MCPcopy Create free account

hub / github.com/PKU-YuanGroup/Helios / functions

Functions847 in github.com/PKU-YuanGroup/Helios

⨍ Functions 847◇ Types & classes 124
↓ 3 callersFunctionapply_rename
(key: str)
tools/others/convert_ckpt.py:52
↓ 3 callersFunctionattn_varlen_func
(q, k, v, attention_mask=None)
helios/modules/helios_kernels/attention_dispatch.py:132
↓ 3 callersMethodavg
(self, name)
eval/utils/third_party/amt/utils/utils.py:51
↓ 3 callersFunctionbackwarp
(tenInput, tenFlow)
eval/utils/third_party/amt/flow_generation/liteflownet/run.py:53
↓ 3 callersFunctioncalculate_settings
( n: int, )
helios/modules/helios_kernels/utils.py:52
↓ 3 callersFunctioncenter_crop_woflow
(img0, imgt, img1, crop_size=(512, 512))
eval/utils/third_party/amt/datasets/gopro_datasets.py:31
↓ 3 callersFunctioncoords_grid
(batch, ht, wd, device)
eval/utils/third_party/amt/networks/blocks/raft.py:27
↓ 3 callersFunctioncreate_ema_final
( accelerator, args, transformer_cpu, model_cls, ds_config, transformer_lora_config,
helios/utils/create_ema_zero3_lora.py:250
↓ 3 callersMethodextract_frame
(self, frame_list, start_from=0)
eval/6_get_drifting_motion_smoothness.py:59
↓ 3 callersMethodextract_frame
(self, frame_list, start_from=0)
eval/2_get_motion_smoothness.py:55
↓ 3 callersFunctionfloor_by_factor
Returns the largest integer less than or equal to 'number' that is divisible by 'factor'.
helios/videoalign/vision_process.py:45
↓ 3 callersMethodfrom_pretrained
( cls, args, path, model_cls, lora_config, transformer_additional_kwargs={} )
helios/utils/create_ema_zero3_lora.py:36
↓ 3 callersMethodget_frequency_batched
(self, freqs_base, pos)
helios/modules/transformer_helios.py:682
↓ 3 callersMethodget_frequency_batched
(self, freqs_base, pos)
helios/diffusers_version/transformer_helios_diffusers.py:340
↓ 3 callersFunctionload_state_dict
(model, state_dict, input_resolution=224, patch_size=16, center=True)
eval/utils/third_party/ViCLIP/viclip_vision.py:231
↓ 3 callersFunctionmulti_flow_combine
A parallel implementation of multiple flow field warping comb_block: An nn.Seqential object. img shape: [b, c, h, w] flow shape: [b,
eval/utils/third_party/amt/networks/blocks/multi_flow.py:12
↓ 3 callersFunctionparse_path
(path)
eval/utils/third_party/amt/benchmarks/snu_film.py:19
↓ 3 callersFunctionprepare_stage1_clean_input_from_latents
( history_latents, # VAE latents, (B, C_latent, F_latent, H_latent, W_latent) target_latents, x0_
helios/utils/utils_helios_base.py:609
↓ 3 callersFunctionprocess_vision_info
( conversations: list[dict] | list[list[dict]], )
helios/videoalign/vision_process.py:358
↓ 3 callersFunctionreadPFM
(file)
eval/utils/third_party/amt/utils/utils.py:139
↓ 3 callersMethodreset
(self)
eval/utils/third_party/amt/utils/utils.py:19
↓ 3 callersMethodreward
videos: List[str] (paths) OR List[torch.Tensor]
helios/videoalign/inference.py:220
↓ 3 callersFunctionsample_dynamic_timestep
( B: int, num_train_timestep: int = 1000, min_timestep: int = 0, max_timestep: int = 1000,
helios/utils/utils_helios_post.py:510
↓ 3 callersFunctionsample_y_error_from_latent_buffer
Specially sample y_error from buffer - can be configured to sample from all grids or custom range.
helios/utils/utils_recycle_single.py:284
↓ 3 callersFunctionsample_y_error_from_latent_buffer
Specially sample y_error from buffer - can be configured to sample from all grids or custom range.
helios/utils/utils_recycle_batch.py:404
↓ 3 callersFunctionsave_extra_components
(args, model=None, model_state_dict=None, output_dir=None)
helios/utils/utils_base.py:157
↓ 3 callersMethodstep
( self, model_output: torch.FloatTensor, timestep: float | torch.FloatTensor = None,
helios/diffusers_version/scheduling_helios_diffusers.py:891
↓ 3 callersMethodtrain
(self)
eval/utils/third_party/amt/trainers/base_trainer.py:128
↓ 2 callersMethod__init__
(self, fmap1, fmap2, num_levels=4, radius=4)
eval/utils/third_party/amt/networks/blocks/raft.py:149
↓ 2 callersMethod__init__
( self, embed_dim: int, context_length: int, vocab_size: int, transfor
eval/utils/third_party/ViCLIP/viclip_text.py:84
↓ 2 callersMethod__init__
(self, width, layers, heads, drop_path=0.0, checkpoint_num=0, dropout=0.0)
eval/utils/third_party/ViCLIP/viclip_vision.py:87
↓ 2 callersMethod__len__
(self)
eval/utils/third_party/amt/datasets/adobe_datasets.py:76
↓ 2 callersMethod_align_sample_counts
(self, is_log=True)
helios/dataset/dataloader_dmd.py:59
↓ 2 callersMethod_clean_message
remove unnecessary keys from message(very very necessary)
helios/videoalign/data.py:151
↓ 2 callersMethod_create_balanced_indices
(self, dataset_groups)
helios/dataset/dataloader_history_latents_dist.py:454
↓ 2 callersMethod_create_balanced_indices
(self, dataset_groups)
helios/dataset/dataloader_mp4_dist.py:609
↓ 2 callersMethod_get_t5_prompt_embeds
( self, prompt: str | list[str] = None, num_videos_per_prompt: int = 1, max_se
helios/diffusers_version/pipeline_helios_diffusers.py:175
↓ 2 callersMethod_get_t5_prompt_embeds
( self, prompt: Union[str, List[str]] = None, num_videos_per_prompt: int = 1,
helios/pipelines/pipeline_helios_ode.py:168
↓ 2 callersMethod_get_t5_prompt_embeds
( self, prompt: Union[str, List[str]] = None, num_videos_per_prompt: int = 1,
helios/pipelines/pipeline_helios.py:169
↓ 2 callersFunction_ode_regression_loss
( args, accelerator, transformer, scheduler, noise, weight_dtype, # For Stage 1
helios/utils/utils_helios_post.py:28
↓ 2 callersMethod_pad_sequence
Pad the sequences to the maximum length.
helios/videoalign/data.py:183
↓ 2 callersMethod_unpad
(self, x)
eval/utils/third_party/amt/utils/utils.py:78
↓ 2 callersFunctionadd_error_to_latent_buffer
Add error sample to buffer using specified replacement strategy based on timestep grid.
helios/utils/utils_recycle_single.py:335
↓ 2 callersFunctionadd_error_to_latent_buffer
Add error sample to buffer using specified replacement strategy based on timestep grid.
helios/utils/utils_recycle_batch.py:467
↓ 2 callersFunctionadd_error_to_y_buffer
Add error sample to buffer using specified replacement strategy based on timestep grid.
helios/utils/utils_recycle_single.py:376
↓ 2 callersFunctionadd_error_to_y_buffer
Add error sample to buffer using specified replacement strategy based on timestep grid.
helios/utils/utils_recycle_batch.py:514
↓ 2 callersFunctionadd_saturation_to_history_latents
( latents_history_short, latents_history_mid, latents_history_long, latent_window_size, is
helios/utils/utils_helios_base.py:469
↓ 2 callersMethodapply_frame_aware_corruption
( self, history_latents: torch.Tensor, corruption_strength: float = 0.1, gener
helios/utils/utils_base.py:731
↓ 2 callersFunctionapply_rotary_emb_transposed
( hidden_states: torch.Tensor, freqs_cis: torch.Tensor, )
helios/diffusers_version/transformer_helios_diffusers.py:53
↓ 2 callersFunctionapply_tiled_linear
Apply tiled MLP computation for memory efficiency. Args: fn: the function to call on sharded inputs (e.g., lambda module, x: module(
helios/modules/helios_kernels/tiled_linear.py:172
↓ 2 callersFunctionbilinear_sampler
Wrapper for grid_sample, uses pixel coordinates
eval/utils/third_party/amt/networks/blocks/raft.py:10
↓ 2 callersFunctionbytes_to_unicode
Returns list of utf-8 byte and a corresponding list of unicode strings. The reversible bpe codes work on unicode strings. This means you
eval/utils/third_party/ViCLIP/simple_tokenizer.py:26
↓ 2 callersFunctioncalculate_shift
( image_seq_len, base_seq_len: int = 256, max_seq_len: int = 4096, base_shift: float = 0.5,
helios/diffusers_version/pipeline_helios_diffusers.py:111
↓ 2 callersFunctioncenter_down_sample_3d
(x, kernel_size)
helios/modules/transformer_helios.py:59
↓ 2 callersFunctioncenter_down_sample_3d
(x, kernel_size)
helios/diffusers_version/transformer_helios_diffusers.py:49
↓ 2 callersFunctioncheck_dim_and_resize
(tensor_list)
eval/utils/third_party/amt/utils/utils.py:300
↓ 2 callersMethodclose
(self)
tools/others/benchmark/benchmark_compile_performance.py:39
↓ 2 callersMethodclose
(self)
eval/utils/third_party/amt/trainers/logger.py:60
↓ 2 callersFunctioncompute_l2_distance
Compute L2 distance between two tensors
helios/utils/utils_recycle_single.py:324
↓ 2 callersFunctioncompute_l2_distance
Compute L2 distance between two tensors
helios/utils/utils_recycle_batch.py:456
↓ 2 callersFunctioncompute_l2_distance_batch
Compute L2 distances between new tensor and all stored tensors efficiently.
helios/utils/utils_recycle_single.py:309
↓ 2 callersFunctioncompute_l2_distance_batch
Compute L2 distances between new tensor and all stored tensors efficiently.
helios/utils/utils_recycle_batch.py:441
↓ 2 callersMethodcompute_latent_statistics
(self, latent_chunk: torch.Tensor)
helios/utils/utils_base.py:704
↓ 2 callersMethodcompute_motion_score_on_frames
Calculate motion smoothness score for a list of frames
eval/6_get_drifting_motion_smoothness.py:107
↓ 2 callersFunctioncorrupt_model_input
( model_input, # choose mode corrupt_mode="noise", # "noise" | "downsample" | "random" noise_
helios/utils/utils_helios_base.py:211
↓ 2 callersFunctioncreate_video_mapping
Create mapping from (step, idx) to filename
tools/gradio/comparison/gradio_compare_diff-ablation.py:34
↓ 2 callersFunctioncreate_video_mapping
Create (step, idx) -> filename mapping
tools/gradio/comparison/gradio_compare_diff-ckpt.py:35
↓ 2 callersMethoddetect_drift
(self, current_mean: torch.Tensor, current_var: torch.Tensor)
helios/utils/utils_base.py:720
↓ 2 callersMethoddisable_kv_cache
(self)
helios/modules/transformer_helios.py:1120
↓ 2 callersMethodenable_kv_cache
(self)
helios/modules/transformer_helios.py:1115
↓ 2 callersFunctionestimate
(tenOne, tenTwo)
eval/utils/third_party/amt/flow_generation/liteflownet/run.py:527
↓ 2 callersMethodeta_format
(self, eta)
eval/utils/third_party/amt/trainers/base_trainer.py:234
↓ 2 callersFunctionevaluate_aesthetic_on_frames
Evaluate aesthetic on a set of frames
eval/5_get_drifting_aesthetic.py:31
↓ 2 callersFunctionevaluate_portion_naturalness
Evaluate naturalness for start or end portion of video
eval/8_get_drifting_naturalness.py:130
↓ 2 callersFunctionevaluate_semantic_on_portion
Evaluate semantic consistency on a portion of the video
eval/7_get_drifting_semantic.py:42
↓ 2 callersFunctionfetch_image
(ele: dict[str, str | Image.Image], size_factor: int = IMAGE_FACTOR)
helios/videoalign/vision_process.py:79
↓ 2 callersFunctiongan_crop_video_spatial
(x, scale=0.5)
helios/utils/utils_helios_post.py:382
↓ 2 callersFunctiongather_zero3ema
(accelerator, ema_model)
helios/utils/create_ema_zero3_lora.py:190
↓ 2 callersFunctiongaussian
(window_size, sigma)
eval/utils/third_party/amt/metrics/psnr_ssim.py:10
↓ 2 callersFunctionget_frame_indices
(num_frames, vlen, sample="rand", fix_start=None, input_fps=1, max_num_frames=-1)
eval/utils/utils.py:43
↓ 2 callersMethodget_mutli_flow_robust_weight
(self, flow_pred, flow_gt)
eval/utils/third_party/amt/losses/loss.py:195
↓ 2 callersFunctionget_optimizer
(args, accelerator, params_to_optimize, use_deepspeed: bool = False)
helios/utils/utils_base.py:76
↓ 2 callersFunctionget_pairs
Return set of symbol pairs in a word. Word is represented as tuple of symbols (symbols being variable-length strings).
eval/utils/third_party/ViCLIP/simple_tokenizer.py:50
↓ 2 callersMethodget_robust_weight
(self, flow_pred, flow_gt)
eval/utils/third_party/amt/losses/loss.py:158
↓ 2 callersFunctionget_video_list
Get all mp4 videos from the folder
tools/gradio/comparison/gradio_compare_diff-ablation.py:17
↓ 2 callersFunctionget_video_list
Get all mp4 videos from folder
tools/gradio/comparison/gradio_compare_diff-ckpt.py:18
↓ 2 callersMethodinit_sigmas
initialize the global timesteps and sigmas
helios/scheduler/scheduling_helios.py:100
↓ 2 callersMethodinit_sigmas
initialize the global timesteps and sigmas
helios/diffusers_version/scheduling_helios_diffusers.py:90
↓ 2 callersFunctionload_model_checkpoint
( args, checkpoint_path, transformer, pipeline_class=None, norm_layer_prefixes=None, c
helios/utils/utils_base.py:455
↓ 2 callersFunctionload_temp_embed_with_mismatch
Add/Remove extra temporal_embeddings as needed. https://arxiv.org/abs/2104.00650 shows adding zero paddings works. temp_embed_old: (1, n
eval/utils/third_party/ViCLIP/viclip_vision.py:17
↓ 2 callersFunctionlog_validation
( pipe, args, accelerator, pipeline_args, )
train_helios.py:2429
↓ 2 callersFunctionmaybe_to_contiguous
(x)
helios/modules/helios_kernels/tiled_linear.py:57
↓ 2 callersFunctionmaybe_zero_3
(param, ignore_status=False, name=None)
helios/videoalign/utils.py:95
↓ 2 callersFunctionmerge_dict_list
(dict_list)
helios/utils/utils_helios_post.py:558
↓ 2 callersFunctionnavigate_idx
Navigate to previous or next index
tools/gradio/comparison/gradio_compare_diff-video.py:239
↓ 2 callersFunctionnavigate_idx
Navigate to previous or next index
tools/gradio/comparison/gradio_compare_diff-ablation.py:292
↓ 2 callersFunctionnavigate_idx
Navigate to previous or next index
tools/gradio/comparison/gradio_compare_diff-ckpt.py:297
↓ 2 callersFunctionnavigate_step
Navigate to previous or next step
tools/gradio/comparison/gradio_compare_diff-ablation.py:256
↓ 2 callersFunctionnavigate_step
Navigate to previous or next step
tools/gradio/comparison/gradio_compare_diff-ckpt.py:261
↓ 2 callersFunctionoptimized_scale
(positive_flat, negative_flat)
helios/diffusers_version/pipeline_helios_diffusers.py:81
← previousnext →101–200 of 847, ranked by callers