesh (line 138) | def get_ulysess_mesh():
  function get_ring_mesh (line 142) | def get_ring_mesh():
  function get_ulysess_group (line 146) | def get_ulysess_group():
  function get_ring_group (line 150) | def get_ring_group():
  function get_ulysess_world_size (line 154) | def get_ulysess_world_size():
  function get_ring_world_size (line 165) | def get_ring_world_size():
  function get_world_mesh (line 176) | def get_world_mesh():
  function get_dp_mesh (line 180) | def get_dp_mesh():
  function get_fsdp_mesh (line 184) | def get_fsdp_mesh():
  function get_sp_mesh (line 188) | def get_sp_mesh():
  function get_tp_mesh (line 192) | def get_tp_mesh():
  function get_sp_group (line 196) | def get_sp_group():
  function get_torch_device_module (line 200) | def get_torch_device_module():
  function profile_time_and_memory (line 212) | def profile_time_and_memory(desc):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/parallel/sampler.py
  class ParallelSampler (line 13) | class ParallelSampler(Sampler):
    method __init__ (line 42) | def __init__(
    method __iter__ (line 78) | def __iter__(self) -> Iterator[int]:
    method __len__ (line 99) | def __len__(self) -> int:
    method set_epoch (line 103) | def set_epoch(self, epoch: int, step=0) -> None:
  function get_length_grouped_indices11 (line 117) | def get_length_grouped_indices11(max_lengths,
  function get_length_grouped_indices (line 145) | def get_length_grouped_indices(max_lengths, group_batch_size, generator=...
  class LengthGroupedSampler (line 207) | class LengthGroupedSampler(Sampler):
    method __init__ (line 209) | def __init__(self,
    method __iter__ (line 262) | def __iter__(self) -> Iterator[int]:
    method __len__ (line 283) | def __len__(self) -> int:
    method set_epoch (line 287) | def set_epoch(self, epoch: int, step=0) -> None:

FILE: xtuner-train_internvideo2_5/xtuner/_lite/parallel/sequence/attention.py
  function pre_process_for_sequence_parallel_attn (line 8) | def pre_process_for_sequence_parallel_attn(query_states,
  function post_process_for_sequence_parallel_attn (line 41) | def post_process_for_sequence_parallel_attn(attn_output,
  function sequence_parallel_wrapper (line 54) | def sequence_parallel_wrapper(local_attn):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/parallel/sequence/data_collate.py
  function pad_for_sequence_parallel (line 7) | def pad_for_sequence_parallel(tensor, padding_value, dim=-1):
  function pad_cumulative_len_for_sequence_parallel (line 26) | def pad_cumulative_len_for_sequence_parallel(cumulative_len):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/parallel/sequence/ops.py
  function split_for_sequence_parallel (line 6) | def split_for_sequence_parallel(input, dim: int, sp_group: dist.ProcessG...
  function gather_for_sequence_parallel (line 33) | def gather_for_sequence_parallel(input, dim: int, sp_group: dist.Process...
  class _GatherForwardSplitBackward (line 60) | class _GatherForwardSplitBackward(torch.autograd.Function):
    method forward (line 68) | def forward(ctx, input, dim, sp_group, grad_scale):
    method backward (line 75) | def backward(ctx, grad_output):
  class _SplitForwardGatherBackward (line 85) | class _SplitForwardGatherBackward(torch.autograd.Function):
    method forward (line 93) | def forward(ctx, input, dim, sp_group, grad_scale):
    method backward (line 100) | def backward(ctx, grad_output):
  function split_forward_gather_backward (line 109) | def split_forward_gather_backward(input, dim, sp_group, grad_scale=None):
  function gather_forward_split_backward (line 123) | def gather_forward_split_backward(input, dim, sp_group, grad_scale=None):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/parallel/sequence/reduce_loss.py
  class _ReduceLoss (line 7) | class _ReduceLoss(torch.autograd.Function):
    method forward (line 10) | def forward(ctx, mean_loss, loss_scale, process_group):
    method backward (line 22) | def backward(ctx, grad_output):
  function reduce_sequence_parallel_loss (line 26) | def reduce_sequence_parallel_loss(mean_loss,

FILE: xtuner-train_internvideo2_5/xtuner/_lite/parallel/setup.py
  function setup_sp (line 15) | def setup_sp(sp_size):
  function setup_tp (line 31) | def setup_tp(tp_size):
  function setup_dp (line 47) | def setup_dp():
  function set_seq_parallel_pg (line 67) | def set_seq_parallel_pg(sp_ulysses_degree, sp_ring_degree):
  function setup_parallel (line 93) | def setup_parallel(sp_size=1, tp_size=1, ring_size=1):
  function get_ulysess_mesh (line 108) | def get_ulysess_mesh():
  function get_ring_mesh (line 112) | def get_ring_mesh():
  function get_ulysess_group (line 116) | def get_ulysess_group():
  function get_ring_group (line 120) | def get_ring_group():
  function get_ulysess_world_size (line 124) | def get_ulysess_world_size():
  function get_ring_world_size (line 135) | def get_ring_world_size():
  function get_dp_mesh (line 146) | def get_dp_mesh():
  function get_dp_group (line 150) | def get_dp_group():
  function get_dp_world_size (line 154) | def get_dp_world_size():
  function get_sp_mesh (line 165) | def get_sp_mesh():
  function get_sp_group (line 169) | def get_sp_group():
  function get_sp_world_size (line 173) | def get_sp_world_size():
  function get_tp_mesh (line 184) | def get_tp_mesh():
  function get_tp_group (line 188) | def get_tp_group():
  function get_tp_world_size (line 192) | def get_tp_world_size():

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/comm/all_to_all.py
  function all_to_all_4D (line 15) | def all_to_all_4D(
  class SeqAllToAll4D (line 96) | class SeqAllToAll4D(torch.autograd.Function):
    method forward (line 98) | def forward(
    method backward (line 113) | def backward(ctx: Any, *grad_output: Tensor) -> Tuple[None, Tensor, No...
  function all_to_all_5D (line 124) | def all_to_all_5D(
  class SeqAllToAll5D (line 207) | class SeqAllToAll5D(torch.autograd.Function):
    method forward (line 209) | def forward(
    method backward (line 224) | def backward(ctx: Any, *grad_output: Tensor) -> Tuple[None, Tensor, No...

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/comm/extract_local.py
  function stripe_extract_local (line 7) | def stripe_extract_local(value, rank, world_size, rd, ud, *args, **kwargs):
  function basic_extract_local (line 29) | def basic_extract_local(value, rank, world_size, *args, **kwargs):
  function zigzag_extract_local (line 33) | def zigzag_extract_local(value, rank, world_size, rd, ud, dim=1, *args, ...

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/globals.py
  class Singleton (line 4) | class Singleton:
    method __new__ (line 7) | def __new__(cls, *args, **kwargs):
  class ProcessGroupSingleton (line 13) | class ProcessGroupSingleton(Singleton):
    method __init__ (line 14) | def __init__(self):
  function set_seq_parallel_pg (line 22) | def set_seq_parallel_pg(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/hybrid/async_attn_layer.py
  class AsyncLongContextAttention (line 11) | class AsyncLongContextAttention(torch.nn.Module):
    method __init__ (line 21) | def __init__(
    method forward (line 42) | def forward(
    method backward (line 195) | def backward(self, *args, **kwargs):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/hybrid/attn_layer.py
  class LongContextAttention (line 15) | class LongContextAttention(torch.nn.Module):
    method __init__ (line 25) | def __init__(
    method forward (line 45) | def forward(
  class LongContextAttentionQKVPacked (line 134) | class LongContextAttentionQKVPacked(torch.nn.Module):
    method __init__ (line 144) | def __init__(
    method forward (line 164) | def forward(
  function llama3_varlen_attention_sp_ulysses_ring (line 225) | def llama3_varlen_attention_sp_ulysses_ring(
  class LongContextVarLenAttentionForLlaMa3 (line 299) | class LongContextVarLenAttentionForLlaMa3(torch.nn.Module):
    method __init__ (line 309) | def __init__(
    method forward (line 324) | def forward(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/ring/llama3_flash_attn_varlen.py
  class AsyncHandles (line 10) | class AsyncHandles:
    method __init__ (line 12) | def __init__(self) -> None:
    method register (line 15) | def register(self, handle):
    method wait (line 18) | def wait(self):
  function llama3_flash_attn_prepare_cu_seqlens (line 24) | def llama3_flash_attn_prepare_cu_seqlens(cu_seqlens, causal, rank, world...
  function llama3_flash_attn_varlen_forward (line 63) | def llama3_flash_attn_varlen_forward(
  function llama3_flash_attn_varlen_backward (line 164) | def llama3_flash_attn_varlen_backward(
  class Llama3FlashAttnVarlenFunc (line 313) | class Llama3FlashAttnVarlenFunc(torch.autograd.Function):
    method forward (line 315) | def forward(
    method backward (line 375) | def backward(ctx, dout, *args):
  function llama3_flash_attn_varlen_qkvpacked_func (line 401) | def llama3_flash_attn_varlen_qkvpacked_func(
  function llama3_flash_attn_varlen_kvpacked_func (line 439) | def llama3_flash_attn_varlen_kvpacked_func(
  function llama3_flash_attn_varlen_func (line 478) | def llama3_flash_attn_varlen_func(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/ring/ring_flash_attn.py
  function ring_flash_attn_forward (line 7) | def ring_flash_attn_forward(
  function ring_flash_attn_backward (line 60) | def ring_flash_attn_backward(
  class RingFlashAttnFunc (line 144) | class RingFlashAttnFunc(torch.autograd.Function):
    method forward (line 146) | def forward(
    method backward (line 190) | def backward(ctx, dout, *args):
  function ring_flash_attn_qkvpacked_func (line 210) | def ring_flash_attn_qkvpacked_func(
  function ring_flash_attn_kvpacked_func (line 236) | def ring_flash_attn_kvpacked_func(
  function ring_flash_attn_func (line 263) | def ring_flash_attn_func(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/ring/ring_flash_attn_varlen.py
  function ring_flash_attn_varlen_forward (line 25) | def ring_flash_attn_varlen_forward(
  function ring_flash_attn_varlen_backward (line 95) | def ring_flash_attn_varlen_backward(
  class RingFlashAttnVarlenFunc (line 184) | class RingFlashAttnVarlenFunc(torch.autograd.Function):
    method forward (line 186) | def forward(
    method backward (line 235) | def backward(ctx, dout, *args):
  function ring_flash_attn_varlen_qkvpacked_func (line 257) | def ring_flash_attn_varlen_qkvpacked_func(
  function ring_flash_attn_varlen_kvpacked_func (line 287) | def ring_flash_attn_varlen_kvpacked_func(
  function ring_flash_attn_varlen_func (line 318) | def ring_flash_attn_varlen_func(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/ring/stripe_flash_attn.py
  function stripe_flash_attn_forward (line 7) | def stripe_flash_attn_forward(
  function stripe_flash_attn_backward (line 81) | def stripe_flash_attn_backward(
  class StripeFlashAttnFunc (line 199) | class StripeFlashAttnFunc(torch.autograd.Function):
    method forward (line 201) | def forward(
    method backward (line 245) | def backward(ctx, dout, *args):
  function stripe_flash_attn_qkvpacked_func (line 265) | def stripe_flash_attn_qkvpacked_func(
  function stripe_flash_attn_kvpacked_func (line 291) | def stripe_flash_attn_kvpacked_func(
  function stripe_flash_attn_func (line 318) | def stripe_flash_attn_func(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/ring/triton_utils.py
  function flatten_kernel (line 7) | def flatten_kernel(
  function flatten_varlen_lse (line 39) | def flatten_varlen_lse(lse, cu_seqlens):
  function unflatten_kernel (line 71) | def unflatten_kernel(
  function unflatten_varlen_lse (line 103) | def unflatten_varlen_lse(lse, cu_seqlens, max_seqlen: int):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/ring/utils.py
  function get_default_args (line 14) | def get_default_args(func):
  function _update_out_and_lse (line 25) | def _update_out_and_lse(
  function update_out_and_lse (line 45) | def update_out_and_lse(
  function flatten_varlen_lse (line 69) | def flatten_varlen_lse(lse, cu_seqlens):
  function unflatten_varlen_lse (line 78) | def unflatten_varlen_lse(lse, cu_seqlens, max_seqlen: int):
  class RingComm (line 90) | class RingComm:
    method __init__ (line 91) | def __init__(self, process_group: dist.ProcessGroup):
    method send_recv (line 105) | def send_recv(
    method commit (line 121) | def commit(self):
    method wait (line 126) | def wait(self):

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/ring/zigzag_ring_flash_attn.py
  function zigzag_ring_flash_attn_forward (line 7) | def zigzag_ring_flash_attn_forward(
  function zigzag_ring_flash_attn_backward (line 81) | def zigzag_ring_flash_attn_backward(
  class ZigZagRingFlashAttnFunc (line 187) | class ZigZagRingFlashAttnFunc(torch.autograd.Function):
    method forward (line 189) | def forward(
    method backward (line 233) | def backward(ctx, dout, *args):
  function zigzag_ring_flash_attn_qkvpacked_func (line 253) | def zigzag_ring_flash_attn_qkvpacked_func(
  function zigzag_ring_flash_attn_kvpacked_func (line 279) | def zigzag_ring_flash_attn_kvpacked_func(
  function zigzag_ring_flash_attn_func (line 306) | def zigzag_ring_flash_attn_func(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/ring/zigzag_ring_flash_attn_varlen.py
  function get_half_index (line 24) | def get_half_index(cu_seqlens, *, front: bool):
  function get_half_lse (line 43) | def get_half_lse(lse, cu_seqlens, *, front: bool):
  function zigzag_ring_flash_attn_varlen_forward (line 74) | def zigzag_ring_flash_attn_varlen_forward(
  function zigzag_ring_flash_attn_varlen_backward (line 184) | def zigzag_ring_flash_attn_varlen_backward(
  class ZigZagRingFlashAttnVarlenFunc (line 305) | class ZigZagRingFlashAttnVarlenFunc(torch.autograd.Function):
    method forward (line 307) | def forward(
    method backward (line 369) | def backward(ctx, dout, *args):
  function zigzag_ring_flash_attn_varlen_qkvpacked_func (line 400) | def zigzag_ring_flash_attn_varlen_qkvpacked_func(
  function zigzag_ring_flash_attn_varlen_kvpacked_func (line 430) | def zigzag_ring_flash_attn_varlen_kvpacked_func(
  function zigzag_ring_flash_attn_varlen_func (line 461) | def zigzag_ring_flash_attn_varlen_func(

FILE: xtuner-train_internvideo2_5/xtuner/_lite/yunchang/ulysses/attn_layer.py
  function torch_attn (line 16) | def torch_attn(query,
  class UlyssesAttention (line 42) | class UlyssesAttention(torch.nn.Module):
    method __init__ (line 52) | def __init__(
    method forward (line 70) | def forward(

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/alpaca.py
  function alpaca_enzh_dataset (line 14) | def alpaca_enzh_dataset(tokenizer,
  function alpaca_enzh_data_collator (line 41) | def alpaca_enzh_data_collator(return_hf_format=False):
  function alpaca_zh_dataset (line 45) | def alpaca_zh_dataset(tokenizer,
  function alpaca_zh_data_collator (line 66) | def alpaca_zh_data_collator(return_hf_format=False):
  function alpaca_dataset (line 70) | def alpaca_dataset(tokenizer,
  function alpaca_data_collator (line 91) | def alpaca_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/arxiv.py
  function arxiv_dataset (line 12) | def arxiv_dataset(tokenizer,
  function arxiv_data_collator (line 37) | def arxiv_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/code_alpaca.py
  function code_alpaca_dataset (line 12) | def code_alpaca_dataset(tokenizer,
  function code_alpaca_data_collator (line 33) | def code_alpaca_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/colorist.py
  function colorist_dataset (line 12) | def colorist_dataset(tokenizer,
  function colorist_data_collator (line 33) | def colorist_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/lawyer.py
  function lawyer_dataset (line 15) | def lawyer_dataset(tokenizer,
  function lawyer_data_collator (line 40) | def lawyer_data_collator(return_hf_format=False):
  function lawyer_crime_dataset (line 44) | def lawyer_crime_dataset(tokenizer,
  function lawyer_crime_data_collator (line 68) | def lawyer_crime_data_collator(return_hf_format=False):
  function lawyer_reference_dataset (line 72) | def lawyer_reference_dataset(tokenizer,
  function lawyer_reference_data_collator (line 96) | def lawyer_reference_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/medical.py
  function medical_dataset (line 12) | def medical_dataset(tokenizer,
  function medical_data_collator (line 33) | def medical_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/moss_003_sft.py
  function moss_003_sft_dataset (line 10) | def moss_003_sft_dataset(tokenizer,
  function moss_003_sft_data_collator (line 29) | def moss_003_sft_data_collator(return_hf_format=False):
  function moss_003_sft_no_plugins_dataset (line 33) | def moss_003_sft_no_plugins_dataset(tokenizer,
  function moss_003_sft_no_plugins_data_collator (line 50) | def moss_003_sft_no_plugins_data_collator(return_hf_format=False):
  function moss_003_sft_plugins_dataset (line 54) | def moss_003_sft_plugins_dataset(tokenizer,
  function moss_003_sft_plugins_data_collator (line 71) | def moss_003_sft_plugins_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/oasst1.py
  function oasst1_dataset (line 12) | def oasst1_dataset(tokenizer,
  function oasst1_data_collator (line 33) | def oasst1_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/open_orca.py
  function openorca_dataset (line 12) | def openorca_dataset(tokenizer,
  function openorca_data_collator (line 33) | def openorca_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/sql.py
  function sql_dataset (line 12) | def sql_dataset(tokenizer,
  function sql_data_collator (line 33) | def sql_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/tiny_codes.py
  function tiny_codes_dataset (line 12) | def tiny_codes_dataset(tokenizer,
  function tiny_codes_data_collator (line 33) | def tiny_codes_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/datasets/wizardlm.py
  function wizardlm_dataset (line 12) | def wizardlm_dataset(tokenizer,
  function wizardlm_data_collator (line 33) | def wizardlm_data_collator(return_hf_format=False):

FILE: xtuner-train_internvideo2_5/xtuner/apis/model.py
  function build_qlora_model (line 12) | def build_qlora_model(model_name_or_path,
  function build_lora_model (line 52) | def build_lora_model(model_name_or_path,
  function build_model (line 78) | def build_model(model_name_or_path, return_tokenizer=True):

FILE: xtuner-train_internvideo2_5/xtuner/apis/training_args.py
  class DefaultTrainingArguments (line 12) | class DefaultTrainingArguments(TrainingArguments):

FILE: xtuner-train_internvideo2_5/xtuner/configs/__init__.py
  function get_cfgs_name_path (line 5) | def get_cfgs_name_path():

FILE: xtuner-train_internvideo2_5/xtuner/configs/llava/llama3_8b_instruct_clip_vit_large_p14_336/convert_xtuner_weights_to_hf.py
  function convert_state_dict_to_hf (line 25) | def convert_state_dict_to_hf(state_dict, mapping):
  function convert_to_hf (line 38) | def convert_to_hf(text_model_id, vision_model_id, projector_weight, save...
  function main (line 131) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/configs/llava/llama3_8b_instruct_clip_vit_large_p14_336/convert_xtuner_weights_to_llava.py
  function convert_state_dict_to_hf (line 27) | def convert_state_dict_to_hf(state_dict, mapping):
  function convert_to_llava (line 39) | def convert_to_llava(text_model_id, vision_model_id, projector_weight,
  function main (line 94) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/configs/llava/phi3_mini_4k_instruct_clip_vit_large_p14_336/convert_phi_to_llama.py
  function convert_phi_to_llama (line 13) | def convert_phi_to_llama(phi_path, save_path):
  function main (line 91) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/configs/llava/phi3_mini_4k_instruct_clip_vit_large_p14_336/convert_xtuner_weights_to_hf.py
  function convert_state_dict_to_hf (line 25) | def convert_state_dict_to_hf(state_dict, mapping):
  function convert_to_hf (line 38) | def convert_to_hf(text_model_id, vision_model_id, projector_weight, save...
  function main (line 128) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/configs/llava/phi3_mini_4k_instruct_clip_vit_large_p14_336/convert_xtuner_weights_to_llava.py
  function convert_state_dict_to_hf (line 27) | def convert_state_dict_to_hf(state_dict, mapping):
  function convert_to_llava (line 39) | def convert_to_llava(text_model_id, vision_model_id, projector_weight,
  function main (line 94) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/dataset/collate_fns/default_collate_fn.py
  function default_collate_fn (line 12) | def default_collate_fn(instances: Sequence[Dict],

FILE: xtuner-train_internvideo2_5/xtuner/dataset/collate_fns/mmlu_collate_fn.py
  function mmlu_collate_fn (line 10) | def mmlu_collate_fn(instances: Sequence[Dict],

FILE: xtuner-train_internvideo2_5/xtuner/dataset/collate_fns/preference_collate_fn.py
  function preference_collate_fn (line 13) | def preference_collate_fn(instances: Sequence[Dict],

FILE: xtuner-train_internvideo2_5/xtuner/dataset/concat_dataset.py
  class ConcatDataset (line 7) | class ConcatDataset(TorchConcatDataset):
    method __init__ (line 9) | def __init__(self, datasets):
    method __repr__ (line 15) | def __repr__(self):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/huggingface.py
  function get_lengths (line 18) | def get_lengths(example):
  function build_origin_dataset (line 22) | def build_origin_dataset(dataset, split):
  function map_dataset (line 39) | def map_dataset(dataset, dataset_map_fn, map_num_proc):
  function add_template_to_dataset (line 54) | def add_template_to_dataset(dataset, template_map_fn, map_num_proc):
  function tokenize_dataset (line 67) | def tokenize_dataset(dataset, tokenizer, max_length, with_image_token,
  function pack_dataset (line 88) | def pack_dataset(dataset, max_length, use_varlen_attn, shuffle_before_pack,
  function process (line 100) | def process(dataset,
  function process_hf_dataset (line 222) | def process_hf_dataset(dataset,

FILE: xtuner-train_internvideo2_5/xtuner/dataset/intern_repo.py
  class JsonlDataset (line 22) | class JsonlDataset(torch.utils.data.Dataset):
    method __init__ (line 36) | def __init__(self, path: str, min_length=50):
    method __getitem__ (line 62) | def __getitem__(self, idx):
    method get_dataset_name (line 84) | def get_dataset_name(self):
    method _get_mmap (line 87) | def _get_mmap(self):
    method __setstate__ (line 100) | def __setstate__(self, state):
    method __getstate__ (line 104) | def __getstate__(self):
    method __del__ (line 111) | def __del__(self):
    method exists (line 118) | def exists(path):
    method __len__ (line 121) | def __len__(self):
  class PackedDataset (line 128) | class PackedDataset(torch.utils.data.Dataset):
    method __init__ (line 137) | def __init__(self, dataset, packed_length: int = 8192, seed: int = 1024):
    method __len__ (line 162) | def __len__(self):
    method search_sample_index (line 165) | def search_sample_index(self, pack_idx: int = 0):
    method mapping (line 172) | def mapping(self, pack_idx: int = 0):
    method build_pack (line 192) | def build_pack(self, begin_sample_idx: int, begin_token_id: int,
    method __getitem__ (line 228) | def __getitem__(self, item: int):
  function load_intern_repo_tokenized_dataset (line 235) | def load_intern_repo_tokenized_dataset(folder,
  function load_intern_repo_untokenized_dataset (line 268) | def load_intern_repo_untokenized_dataset(processed_dataset_dict_path=None,
  function build_packed_dataset_rank0 (line 332) | def build_packed_dataset_rank0(dataset_cfg, packed_length=8192, seed=1024):
  function build_packed_dataset (line 352) | def build_packed_dataset(*args, **kwargs):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/json_dataset.py
  function load_json_file (line 7) | def load_json_file(data_files=None, data_dir=None, suffix=None):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/llava.py
  function load_jsonl (line 19) | def load_jsonl(json_file):
  class LLaVADataset (line 28) | class LLaVADataset(Dataset):
    method __init__ (line 30) | def __init__(self,
    method modality_length (line 89) | def modality_length(self):
    method __len__ (line 98) | def __len__(self):
    method __getitem__ (line 101) | def __getitem__(self, index):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/alpaca_map_fn.py
  function alpaca_map_fn (line 4) | def alpaca_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/alpaca_zh_map_fn.py
  function alpaca_zh_map_fn (line 4) | def alpaca_zh_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/arxiv_map_fn.py
  function arxiv_map_fn (line 5) | def arxiv_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/code_alpaca_map_fn.py
  function code_alpaca_map_fn (line 5) | def code_alpaca_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/colors_map_fn.py
  function colors_map_fn (line 5) | def colors_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/crime_kg_assitant_map_fn.py
  function crime_kg_assitant_map_fn (line 5) | def crime_kg_assitant_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/default_map_fn.py
  function default_map_fn (line 2) | def default_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/law_reference_map_fn.py
  function law_reference_map_fn (line 5) | def law_reference_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/llava_map_fn.py
  function llava_image_only_map_fn (line 5) | def llava_image_only_map_fn(example):
  function llava_map_fn (line 25) | def llava_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/medical_map_fn.py
  function medical_map_fn (line 5) | def medical_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/msagent_map_fn.py
  function replace_think (line 9) | def replace_think(match):
  function replace_exec (line 22) | def replace_exec(match):
  function extract_json_objects (line 34) | def extract_json_objects(text, decoder=json.JSONDecoder()):
  function msagent_react_map_fn (line 53) | def msagent_react_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/oasst1_map_fn.py
  function oasst1_map_fn (line 2) | def oasst1_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/openai_map_fn.py
  function openai_map_fn (line 2) | def openai_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/openorca_map_fn.py
  function openorca_map_fn (line 2) | def openorca_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/pretrain_map_fn.py
  function pretrain_map_fn (line 2) | def pretrain_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/sql_map_fn.py
  function sql_map_fn (line 5) | def sql_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/stack_exchange_map_fn.py
  function stack_exchange_map_fn (line 2) | def stack_exchange_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/tiny_codes_map_fn.py
  function tiny_codes_map_fn (line 5) | def tiny_codes_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/dataset_map_fns/wizardlm_map_fn.py
  function wizardlm_map_fn (line 2) | def wizardlm_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/map_fns/template_map_fn.py
  function template_map_fn (line 7) | def template_map_fn(example, template):
  function template_map_fn_factory (line 33) | def template_map_fn_factory(template):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/modelscope.py
  function process_ms_dataset (line 8) | def process_ms_dataset(dataset, split='train', *args, **kwargs):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/moss_sft.py
  class MOSSSFTDataset (line 15) | class MOSSSFTDataset(Dataset):
    method __init__ (line 17) | def __init__(self, data_file, tokenizer, max_length=2048, bot_name=None):
    method load_data (line 43) | def load_data(self):
    method process_data (line 118) | def process_data(self):
    method __len__ (line 125) | def __len__(self):
    method __getitem__ (line 128) | def __getitem__(self, index):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/preference_dataset.py
  function _worker (line 24) | def _worker(
  function _chunk_data_to_queue (line 41) | def _chunk_data_to_queue(data_queue: Queue, data: List[Dict], chunk_size...
  function _multi_progress (line 61) | def _multi_progress(tokenize_fun_p, dataset, nproc, task_num, chunksize,
  function load_jsonl_dataset (line 90) | def load_jsonl_dataset(data_files=None, data_dir=None, suffix=None):
  function tokenize (line 110) | def tokenize(pair: str,
  class PreferenceDataset (line 156) | class PreferenceDataset(Dataset):
    method __init__ (line 158) | def __init__(
    method __len__ (line 194) | def __len__(self):
    method __getitem__ (line 197) | def __getitem__(self, idx):
  class PackedDatasetWrapper (line 201) | class PackedDatasetWrapper(Dataset):
    method __init__ (line 203) | def __init__(self,
    method __len__ (line 252) | def __len__(self):
    method __getitem__ (line 255) | def __getitem__(self, index):
  function unpack_seq (line 280) | def unpack_seq(seq, cu_seqlens):
  function broad_cast_dataset (line 288) | def broad_cast_dataset(dataset):
  function map_dataset (line 309) | def map_dataset(dataset, dataset_map_fn, map_num_proc):
  function build_preference_dataset (line 324) | def build_preference_dataset(
  function intel_orca_dpo_map_fn (line 368) | def intel_orca_dpo_map_fn(example):
  function orpo_dpo_mix_40k_map_fn (line 381) | def orpo_dpo_mix_40k_map_fn(example):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/refcoco_json.py
  class RefCOCOJsonDataset (line 26) | class RefCOCOJsonDataset(LLaVADataset):
    method __init__ (line 37) | def __init__(
    method reformat_data (line 81) | def reformat_data(self, json_data):
    method gen_refcoco_conversations (line 91) | def gen_refcoco_conversations(cls, data, instruction_template='{}'):
    method get_data_json (line 127) | def get_data_json(
    method normalize_bbox (line 168) | def normalize_bbox(cls, bbox, height, width):
  class RefCOCOJsonEvalDataset (line 176) | class RefCOCOJsonEvalDataset(RefCOCOJsonDataset):
    method reformat_data (line 179) | def reformat_data(self, json_data):
  class InvRefCOCOJsonDataset (line 188) | class InvRefCOCOJsonDataset(RefCOCOJsonDataset):
    method gen_refcoco_conversations (line 200) | def gen_refcoco_conversations(cls, data, instruction_template='{}'):
  class REFER (line 241) | class REFER:
    method __init__ (line 243) | def __init__(self, data_root, vis_root, dataset='refcoco', splitBy='un...
    method createIndex (line 276) | def createIndex(self):
    method getRefIds (line 340) | def getRefIds(self, image_ids=[], cat_ids=[], ref_ids=[], split=''):
    method getAnnIds (line 372) | def getAnnIds(self, image_ids=[], cat_ids=[], ref_ids=[]):
    method getImgIds (line 397) | def getImgIds(self, ref_ids=[]):
    method getCatIds (line 408) | def getCatIds(self):
    method loadRefs (line 411) | def loadRefs(self, ref_ids=[]):
    method loadAnns (line 417) | def loadAnns(self, ann_ids=[]):
    method loadImgs (line 423) | def loadImgs(self, image_ids=[]):
    method loadCats (line 429) | def loadCats(self, cat_ids=[]):
    method getRefBox (line 435) | def getRefBox(self, ref_id):
    method showRef (line 440) | def showRef(self, ref, seg_box='box'):

FILE: xtuner-train_internvideo2_5/xtuner/dataset/samplers/intern_repo.py
  class InternRepoSampler (line 13) | class InternRepoSampler(Sampler):
    method __init__ (line 15) | def __init__(self,
    method __iter__ (line 35) | def __iter__(self) -> Iterator[int]:
    method __len__ (line 54) | def __len__(self) -> int:
    method set_epoch (line 58) | def set_epoch(self, epoch: int) -> None:
  class InternlmRepoSampler (line 71) | class InternlmRepoSampler(InternRepoSampler):
    method __init__ (line 73) | def __init__(self,

FILE: xtuner-train_internvideo2_5/xtuner/dataset/samplers/length_grouped.py
  function get_length_grouped_indices (line 11) | def get_length_grouped_indices(lengths, group_batch_size, generator=None):
  class LengthGroupedSampler (line 72) | class LengthGroupedSampler(Sampler):
    method __init__ (line 74) | def __init__(self,
    method __iter__ (line 124) | def __iter__(self) -> Iterator[int]:
    method __len__ (line 144) | def __len__(self) -> int:
    method set_epoch (line 148) | def set_epoch(self, epoch: int) -> None:

FILE: xtuner-train_internvideo2_5/xtuner/dataset/utils.py
  function get_bos_eos_token_ids (line 15) | def get_bos_eos_token_ids(tokenizer):
  function encode_fn (line 36) | def encode_fn(example,
  class Packer (line 132) | class Packer:
    method __init__ (line 135) | def __init__(self,
    method get_cumulative_len (line 146) | def get_cumulative_len(self, chunk_num):
    method get_position_ids (line 179) | def get_position_ids(self, cumulative_len):
    method __call__ (line 191) | def __call__(self, batch):
  function expand2square (line 245) | def expand2square(pil_img, background_color):
  function load_image (line 259) | def load_image(image_file):
  function decode_base64_to_image (line 268) | def decode_base64_to_image(base64_string):

FILE: xtuner-train_internvideo2_5/xtuner/engine/_strategy/deepspeed.py
  class DeepSpeedStrategy (line 11) | class DeepSpeedStrategy(MMEngineDeepSpeedStrategy):
    method __init__ (line 13) | def __init__(self, *args, **kwargs):
    method _wrap_model (line 24) | def _wrap_model(self, model):
    method save_checkpoint (line 33) | def save_checkpoint(self, *args, **kwargs) -> None:
    method load_checkpoint (line 45) | def load_checkpoint(self, *args, **kwargs) -> None:
    method resume (line 54) | def resume(self, *args, **kwargs) -> None:
    method _setup_distributed (line 63) | def _setup_distributed(  # type: ignore

FILE: xtuner-train_internvideo2_5/xtuner/engine/hooks/dataset_info_hook.py
  function split_list (line 8) | def split_list(lst, value):
  class DatasetInfoHook (line 21) | class DatasetInfoHook(Hook):
    method __init__ (line 23) | def __init__(self, tokenizer, is_intern_repo_dataset=False):
    method log (line 27) | def log(self, runner, dataset, mode='train'):
    method before_train (line 42) | def before_train(self, runner) -> None:
    method before_val (line 52) | def before_val(self, runner) -> None:
    method before_test (line 56) | def before_test(self, runner) -> None:

FILE: xtuner-train_internvideo2_5/xtuner/engine/hooks/evaluate_chat_hook.py
  class EvaluateChatHook (line 20) | class EvaluateChatHook(Hook):
    method __init__ (line 24) | def __init__(self,
    method _save_eval_output (line 98) | def _save_eval_output(self, runner, eval_outputs):
    method _eval_images (line 106) | def _eval_images(self,
    method _eval_language (line 167) | def _eval_language(self,
    method _generate_samples (line 194) | def _generate_samples(self,
    method before_train (line 232) | def before_train(self, runner):
    method _is_save_checkpoint (line 236) | def _is_save_checkpoint(self, runner):
    method after_train_iter (line 254) | def after_train_iter(self,
    method after_train (line 273) | def after_train(self, runner):
    method after_val (line 277) | def after_val(self, runner) -> None:

FILE: xtuner-train_internvideo2_5/xtuner/engine/hooks/hf_checkpoint_hook.py
  class HFCheckpointHook (line 19) | class HFCheckpointHook(Hook):
    method __init__ (line 23) | def __init__(self, out_dir: Optional[Union[str, Path]] = None) -> None:
    method _use_shard_moe (line 27) | def _use_shard_moe(llm):
    method after_run (line 32) | def after_run(self, runner) -> None:

FILE: xtuner-train_internvideo2_5/xtuner/engine/hooks/throughput_hook.py
  class ThroughputHook (line 16) | class ThroughputHook(Hook):
    method __init__ (line 22) | def __init__(self,
    method _guess_is_casual_attn (line 37) | def _guess_is_casual_attn(model):
    method _get_batch_size_and_sequence_len (line 47) | def _get_batch_size_and_sequence_len(data_batch):
    method _guess_use_activation_checkpointing (line 55) | def _guess_use_activation_checkpointing(model):
    method before_run (line 61) | def before_run(self, runner) -> None:
    method after_train_iter (line 88) | def after_train_iter(self,

FILE: xtuner-train_internvideo2_5/xtuner/engine/hooks/varlen_attn_args_to_messagehub_hook.py
  class VarlenAttnArgsToMessageHubHook (line 11) | class VarlenAttnArgsToMessageHubHook(Hook):
    method before_train_iter (line 13) | def before_train_iter(self,
    method after_train_iter (line 31) | def after_train_iter(self,
    method before_val_iter (line 41) | def before_val_iter(self,
    method after_val_iter (line 68) | def after_val_iter(self,

FILE: xtuner-train_internvideo2_5/xtuner/engine/runner/loops.py
  class TrainLoop (line 8) | class TrainLoop(IterBasedTrainLoop):
    method __init__ (line 10) | def __init__(self,

FILE: xtuner-train_internvideo2_5/xtuner/entry_point.py
  function list_dataset_format (line 127) | def list_dataset_format():
  function list_cfg (line 132) | def list_cfg():
  function copy_cfg (line 137) | def copy_cfg():
  function log_dataset (line 142) | def log_dataset():
  function check_custom_dataset (line 147) | def check_custom_dataset():
  function train (line 152) | def train():
  function test (line 157) | def test():
  function chat (line 162) | def chat():
  function mmbench (line 167) | def mmbench():
  function pth_to_hf (line 172) | def pth_to_hf():
  function merge (line 177) | def merge():
  function split (line 182) | def split():
  function arxiv_preprocess (line 187) | def arxiv_preprocess():
  function convert_refcoco (line 192) | def convert_refcoco():
  function convert_help_msg (line 197) | def convert_help_msg():
  function preprocess_help_msg (line 201) | def preprocess_help_msg():
  function eval_refcoco (line 205) | def eval_refcoco():
  function cli (line 244) | def cli():

FILE: xtuner-train_internvideo2_5/xtuner/evaluation/metrics/mmlu_metric.py
  class MMLUMetric (line 14) | class MMLUMetric(BaseMetric):
    method __init__ (line 91) | def __init__(self, tokenizer, *args, **kwargs):
    method ABCD_to_0123 (line 102) | def ABCD_to_0123(abcd):
    method find_first_zero_index (line 106) | def find_first_zero_index(tensor):
    method accuracy (line 114) | def accuracy(preds, gts):
    method process (line 120) | def process(self, data_batch: Any, data_samples: Sequence[dict]) -> None:
    method compute_metrics (line 147) | def compute_metrics(self, results: list) -> dict:
    method _print_results (line 236) | def _print_results(self, table_metrics: dict) -> None:

FILE: xtuner-train_internvideo2_5/xtuner/evaluation/metrics/reward_metric.py
  class RewardMetric (line 12) | class RewardMetric(BaseMetric):
    method __init__ (line 17) | def __init__(self,
    method process (line 22) | def process(self, data_batch, data_samples: Sequence[dict]):
    method compute_metrics (line 49) | def compute_metrics(self, results: List):

FILE: xtuner-train_internvideo2_5/xtuner/model/dpo.py
  function create_reference_model (line 19) | def create_reference_model(model):
  class DPO (line 36) | class DPO(SupervisedFinetune):
    method __init__ (line 39) | def __init__(self,
    method _gather_masked_logits (line 55) | def _gather_masked_logits(self, logits, labels, mask):
    method get_logps (line 61) | def get_logps(
    method get_var_len_atten_logps (line 89) | def get_var_len_atten_logps(self, all_logits, all_ref_logits, labels,
    method _split_for_sequence_parallel (line 141) | def _split_for_sequence_parallel(data):
    method compute_loss (line 153) | def compute_loss(self, data, data_samples=None):

FILE: xtuner-train_internvideo2_5/xtuner/model/llava.py
  function convert_state_dict_to_hf (line 30) | def convert_state_dict_to_hf(state_dict, mapping):
  class LLaVAModel (line 42) | class LLaVAModel(BaseModel):
    method __init__ (line 44) | def __init__(self,
    method _parse_lora_config (line 121) | def _parse_lora_config(self, lora_config):
    method _prepare_llm_for_lora (line 127) | def _prepare_llm_for_lora(self,
    method _prepare_visual_encoder_for_lora (line 138) | def _prepare_visual_encoder_for_lora(self,
    method gradient_checkpointing_enable (line 147) | def gradient_checkpointing_enable(self):
    method activation_checkpointing_enable (line 150) | def activation_checkpointing_enable(self):
    method gradient_checkpointing_disable (line 155) | def gradient_checkpointing_disable(self):
    method activation_checkpointing_disable (line 158) | def activation_checkpointing_disable(self):
    method init_weights (line 163) | def init_weights(self):
    method state_dict (line 166) | def state_dict(self, *args, **kwargs):
    method _prepare_for_long_context_training (line 194) | def _prepare_for_long_context_training(cfg, llm_cfg,
    method _prepare_for_flash_attn (line 221) | def _prepare_for_flash_attn(cfg, llm_cfg):
    method _prepare_for_qlora_zero3 (line 250) | def _prepare_for_qlora_zero3(cfg):
    method _dispatch_lm_model_cfg (line 266) | def _dispatch_lm_model_cfg(self, cfg, max_position_embeddings=None):
    method _build_from_cfg_or_module (line 277) | def _build_from_cfg_or_module(self, cfg_or_mod):
    method forward (line 286) | def forward(self, data, data_samples=None, mode='loss'):
    method _forward (line 313) | def _forward(self, data, data_samples=None):
    method predict (line 319) | def predict(self, data, data_samples=None):
    method compute_loss (line 324) | def compute_loss(self, data, data_samples=None):
    method __getattr__ (line 329) | def __getattr__(self, name: str):
    method to_hf (line 335) | def to_hf(self,
    method to_xtuner_llava (line 352) | def to_xtuner_llava(self,
    method to_huggingface_llava (line 402) | def to_huggingface_llava(self,
    method to_official_llava (line 536) | def to_official_llava(self,

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/__init__.py
  function log_once (line 137) | def log_once(func):
  function dispatch_attn_forward (line 150) | def dispatch_attn_forward(model):
  function dispatch_varlen_attn_forward (line 178) | def dispatch_varlen_attn_forward(model):
  function dispatch_rmsnorm_forward (line 207) | def dispatch_rmsnorm_forward(model):
  function replace_rote (line 226) | def replace_rote(model):
  function dispatch_modules (line 254) | def dispatch_modules(model, use_varlen_attn=False):

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/attention.py
  function flash_attn_wo_mask (line 15) | def flash_attn_wo_mask(
  function flash_attn_w_mask (line 36) | def flash_attn_w_mask(
  function varlen_flash_attn (line 68) | def varlen_flash_attn(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/baichuan.py
  function baichuan2_norm_head_forward (line 9) | def baichuan2_norm_head_forward(self, hidden_states):
  function rotate_half (line 14) | def rotate_half(x):
  function apply_rotary_pos_emb (line 21) | def apply_rotary_pos_emb(q, k, cos_, sin_, position_ids):
  function baichuan_7b_attn_forward (line 31) | def baichuan_7b_attn_forward(
  function baichuan_13b_attn_forward (line 75) | def baichuan_13b_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/cohere.py
  class Cache (line 17) | class Cache:
  function cohere_attn_forward (line 21) | def cohere_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/deepseek_v2.py
  function deepseek_attn_forward (line 19) | def deepseek_attn_forward(
  function deepseek_varlen_attn_forward (line 158) | def deepseek_varlen_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/internlm.py
  class InternLMRotaryEmbedding (line 21) | class InternLMRotaryEmbedding(torch.nn.Module):
    method __init__ (line 23) | def __init__(self,
    method forward (line 43) | def forward(self, x, seq_len):
  function rotate_half (line 64) | def rotate_half(x):
  function apply_rotary_pos_emb (line 71) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids):
  function internlm_attn_forward (line 79) | def internlm_attn_forward(
  function internlm_varlen_attn_forward (line 140) | def internlm_varlen_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/internlm2.py
  function rotate_half (line 16) | def rotate_half(x):
  function apply_rotary_pos_emb (line 23) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  function repeat_kv (line 31) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  function repeat_kv_bshd (line 49) | def repeat_kv_bshd(hidden_states: torch.Tensor, n_rep: int) -> torch.Ten...
  function internlm2_attn_forward (line 63) | def internlm2_attn_forward(
  function internlm2_varlen_attn_forward (line 174) | def internlm2_varlen_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/llama.py
  class Cache (line 20) | class Cache:
  function repeat_kv_bshd (line 24) | def repeat_kv_bshd(hidden_states: torch.Tensor, n_rep: int) -> torch.Ten...
  function llama_attn_forward (line 38) | def llama_attn_forward(
  function llama_attn_forward_legacy (line 160) | def llama_attn_forward_legacy(
  function llama_varlen_attn_forward (line 290) | def llama_varlen_attn_forward(
  function llama_varlen_attn_forward_legacy (line 399) | def llama_varlen_attn_forward_legacy(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/mistral.py
  class MistralRotaryEmbedding (line 32) | class MistralRotaryEmbedding(nn.Module):
    method __init__ (line 34) | def __init__(self,
    method _set_cos_sin_cache (line 53) | def _set_cos_sin_cache(self, seq_len, device, dtype):
    method forward (line 64) | def forward(self, x, seq_len=None):
  function repeat_kv_bshd (line 78) | def repeat_kv_bshd(hidden_states: torch.Tensor, n_rep: int) -> torch.Ten...
  function mistral_attn_forward (line 92) | def mistral_attn_forward(
  function mistral_varlen_attn_forward (line 241) | def mistral_varlen_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/phi3.py
  class Cache (line 18) | class Cache:
  function repeat_kv (line 40) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  function rotate_half (line 59) | def rotate_half(x):
  function apply_rotary_pos_emb (line 67) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_di...
  function phi3_attn_forward (line 94) | def phi3_attn_forward(
  function phi3_varlen_attn_forward (line 261) | def phi3_varlen_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/qwen2.py
  function qwen2_attn_forward (line 30) | def qwen2_attn_forward(
  function qwen2_varlen_attn_forward (line 177) | def qwen2_varlen_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/triton_kernels/layer_norm.py
  function layer_norm_forward (line 6) | def layer_norm_forward(self, hidden_states):

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/triton_kernels/rms_norm.py
  function _rms_norm_fwd_fused (line 8) | def _rms_norm_fwd_fused(
  function _rms_norm_bwd_dx_fused (line 45) | def _rms_norm_bwd_dx_fused(
  function _rms_norm_bwd_dwdb (line 100) | def _rms_norm_bwd_dwdb(
  class RMSNorm (line 122) | class RMSNorm(torch.autograd.Function):
    method forward (line 125) | def forward(ctx, x, weight, eps):
    method backward (line 159) | def backward(ctx, dy):
  function rms_norm_forward (line 214) | def rms_norm_forward(self, hidden_states):

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/triton_kernels/rotary.py
  function rotary_kernel (line 11) | def rotary_kernel(
  function apply_rotary (line 163) | def apply_rotary(
  class ApplyRotaryEmb (line 265) | class ApplyRotaryEmb(torch.autograd.Function):
    method forward (line 268) | def forward(
    method backward (line 302) | def backward(ctx, do):

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/utils.py
  function _get_unpad_data (line 10) | def _get_unpad_data(attention_mask):
  function upad_qkv (line 23) | def upad_qkv(query_layer, key_layer, value_layer, attention_mask,

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/dispatch/yi.py
  function rotate_half (line 8) | def rotate_half(x):
  function apply_rotary_pos_emb (line 15) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids):
  function repeat_kv (line 27) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  function yi_attn_forward (line 45) | def yi_attn_forward(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/projector/configuration_projector.py
  class ProjectorConfig (line 5) | class ProjectorConfig(PretrainedConfig):
    method __init__ (line 9) | def __init__(

FILE: xtuner-train_internvideo2_5/xtuner/model/modules/projector/modeling_projector.py
  class ProjectorModel (line 10) | class ProjectorModel(PreTrainedModel):
    method __init__ (line 16) | def __init__(self, config: ProjectorConfig) -> None:
    method enable_input_require_grads (line 35) | def enable_input_require_grads(self):
    method _set_gradient_checkpointing (line 42) | def _set_gradient_checkpointing(self, module, value=False):
    method forward (line 46) | def forward(self, x):

FILE: xtuner-train_internvideo2_5/xtuner/model/orpo.py
  class ORPO (line 17) | class ORPO(SupervisedFinetune):
    method __init__ (line 25) | def __init__(self, *args, beta=0.1, **kwargs):
    method _gather_masked_logits (line 29) | def _gather_masked_logits(self, logits, labels, mask):
    method get_logps (line 35) | def get_logps(
    method get_var_len_atten_logps (line 56) | def get_var_len_atten_logps(self, all_logits, average_log_prob, labels,
    method cross_entropy_loss (line 94) | def cross_entropy_loss(self, logits, labels):
    method odds_ratio_loss (line 106) | def odds_ratio_loss(
    method _split_for_sequence_parallel (line 127) | def _split_for_sequence_parallel(data):
    method compute_loss (line 139) | def compute_loss(self, data, data_samples=None):

FILE: xtuner-train_internvideo2_5/xtuner/model/reward.py
  function reduce_mean (line 35) | def reduce_mean(tensor):
  function smart_tokenizer_and_embedding_resize (line 44) | def smart_tokenizer_and_embedding_resize(
  class RewardModel (line 87) | class RewardModel(BaseModel):
    method __init__ (line 89) | def __init__(
    method gradient_checkpointing_enable (line 157) | def gradient_checkpointing_enable(self):
    method activation_checkpointing_enable (line 160) | def activation_checkpointing_enable(self):
    method gradient_checkpointing_disable (line 163) | def gradient_checkpointing_disable(self):
    method activation_checkpointing_disable (line 166) | def activation_checkpointing_disable(self):
    method _prepare_for_lora (line 169) | def _prepare_for_lora(self,
    method init_weights (line 182) | def init_weights(self):
    method _prepare_for_long_context_training (line 186) | def _prepare_for_long_context_training(cfg, llm_cfg,
    method _prepare_for_flash_attn (line 211) | def _prepare_for_flash_attn(cfg, llm_cfg):
    method _prepare_for_qlora_zero3 (line 240) | def _prepare_for_qlora_zero3(cfg):
    method _dispatch_lm_model_cfg (line 256) | def _dispatch_lm_model_cfg(self, cfg, max_position_embeddings=None):
    method _build_from_cfg_or_module (line 267) | def _build_from_cfg_or_module(self, cfg_or_mod):
    method forward (line 276) | def forward(self, data, data_samples=None, mode='loss'):
    method _forward (line 287) | def _forward(self, data, data_samples=None):
    method predict (line 292) | def predict(self, data, data_samples=None):
    method _split_for_sequence_parallel (line 299) | def _split_for_sequence_parallel(data):
    method compute_loss (line 311) | def compute_loss(self, data, labels=None):
    method ranking_loss (line 381) | def ranking_loss(self, chosen_logits, rejected_logits, avg_factor):
    method focal_loss (line 385) | def focal_loss(self, chosen_logits, rejected_logits, avg_factor):
    method log_barrier_penalty (line 394) | def log_barrier_penalty(self,
    method l2_penalty (line 408) | def l2_penalty(self, logits, avg_factor=1):
    method state_dict (line 411) | def state_dict(self, *args, **kwargs):
    method __getattr__ (line 418) | def __getattr__(self, name: str):
    method to_hf (line 424) | def to_hf(self,

FILE: xtuner-train_internvideo2_5/xtuner/model/sft.py
  function smart_tokenizer_and_embedding_resize (line 28) | def smart_tokenizer_and_embedding_resize(
  class SupervisedFinetune (line 71) | class SupervisedFinetune(BaseModel):
    method __init__ (line 73) | def __init__(self,
    method gradient_checkpointing_enable (line 122) | def gradient_checkpointing_enable(self):
    method activation_checkpointing_enable (line 125) | def activation_checkpointing_enable(self):
    method gradient_checkpointing_disable (line 128) | def gradient_checkpointing_disable(self):
    method activation_checkpointing_disable (line 131) | def activation_checkpointing_disable(self):
    method _prepare_for_lora (line 134) | def _prepare_for_lora(self,
    method init_weights (line 147) | def init_weights(self):
    method _prepare_for_long_context_training (line 151) | def _prepare_for_long_context_training(cfg, llm_cfg,
    method _prepare_for_flash_attn (line 176) | def _prepare_for_flash_attn(cfg, llm_cfg):
    method _prepare_for_qlora_zero3 (line 206) | def _prepare_for_qlora_zero3(cfg):
    method _dispatch_lm_model_cfg (line 222) | def _dispatch_lm_model_cfg(self, cfg, max_position_embeddings=None):
    method _build_from_cfg_or_module (line 233) | def _build_from_cfg_or_module(self, cfg_or_mod):
    method forward (line 242) | def forward(self, data, data_samples=None, mode='loss'):
    method _forward (line 253) | def _forward(self, data, data_samples=None):
    method predict (line 259) | def predict(self, data, data_samples=None):
    method _split_for_sequence_parallel (line 265) | def _split_for_sequence_parallel(data):
    method _compute_sequence_parallel_loss (line 277) | def _compute_sequence_parallel_loss(self, data):
    method compute_loss (line 287) | def compute_loss(self, data, data_samples=None):
    method state_dict (line 295) | def state_dict(self, *args, **kwargs):
    method __getattr__ (line 302) | def __getattr__(self, name: str):
    method to_hf (line 308) | def to_hf(self,

FILE: xtuner-train_internvideo2_5/xtuner/model/transformers_models/deepseek_v2/configuration_deepseek.py
  class DeepseekV2Config (line 11) | class DeepseekV2Config(PretrainedConfig):
    method __init__ (line 124) | def __init__(

FILE: xtuner-train_internvideo2_5/xtuner/model/transformers_models/deepseek_v2/modeling_deepseek.py
  function _get_unpad_data (line 76) | def _get_unpad_data(attention_mask):
  class DeepseekV2RMSNorm (line 89) | class DeepseekV2RMSNorm(nn.Module):
    method __init__ (line 91) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 97) | def forward(self, hidden_states):
  class DeepseekV2RotaryEmbedding (line 109) | class DeepseekV2RotaryEmbedding(nn.Module):
    method __init__ (line 111) | def __init__(self,
    method _set_cos_sin_cache (line 134) | def _set_cos_sin_cache(self, seq_len, device, dtype):
    method forward (line 147) | def forward(self, x, seq_len=None):
  class DeepseekV2LinearScalingRotaryEmbedding (line 160) | class DeepseekV2LinearScalingRotaryEmbedding(DeepseekV2RotaryEmbedding):
    method __init__ (line 166) | def __init__(
    method _set_cos_sin_cache (line 177) | def _set_cos_sin_cache(self, seq_len, device, dtype):
  class DeepseekV2DynamicNTKScalingRotaryEmbedding (line 193) | class DeepseekV2DynamicNTKScalingRotaryEmbedding(DeepseekV2RotaryEmbeddi...
    method __init__ (line 199) | def __init__(
    method _set_cos_sin_cache (line 210) | def _set_cos_sin_cache(self, seq_len, device, dtype):
  function yarn_find_correction_dim (line 236) | def yarn_find_correction_dim(num_rotations,
  function yarn_find_correction_range (line 246) | def yarn_find_correction_range(low_rot,
  function yarn_get_mscale (line 258) | def yarn_get_mscale(scale=1, mscale=1):
  function yarn_linear_ramp_mask (line 264) | def yarn_linear_ramp_mask(min, max, dim):
  class DeepseekV2YarnRotaryEmbedding (line 273) | class DeepseekV2YarnRotaryEmbedding(DeepseekV2RotaryEmbedding):
    method __init__ (line 275) | def __init__(
    method _set_cos_sin_cache (line 296) | def _set_cos_sin_cache(self, seq_len, device, dtype):
  function rotate_half (line 336) | def rotate_half(x):
  function apply_rotary_pos_emb (line 344) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
  class DeepseekV2MLP (line 379) | class DeepseekV2MLP(nn.Module):
    method __init__ (line 381) | def __init__(self, config, hidden_size=None, intermediate_size=None):
    method forward (line 397) | def forward(self, x):
  class MoEGate (line 403) | class MoEGate(nn.Module):
    method __init__ (line 405) | def __init__(self, config):
    method reset_parameters (line 425) | def reset_parameters(self) -> None:
    method forward (line 430) | def forward(self, hidden_states):
  class AddAuxiliaryLoss (line 502) | class AddAuxiliaryLoss(torch.autograd.Function):
    method forward (line 507) | def forward(ctx, x, loss):
    method backward (line 514) | def backward(ctx, grad_output):
  class ExpertShard (line 522) | class ExpertShard(nn.Module):
    method __init__ (line 524) | def __init__(self, config, shard_idx, expert_in_one_shard=10):
    method reset_parameters (line 539) | def reset_parameters(self) -> None:
    method expert_forward (line 545) | def expert_forward(self, current_state, expert_idx):
    method forward (line 555) | def forward(self, hidden_states, flat_topk_idx, y):
  class DeepseekV2MoEShard (line 563) | class DeepseekV2MoEShard(nn.Module):
    method __init__ (line 566) | def __init__(self, config):
    method forward (line 597) | def forward(self, hidden_states):
  class DeepseekV2MoE (line 623) | class DeepseekV2MoE(nn.Module):
    method __init__ (line 626) | def __init__(self, config):
    method forward (line 658) | def forward(self, hidden_states):
    method moe_infer (line 684) | def moe_infer(self, x, topk_ids, topk_weight):
  function repeat_kv (line 754) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class DeepseekV2Attention (line 773) | class DeepseekV2Attention(nn.Module):
    method __init__ (line 776) | def __init__(self,
    method _init_rope (line 845) | def _init_rope(self):
    method _shape (line 890) | def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
    method forward (line 894) | def forward(
  class DeepseekV2FlashAttention2 (line 999) | class DeepseekV2FlashAttention2(DeepseekV2Attention):
    method __init__ (line 1008) | def __init__(self, *args, **kwargs):
    method forward (line 1017) | def forward(
    method _flash_attention_forward (line 1146) | def _flash_attention_forward(
    method _upad_input (line 1224) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class DeepseekV2DecoderLayer (line 1278) | class DeepseekV2DecoderLayer(nn.Module):
    method __init__ (line 1280) | def __init__(self, config: DeepseekV2Config, layer_idx: int):
    method forward (line 1305) | def forward(
  function _load_pretrained_model (line 1367) | def _load_pretrained_model(
  class DeepseekV2PreTrainedModel (line 1421) | class DeepseekV2PreTrainedModel(PreTrainedModel):
    method _init_weights (line 1431) | def _init_weights(self, module):
    method from_pretrained (line 1443) | def from_pretrained(cls, pretrained_model_name_or_path, *args, **kwargs):
  class DeepseekV2Model (line 1529) | class DeepseekV2Model(DeepseekV2PreTrainedModel):
    method __init__ (line 1537) | def __init__(self, config: DeepseekV2Config):
    method get_input_embeddings (line 1557) | def get_input_embeddings(self):
    method set_input_embeddings (line 1560) | def set_input_embeddings(self, value):
    method forward (line 1564) | def forward(
  class DeepseekV2ForCausalLM (line 1718) | class DeepseekV2ForCausalLM(DeepseekV2PreTrainedModel):
    method __init__ (line 1721) | def __init__(self, config):
    method get_input_embeddings (line 1731) | def get_input_embeddings(self):
    method set_input_embeddings (line 1734) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1737) | def get_output_embeddings(self):
    method set_output_embeddings (line 1740) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1743) | def set_decoder(self, decoder):
    method get_decoder (line 1746) | def get_decoder(self):
    method forward (line 1752) | def forward(
    method prepare_inputs_for_generation (line 1842) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1901) | def _reorder_cache(past_key_values, beam_idx):
  class DeepseekV2ForSequenceClassification (line 1925) | class DeepseekV2ForSequenceClassification(DeepseekV2PreTrainedModel):
    method __init__ (line 1927) | def __init__(self, config):
    method get_input_embeddings (line 1936) | def get_input_embeddings(self):
    method set_input_embeddings (line 1939) | def set_input_embeddings(self, value):
    method forward (line 1943) | def forward(

FILE: xtuner-train_internvideo2_5/xtuner/model/transformers_models/deepseek_v2/tokenization_deepseek_fast.py
  class DeepseekTokenizerFast (line 6) | class DeepseekTokenizerFast(LlamaTokenizerFast):
    method convert_ids_to_tokens (line 8) | def convert_ids_to_tokens(
    method _convert_id_to_token (line 35) | def _convert_id_to_token(self, index: int) -> Optional[str]:

FILE: xtuner-train_internvideo2_5/xtuner/model/transformers_models/mixtral/configuration_mixtral.py
  class MixtralConfig (line 22) | class MixtralConfig(PretrainedConfig):
    method __init__ (line 114) | def __init__(

FILE: xtuner-train_internvideo2_5/xtuner/model/transformers_models/mixtral/modeling_mixtral.py
  function load_balancing_loss_func (line 56) | def load_balancing_loss_func(
  function _get_unpad_data (line 138) | def _get_unpad_data(attention_mask):
  class MixtralRMSNorm (line 152) | class MixtralRMSNorm(nn.Module):
    method __init__ (line 154) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 160) | def forward(self, hidden_states):
  class MixtralRotaryEmbedding (line 170) | class MixtralRotaryEmbedding(nn.Module):
    method __init__ (line 172) | def __init__(self,
    method _set_cos_sin_cache (line 194) | def _set_cos_sin_cache(self, seq_len, device, dtype):
    method forward (line 208) | def forward(self, x, seq_len=None):
  function rotate_half (line 221) | def rotate_half(x):
  function apply_rotary_pos_emb (line 229) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
  function repeat_kv (line 258) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class MixtralAttention (line 277) | class MixtralAttention(nn.Module):
    method __init__ (line 284) | def __init__(self, config: MixtralConfig, layer_idx: Optional[int] = N...
    method _shape (line 327) | def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
    method forward (line 331) | def forward(
  class MixtralFlashAttention2 (line 416) | class MixtralFlashAttention2(MixtralAttention):
    method __init__ (line 426) | def __init__(self, *args, **kwargs):
    method forward (line 435) | def forward(
    method _flash_attention_forward (line 567) | def _flash_attention_forward(
    method _upad_input (line 670) | def _upad_input(self, query_layer, key_layer, value_layer, attention_m...
  class MixtralSdpaAttention (line 722) | class MixtralSdpaAttention(MixtralAttention):
    method forward (line 732) | def forward(
  class MixtralBlockSparseTop2MLP (line 825) | class MixtralBlockSparseTop2MLP(nn.Module):
    method __init__ (line 827) | def __init__(self, config: MixtralConfig):
    method forward (line 838) | def forward(self, hidden_states):
  class MixtralSparseMoeBlock (line 845) | class MixtralSparseMoeBlock(nn.Module):
    method __init__ (line 856) | def __init__(self, config):
    method forward (line 873) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
  class ExpertShard (line 921) | class ExpertShard(nn.Module):
    method __init__ (line 923) | def __init__(self, config, expert_in_one_shard=1):
    method forward (line 934) | def forward(self, hidden_states, expert_mask, routing_weights,
  class MixtralSparseShardMoeBlock (line 955) | class MixtralSparseShardMoeBlock(nn.Module):
    method __init__ (line 957) | def __init__(self, config):
    method forward (line 981) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
  class MixtralDecoderLayer (line 1021) | class MixtralDecoderLayer(nn.Module):
    method __init__ (line 1023) | def __init__(self, config: MixtralConfig, layer_idx: int):
    method forward (line 1044) | def forward(
  function _load_pretrained_model (line 1107) | def _load_pretrained_model(
  class MixtralPreTrainedModel (line 1162) | class MixtralPreTrainedModel(PreTrainedModel):
    method _init_weights (line 1172) | def _init_weights(self, module):
    method from_pretrained (line 1184) | def from_pretrained(cls, pretrained_model_name_or_path, *args, **kwargs):
  class MixtralModel (line 1268) | class MixtralModel(MixtralPreTrainedModel):
    method __init__ (line 1276) | def __init__(self, config: MixtralConfig):
    method get_input_embeddings (line 1294) | def get_input_embeddings(self):
    method set_input_embeddings (line 1297) | def set_input_embeddings(self, value):
    method forward (line 1302) | def forward(
  class MixtralForCausalLM (line 1475) | class MixtralForCausalLM(MixtralPreTrainedModel):
    method __init__ (line 1478) | def __init__(self, config):
    method get_input_embeddings (line 1490) | def get_input_embeddings(self):
    method set_input_embeddings (line 1493) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1496) | def get_output_embeddings(self):
    method set_output_embeddings (line 1499) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1502) | def set_decoder(self, decoder):
    method get_decoder (line 1505) | def get_decoder(self):
    method forward (line 1512) | def forward(
    method prepare_inputs_for_generation (line 1621) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1683) | def _reorder_cache(past_key_values, beam_idx):
  class MixtralForSequenceClassification (line 1708) | class MixtralForSequenceClassification(MixtralPreTrainedModel):
    method __init__ (line 1710) | def __init__(self, config):
    method get_input_embeddings (line 1719) | def get_input_embeddings(self):
    method set_input_embeddings (line 1722) | def set_input_embeddings(self, value):
    method forward (line 1726) | def forward(

FILE: xtuner-train_internvideo2_5/xtuner/model/utils.py
  function set_obj_dtype (line 14) | def set_obj_dtype(d):
  function try_build_module (line 20) | def try_build_module(cfg):
  function traverse_dict (line 33) | def traverse_dict(d):
  function find_all_linear_names (line 47) | def find_all_linear_names(model):
  class LoadWoInit (line 61) | class LoadWoInit:
    method __init__ (line 64) | def __init__(self):
    method __enter__ (line 73) | def __enter__(self, *args, **kwargs):
    method __exit__ (line 82) | def __exit__(self, *args, **kwargs):
  function get_peft_model_state_dict (line 92) | def get_peft_model_state_dict(model, state_dict=None, adapter_name='defa...
  function prepare_inputs_labels_for_multimodal (line 138) | def prepare_inputs_labels_for_multimodal(
  function make_inputs_require_grad (line 293) | def make_inputs_require_grad(module, input, output):
  function guess_load_checkpoint (line 297) | def guess_load_checkpoint(pth_model):

FILE: xtuner-train_internvideo2_5/xtuner/parallel/sequence/attention.py
  function pre_process_for_sequence_parallel_attn (line 16) | def pre_process_for_sequence_parallel_attn(query_states,
  function post_process_for_sequence_parallel_attn (line 96) | def post_process_for_sequence_parallel_attn(attn_output,
  function sequence_parallel_wrapper (line 130) | def sequence_parallel_wrapper(local_attn):

FILE: xtuner-train_internvideo2_5/xtuner/parallel/sequence/comm.py
  function _all_to_all (line 9) | def _all_to_all(
  class _AllToAll (line 25) | class _AllToAll(torch.autograd.Function):
    method forward (line 36) | def forward(ctx: Any, input: Tensor, sp_group: dist.ProcessGroup,
    method backward (line 47) | def backward(ctx: Any, grad_output: Tensor) -> Tuple:
  function all_to_all (line 63) | def all_to_all(
  function split_for_sequence_parallel (line 90) | def split_for_sequence_parallel(input, dim: int, sp_group: dist.ProcessG...
  function gather_for_sequence_parallel (line 117) | def gather_for_sequence_parallel(input, dim: int, sp_group: dist.Process...
  class _GatherForwardSplitBackward (line 144) | class _GatherForwardSplitBackward(torch.autograd.Function):
    method forward (line 152) | def forward(ctx, input, dim, sp_group, grad_scale):
    method backward (line 159) | def backward(ctx, grad_output):
  class _SplitForwardGatherBackward (line 169) | class _SplitForwardGatherBackward(torch.autograd.Function):
    method forward (line 177) | def forward(ctx, input, dim, sp_group, grad_scale):
    method backward (line 184) | def backward(ctx, grad_output):
  function split_forward_gather_backward (line 193) | def split_forward_gather_backward(input, dim, sp_group, grad_scale=None):
  function gather_forward_split_backward (line 207) | def gather_forward_split_backward(input, dim, sp_group, grad_scale=None):

FILE: xtuner-train_internvideo2_5/xtuner/parallel/sequence/data_collate.py
  function pad_for_sequence_parallel (line 7) | def pad_for_sequence_parallel(tensor, padding_value, dim=-1):
  function pad_cumulative_len_for_sequence_parallel (line 26) | def pad_cumulative_len_for_sequence_parallel(cumulative_len):

FILE: xtuner-train_internvideo2_5/xtuner/parallel/sequence/reduce_loss.py
  class _ReduceLoss (line 7) | class _ReduceLoss(torch.autograd.Function):
    method forward (line 10) | def forward(ctx, mean_loss, loss_scale, process_group):
    method backward (line 22) | def backward(ctx, grad_output):
  function reduce_sequence_parallel_loss (line 26) | def reduce_sequence_parallel_loss(mean_loss,

FILE: xtuner-train_internvideo2_5/xtuner/parallel/sequence/sampler.py
  class SequenceParallelSampler (line 12) | class SequenceParallelSampler(DefaultSampler):
    method __init__ (line 14) | def __init__(self,

FILE: xtuner-train_internvideo2_5/xtuner/parallel/sequence/setup_distributed.py
  function init_sequence_parallel (line 17) | def init_sequence_parallel(sequence_parallel_size: int = 1):
  function init_inner_sequence_parallel (line 56) | def init_inner_sequence_parallel(inner_sequence_parallel_size: int = 1):
  function is_inner_sequence_parallel_initialized (line 82) | def is_inner_sequence_parallel_initialized():
  function get_inner_sequence_parallel_group (line 86) | def get_inner_sequence_parallel_group():
  function get_inner_sequence_parallel_world_size (line 90) | def get_inner_sequence_parallel_world_size():
  function get_inner_sequence_parallel_rank (line 102) | def get_inner_sequence_parallel_rank():
  function get_sequence_parallel_group (line 114) | def get_sequence_parallel_group():
  function get_sequence_parallel_world_size (line 119) | def get_sequence_parallel_world_size():
  function get_sequence_parallel_rank (line 132) | def get_sequence_parallel_rank():
  function get_data_parallel_group (line 145) | def get_data_parallel_group():
  function get_data_parallel_world_size (line 152) | def get_data_parallel_world_size():
  function get_data_parallel_rank (line 165) | def get_data_parallel_rank():

FILE: xtuner-train_internvideo2_5/xtuner/tools/chat.py
  function remove_prefix (line 26) | def remove_prefix(state_dict, prefix):
  function parse_args (line 37) | def parse_args():
  function get_input (line 131) | def get_input():
  function main (line 146) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/check_custom_dataset.py
  function parse_args (line 13) | def parse_args():
  function is_standard_format (line 22) | def is_standard_format(dataset):
  function main (line 40) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/copy_cfg.py
  function parse_args (line 11) | def parse_args():
  function add_copy_suffix (line 19) | def add_copy_suffix(string):
  function main (line 24) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/data_preprocess/arxiv.py
  function parse_args (line 7) | def parse_args():
  function has_intersection (line 25) | def has_intersection(list1, list2):
  function read_json_file (line 31) | def read_json_file(file_path):
  function main (line 43) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/data_preprocess/convert_refcoco.py
  function parse_args (line 8) | def parse_args():

FILE: xtuner-train_internvideo2_5/xtuner/tools/eval_refcoco.py
  function merge_outputs (line 31) | def merge_outputs(otuputs):
  function master_print (line 44) | def master_print(msg):
  function parse_args (line 48) | def parse_args():
  function eval_iou (line 110) | def eval_iou(answers):
  function build_model (line 143) | def build_model(args):
  function generate (line 233) | def generate(
  function main (line 299) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/get_data_order.py
  function parse_args (line 6) | def parse_args():
  function save_data_order (line 18) | def save_data_order(data_folder, save_folder, file_type='.bin'):

FILE: xtuner-train_internvideo2_5/xtuner/tools/list_cfg.py
  function parse_args (line 7) | def parse_args():
  function main (line 15) | def main(pattern=None):

FILE: xtuner-train_internvideo2_5/xtuner/tools/list_dataset_format.py
  function main (line 5) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/log_dataset.py
  function parse_args (line 9) | def parse_args():
  function main (line 22) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/mmbench.py
  function parse_args (line 38) | def parse_args():
  function master_print (line 95) | def master_print(msg):
  class MMBenchDataset (line 99) | class MMBenchDataset(Dataset):
    method __init__ (line 116) | def __init__(self, data_file):
    method get_image (line 122) | def get_image(self, image):
    method __len__ (line 130) | def __len__(self):
    method __getitem__ (line 133) | def __getitem__(self, idx):
    method load_from_df (line 166) | def load_from_df(self, idx, key):
    method eval_result (line 173) | def eval_result(self, result_df, show=True):
  function main (line 271) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/model_converters/merge.py
  function parse_args (line 12) | def parse_args():
  function main (line 43) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/model_converters/modeling_internlm2_reward/configuration_internlm2.py
  class InternLM2Config (line 28) | class InternLM2Config(PretrainedConfig):
    method __init__ (line 78) | def __init__(  # pylint: disable=W0102
    method _rope_scaling_validation (line 135) | def _rope_scaling_validation(self):

FILE: xtuner-train_internvideo2_5/xtuner/tools/model_converters/modeling_internlm2_reward/modeling_internlm2.py
  function _import_flash_attn (line 56) | def _import_flash_attn():
  function _get_unpad_data (line 68) | def _get_unpad_data(attention_mask):
  function _make_causal_mask (line 81) | def _make_causal_mask(
  function _expand_mask (line 99) | def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Option...
  class InternLM2RMSNorm (line 114) | class InternLM2RMSNorm(nn.Module):
    method __init__ (line 115) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 123) | def forward(self, hidden_states):
  class InternLM2RotaryEmbedding (line 132) | class InternLM2RotaryEmbedding(nn.Module):
    method __init__ (line 133) | def __init__(self, dim, max_position_embeddings=2048, base=10000, devi...
    method _set_cos_sin_cache (line 147) | def _set_cos_sin_cache(self, seq_len, device, dtype):
    method forward (line 157) | def forward(self, x, seq_len=None):
  class InternLM2LinearScalingRotaryEmbedding (line 169) | class InternLM2LinearScalingRotaryEmbedding(InternLM2RotaryEmbedding):
    method __init__ (line 172) | def __init__(self, dim, max_position_embeddings=2048, base=10000, devi...
    method _set_cos_sin_cache (line 176) | def _set_cos_sin_cache(self, seq_len, device, dtype):
  class InternLM2DynamicNTKScalingRotaryEmbedding (line 189) | class InternLM2DynamicNTKScalingRotaryEmbedding(InternLM2RotaryEmbedding):
    method __init__ (line 194) | def __init__(self, dim, max_position_embeddings=2048, base=10000, devi...
    method _set_cos_sin_cache (line 198) | def _set_cos_sin_cache(self, seq_len, device, dtype):
  function rotate_half (line 218) | def rotate_half(x):
  function apply_rotary_pos_emb (line 226) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids, unsqueeze_dim=1):
  class InternLM2MLP (line 235) | class InternLM2MLP(nn.Module):
    method __init__ (line 236) | def __init__(self, config):
    method forward (line 246) | def forward(self, x):
  function repeat_kv (line 253) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  class InternLM2Attention (line 266) | class InternLM2Attention(nn.Module):
    method __init__ (line 269) | def __init__(self, config: InternLM2Config):
    method _init_rope (line 295) | def _init_rope(self):
    method _shape (line 323) | def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
    method forward (line 326) | def forward(
  class InternLM2FlashAttention2 (line 415) | class InternLM2FlashAttention2(InternLM2Attention):
    method forward (line 422) | def forward(
    method _flash_attention_forward (line 494) | def _flash_attention_forward(
    method _unpad_input (line 548) | def _unpad_input(self, query_layer, key_layer, value_layer, attention_...
  class InternLM2DecoderLayer (line 593) | class InternLM2DecoderLayer(nn.Module):
    method __init__ (line 594) | def __init__(self, config: InternLM2Config):
    method forward (line 604) | def forward(
  class InternLM2PreTrainedModel (line 689) | class InternLM2PreTrainedModel(PreTrainedModel):
    method _init_weights (line 696) | def _init_weights(self, module):
  class InternLM2Model (line 778) | class InternLM2Model(InternLM2PreTrainedModel):
    method __init__ (line 788) | def __init__(self, config: InternLM2Config):
    method get_input_embeddings (line 803) | def get_input_embeddings(self):
    method set_input_embeddings (line 806) | def set_input_embeddings(self, value):
    method _prepare_decoder_attention_mask (line 809) | def _prepare_decoder_attention_mask(self, attention_mask, input_shape,...
    method forward (line 833) | def forward(
  class InternLM2ForCausalLM (line 967) | class InternLM2ForCausalLM(InternLM2PreTrainedModel):
    method __init__ (line 972) | def __init__(self, config):
    method get_input_embeddings (line 981) | def get_input_embeddings(self):
    method set_input_embeddings (line 984) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 987) | def get_output_embeddings(self):
    method set_output_embeddings (line 990) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 993) | def set_decoder(self, decoder):
    method get_decoder (line 996) | def get_decoder(self):
    method forward (line 1001) | def forward(
    method prepare_inputs_for_generation (line 1088) | def prepare_inputs_for_generation(
    method _reorder_cache (line 1128) | def _reorder_cache(past_key_values, beam_idx):
    method build_inputs (line 1136) | def build_inputs(self, tokenizer, query: str, history: List[Tuple[str,...
    method chat (line 1149) | def chat(
    method stream_chat (line 1185) | def stream_chat(
  class InternLM2ForRewardModel (line 1271) | class InternLM2ForRewardModel(InternLM2PreTrainedModel):
    method __init__ (line 1276) | def __init__(self, config):
    method get_input_embeddings (line 1286) | def get_input_embeddings(self):
    method set_input_embeddings (line 1289) | def set_input_embeddings(self, value):
    method get_output_embeddings (line 1292) | def get_output_embeddings(self):
    method set_output_embeddings (line 1295) | def set_output_embeddings(self, new_embeddings):
    method set_decoder (line 1298) | def set_decoder(self, decoder):
    method get_decoder (line 1301) | def get_decoder(self):
    method forward (line 1306) | def forward(
    method get_score (line 1386) | def get_score(
    method get_scores (line 1403) | def get_scores(
    method compare (line 1428) | def compare(
    method rank (line 1444) | def rank(
  class InternLM2ForSequenceClassification (line 1474) | class InternLM2ForSequenceClassification(InternLM2PreTrainedModel):
    method __init__ (line 1475) | def __init__(self, config):
    method get_input_embeddings (line 1484) | def get_input_embeddings(self):
    method set_input_embeddings (line 1487) | def set_input_embeddings(self, value):
    method forward (line 1491) | def forward(

FILE: xtuner-train_internvideo2_5/xtuner/tools/model_converters/pth_to_hf.py
  function parse_args (line 20) | def parse_args():
  function main (line 60) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/model_converters/split.py
  function parse_args (line 13) | def parse_args():
  function main (line 22) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/plugins/api.py
  function plugins_api (line 5) | def plugins_api(input_str,

FILE: xtuner-train_internvideo2_5/xtuner/tools/plugins/calculate.py
  function Calculate (line 5) | def Calculate(expression):

FILE: xtuner-train_internvideo2_5/xtuner/tools/plugins/search.py
  function parse_results (line 15) | def parse_results(results, k=10):
  function search (line 26) | def search(api_key, search_term, **kwargs):
  function Search (line 47) | def Search(q, k=10):

FILE: xtuner-train_internvideo2_5/xtuner/tools/plugins/solve.py
  function Solve (line 11) | def Solve(equations_str):

FILE: xtuner-train_internvideo2_5/xtuner/tools/process_untokenized_datasets.py
  function parse_args (line 15) | def parse_args():
  function modify_config (line 23) | def modify_config(config, dataset_save_folder):
  function process_untokenized_dataset (line 45) | def process_untokenized_dataset(config):

FILE: xtuner-train_internvideo2_5/xtuner/tools/process_untokenized_datasets_legacy.py
  function parse_args (line 42) | def parse_args():
  function process_one (line 78) | def process_one(fp,
  function process_untokenized_dataset (line 109) | def process_untokenized_dataset(folder,

FILE: xtuner-train_internvideo2_5/xtuner/tools/process_untokenized_llava_data.py
  function parse_args (line 13) | def parse_args():
  function build_llava_dataset (line 21) | def build_llava_dataset(config):

FILE: xtuner-train_internvideo2_5/xtuner/tools/test.py
  function parse_args (line 16) | def parse_args():
  function register_function (line 45) | def register_function(cfg_dict):
  function main (line 60) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/tokenize_ftdp_datasets.py
  function chatml_format (line 95) | def chatml_format(
  function write_bin_meta_bin (line 271) | def write_bin_meta_bin(path, dataset_name, filename, samples):
  function tokenize_and_save (line 320) | def tokenize_and_save(tokenizer, processed_dir, tokenized_dir):
  function tokenizer_add_special_tokens (line 389) | def tokenizer_add_special_tokens(tokenizer):
  function save_new_tokenizer (line 397) | def save_new_tokenizer(tokenizer, save_dir):
  function parse_args (line 402) | def parse_args():
  function main (line 419) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/train.py
  function parse_args (line 29) | def parse_args():
  function register_function (line 65) | def register_function(cfg_dict):
  function check_cfg (line 80) | def check_cfg(cfg, args):
  function main (line 127) | def main():

FILE: xtuner-train_internvideo2_5/xtuner/tools/utils.py
  function get_base_model (line 13) | def get_base_model(model):
  function get_streamer (line 21) | def get_streamer(model):
  class DecodeOutputStreamer (line 41) | class DecodeOutputStreamer(BaseStreamer):
    method __init__ (line 44) | def __init__(self, tokenizer, skip_prompt=True) -> None:
    method _decode_with_raw_id (line 59) | def _decode_with_raw_id(self, value):
    method _decode_fallback (line 74) | def _decode_fallback(self, value):
    method put (line 83) | def put(self, value):
    method end (line 94) | def end(self):
  class NoSpaceStreamer (line 100) | class NoSpaceStreamer(DecodeOutputStreamer):
    method __init__ (line 102) | def __init__(self, tokenizer, skip_prompt=True) -> None:
    method decode (line 113) | def decode(self, value):
  function get_stop_criteria (line 123) | def get_stop_criteria(
  function auto_dtype_of_deepspeed_config (line 133) | def auto_dtype_of_deepspeed_config(ds_config):
  function is_cn_string (line 152) | def is_cn_string(s):
  function get_seed_from_checkpoint (line 158) | def get_seed_from_checkpoint(pth_model):

FILE: xtuner-train_internvideo2_5/xtuner/utils/fileio.py
  function patch_func (line 8) | def patch_func(module, fn_name_to_wrap):
  function patch_fileio (line 23) | def patch_fileio(global_vars=None):
  function patch_hf_auto_from_pretrained (line 231) | def patch_hf_auto_from_pretrained(petrel_hub):
  function patch_hf_save_pretrained (line 279) | def patch_hf_save_pretrained():
  function patch_deepspeed_engine (line 320) | def patch_deepspeed_engine():

FILE: xtuner-train_internvideo2_5/xtuner/utils/handle_moe_load_and_save.py
  function print_on_rank0 (line 33) | def print_on_rank0(info):
  function get_expert_num_per_shard (line 38) | def get_expert_num_per_shard(model):
  function mix_sort (line 44) | def mix_sort(expert_name):
  function _get_merged_param_name (line 50) | def _get_merged_param_name(origin_param_name, expert_num_per_shard):
  function _merge_experts_weight (line 61) | def _merge_experts_weight(state_dict, expert_num_per_shard, order_mapping):
  function load_state_dict_into_model (line 100) | def load_state_dict_into_model(model_to_load, pretrained_model_path):
  function _get_origin_param_name (line 172) | def _get_origin_param_name(merged_param_name, expert_num_per_shard, is_w...
  function _split_param (line 197) | def _split_param(merged_param, is_w1w3):
  function get_origin_state_dict (line 207) | def get_origin_state_dict(state_dict, model):

FILE: xtuner-train_internvideo2_5/xtuner/utils/stop_criteria.py
  class StopWordStoppingCriteria (line 5) | class StopWordStoppingCriteria(StoppingCriteria):
    method __init__ (line 8) | def __init__(self, tokenizer, stop_word):
    method __call__ (line 13) | def __call__(self, input_ids, *args, **kwargs) -> bool:

FILE: xtuner-train_internvideo2_5/xtuner/utils/zero_to_any_dtype.py
  class zero_model_state (line 43) | class zero_model_state:
  function atoi (line 60) | def atoi(text):
  function natural_keys (line 64) | def natural_keys(text):
  function get_model_state_file (line 71) | def get_model_state_file(checkpoint_dir, zero_stage):
  function get_checkpoint_files (line 88) | def get_checkpoint_files(checkpoint_dir, glob_pattern):
  function get_optim_files (line 102) | def get_optim_files(checkpoint_dir):
  function get_model_state_files (line 106) | def get_model_state_files(checkpoint_dir):
  function parse_model_states (line 110) | def parse_model_states(files, dtype=DEFAULT_DTYPE):
  function parse_optim_states (line 161) | def parse_optim_states(files, ds_checkpoint_dir, dtype=DEFAULT_DTYPE):
  function _get_state_dict_from_zero_checkpoint (line 217) | def _get_state_dict_from_zero_checkpoint(ds_checkpoint_dir,
  function _zero2_merge_frozen_params (line 250) | def _zero2_merge_frozen_params(state_dict, zero_model_states):
  function _has_callable (line 286) | def _has_callable(obj, fn):
  function _zero2_merge_trainable_params (line 291) | def _zero2_merge_trainable_params(state_dict, world_size, flat_groups,
  function _get_state_dict_from_zero2_checkpoint (line 378) | def _get_state_dict_from_zero2_checkpoint(world_size, flat_groups,
  function zero3_partitioned_param_info (line 403) | def zero3_partitioned_param_info(unpartitioned_numel, world_size):
  function _zero3_merge_frozen_params (line 410) | def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
  function _zero3_merge_trainable_params (line 457) | def _zero3_merge_trainable_params(state_dict, world_size, flat_groups,
  function _get_state_dict_from_zero3_checkpoint (line 536) | def _get_state_dict_from_zero3_checkpoint(world_size, flat_groups,
  function get_state_dict_from_zero_checkpoint (line 561) | def get_state_dict_from_zero_checkpoint(checkpoint_dir,
  function convert_zero_checkpoint_to_state_dict (line 622) | def convert_zero_checkpoint_to_state_dict(checkpoint_dir,
  function load_state_dict_from_zero_checkpoint (line 648) | def load_state_dict_from_zero_checkpoint(model,

FILE: xtuner-train_internvideo2_5/xtuner/version.py
  function parse_version_info (line 6) | def parse_version_info(version_str):