MCPcopy Create free account

hub / github.com/allenai/molmo2 / functions

Functions2,270 in github.com/allenai/molmo2

⨍ Functions 2,270◇ Types & classes 480
↓ 1 callersMethod_get_frames_dir
(cls, data_split, video_name)
olmo/data/academic_video_track_datasets.py:568
↓ 1 callersFunction_get_gcs_conditional_retry
()
olmo/io.py:618
↓ 1 callersFunction_get_key
(state_dict: Dict[str, Any], key: str, pop: bool = False)
olmo/train/distributed_checkpointing.py:722
↓ 1 callersFunction_get_koala_lookup
Lazy-load koala YouTube ID -> filename mapping from metadata JSONs.
olmo/data/molmo2_datasets.py:366
↓ 1 callersFunction_get_model_mesh
(device_mesh: DeviceMesh)
olmo/dist_util.py:108
↓ 1 callersMethod_get_next_source
(self, rng, counts)
olmo/data/iterable_dataset_mixture.py:115
↓ 1 callersMethod_get_order
(self, dataset_ix, epoch)
olmo/data/iterable_dataset_mixture.py:123
↓ 1 callersFunction_get_s3_endpoint_url
(scheme: str)
olmo/io.py:818
↓ 1 callersFunction_get_s3_profile_name
(scheme: str)
olmo/io.py:794
↓ 1 callersMethod_get_style
(self)
olmo/data/academic_video_track_datasets.py:462
↓ 1 callersMethod_get_tmp_dir
(self, dir: PathOrStr)
olmo/train/checkpointer.py:257
↓ 1 callersMethod_get_video_dir
Default: VIDEO_HOME/{data_split}/videos/
olmo/data/academic_video_track_datasets.py:254
↓ 1 callersMethod_get_video_dir_for_source
(cls, source)
olmo/data/molmo2_video_track_datasets.py:163
↓ 1 callersMethod_get_video_dir_for_source
(cls, source)
olmo/data/molmo2_video_track_datasets.py:280
↓ 1 callersMethod_get_video_dir_for_source
(cls, source, split='test')
olmo/data/molmo2_video_track_datasets.py:461
↓ 1 callersMethod_get_video_path
Get the appropriate video path, preferring extracted clips when available.
olmo/data/academic_video_datasets.py:2107
↓ 1 callersFunction_http_file_exists
(url: str)
olmo/io.py:575
↓ 1 callersFunction_http_file_size
(url: str)
olmo/io.py:550
↓ 1 callersFunction_http_get_bytes_range
(url: str, bytes_start: int, num_bytes: int)
olmo/io.py:558
↓ 1 callersFunction_item_size
(item: WriteItem)
olmo/train/remote_filesystem.py:103
↓ 1 callersFunction_iter
()
olmo/preprocessing/image_preprocessor.py:116
↓ 1 callersFunction_iter
()
olmo/data/utils.py:192
↓ 1 callersFunction_iter
()
olmo/data/dataset_builders/download_urls.py:171
↓ 1 callersFunction_key
(path)
scripts/prepare_pretrained_model.py:18
↓ 1 callersMethod_load_extensions
(self)
olmo/data/academic_video_datasets.py:1889
↓ 1 callersMethod_load_extensions
Load cached extension mapping.
olmo/data/academic_video_datasets.py:2739
↓ 1 callersMethod_load_mappings
Load all label mappings in one place
olmo/data/academic_video_datasets.py:2499
↓ 1 callersMethod_load_missing_videos
Load cached set of missing video IDs.
olmo/data/academic_video_datasets.py:1669
↓ 1 callersFunction_make_video_path
Reconstruct full video path from HF row fields.
olmo/data/academic_video_datasets.py:1353
↓ 1 callersMethod_maybe_truncate_example
(self, ex)
olmo/models/molmo_point/molmo_point_text_preprocessor.py:181
↓ 1 callersMethod_merge_heads
(self, hidden_states)
olmo/hf_model/modeling_molmo_point.py:261
↓ 1 callersMethod_merge_heads
(self, hidden_states)
olmo/hf_model/modeling_molmo2.py:153
↓ 1 callersMethod_merge_heads
(self, hidden_states)
olmo/nn/image_vit.py:202
↓ 1 callersFunction_narrow_tensor_by_index
Narrow the tensor according to ``offsets`` and ``sizes``.
olmo/train/remote_filesystem.py:178
↓ 1 callersMethod_points_from_num_str
(self, text, image_w, image_h, extract_ids=False)
olmo/preprocessing/point_formatter.py:146
↓ 1 callersMethod_precompute_gt_masks
(cls)
olmo/data/academic_video_track_datasets.py:617
↓ 1 callersMethod_prepare_annotation_dir
Download and extract MeViS frames for all splits including "valid" (for MevisChallenge). Fallback chain per split: 1. frames_root exi
olmo/data/academic_video_track_datasets.py:572
↓ 1 callersMethod_prepare_cp_inputs
( self, input_ids: torch.Tensor, labels: Optional[torch.Tensor] = None, input_
olmo/models/molmo2/molmo2.py:432
↓ 1 callersMethod_prepare_dir
(self, dir: PathOrStr, ensure_exists: bool = True)
olmo/train/checkpointer.py:276
↓ 1 callersMethod_prepare_input_videos
( self, videos: VideoInput, **kwargs, )
olmo/hf_model/video_processing_molmo2.py:814
↓ 1 callersMethod_prepare_test_frames
Extract sav_{split}.tar to get JPEGImages_24fps/
olmo/data/molmo2_video_track_datasets.py:1144
↓ 1 callersMethod_prepare_train_frames
Extract sav_000..003.tar -> decode mp4s -> JPEGImages_24fps/
olmo/data/molmo2_video_track_datasets.py:1119
↓ 1 callersMethod_preprocess
Preprocess a video for the model. Args: videos (`VideoInput`): Video to preprocess. size (`Si
olmo/hf_model/video_processing_molmo2.py:860
↓ 1 callersMethod_process_mm_data
(self, example, rng)
olmo/models/molmo_point/molmo_point_example_preprocessor.py:273
↓ 1 callersMethod_reconstruct_sequences
(predictions, backpointers)
olmo/nn/beam_search.py:976
↓ 1 callersMethod_remove_sharded_checkpoint
(self, idx: int, checkpoints: List[Path])
olmo/train/trainer.py:665
↓ 1 callersFunction_remove_video_text
(text)
tests/data/test_preprocessor.py:8
↓ 1 callersFunction_resolve_sources_from_configs
For eval dataset, resolve config names to video source names.
olmo/data/molmo2_video_track_datasets.py:2053
↓ 1 callersFunction_s3_clear_directory
(scheme: str, bucket_name: str, prefix: str)
olmo/io.py:918
↓ 1 callersFunction_s3_get_bytes_range
( scheme: str, bucket_name: str, key: str, bytes_start: int, num_bytes: Optional[int] )
olmo/io.py:863
↓ 1 callersFunction_s3_list_directory
( scheme: str, bucket_name: str, prefix: str, recurse: bool = False, include_files: bool =
olmo/io.py:937
↓ 1 callersFunction_s3_upload
( source: Path, scheme: str, bucket_name: str, key: str, save_overwrite: bool = False, )
olmo/io.py:893
↓ 1 callersFunction_sampler_with_overrides
(sampler, **sampler_overrides)
olmo/data/video_loader.py:106
↓ 1 callersMethod_save_train_state
(self, dir: PathOrStr, wd: Path, train_state: Dict[str, Any])
olmo/train/checkpointer.py:249
↓ 1 callersMethod_scale_point
(self, point, scale)
olmo/models/molmo_point/molmo_point_data_formatter.py:655
↓ 1 callersMethod_scaled_dot_product_attention
Computes scaled dot product attention on query, key and value tensors, using an optional attention mask if passed, and applying dropo
olmo/nn/llm.py:1698
↓ 1 callersMethod_select_normalized_human_readable_label
Select a normalized label and find its human-readable equivalent.
olmo/models/molmo_point/molmo_point_data_formatter.py:87
↓ 1 callersMethod_shard_by_image_bounds
Shard images and pooled_patches_idx based on cumulative image bounds for balanced processing. Args: images: Imag
olmo/nn/vision_backbone.py:454
↓ 1 callersMethod_split_cloud_path
Split a full s3 path into the bucket name and path.
olmo/io.py:1015
↓ 1 callersFunction_step_function
( last_predictions: torch.Tensor, state: dict[str, torch.Tensor] )
tests/beam_search_test.py:74
↓ 1 callersMethod_teardown_tmp_dir
(self, dir: PathOrStr, tmp_dir: Path)
olmo/train/checkpointer.py:297
↓ 1 callersMethod_temporary_wd
(self, dir: PathOrStr)
olmo/train/checkpointer.py:332
↓ 1 callersFunction_test_template
(messages, formatter, tokenizer)
tests/preprocessing/test_data_formatter.py:6
↓ 1 callersMethod_update_initial_state
Expand tensors in a state dictionary from `(batch_size, *)` to `(batch_size * beam_size, *)`.
olmo/nn/beam_search.py:1320
↓ 1 callersMethod_update_state
( self, state: ConstraintStateType, last_prediction: torch.Tensor, )
olmo/nn/beam_search.py:714
↓ 1 callersMethod_update_state
(self, state: StateType, backpointer: torch.Tensor)
olmo/nn/beam_search.py:1335
↓ 1 callersFunction_wait_before_retry
(attempt: int)
olmo/io.py:490
↓ 1 callersFunction_write_items
( path: str, storage_key: str, items: List[WriteItem], planner: SavePlanner )
olmo/train/remote_filesystem.py:138
↓ 1 callersFunctiona_okvqa_score
(target, pred)
olmo/eval/vqa.py:112
↓ 1 callersFunctionabs_dist_norm
(pred: float, target: float)
olmo/eval/vsi_bench_utils.py:13
↓ 1 callersMethodadd_pos_emb
(self, x: torch.Tensor, patch_num: int)
olmo/hf_model/modeling_molmo2.py:306
↓ 1 callersMethodadd_pos_emb
(self, x: torch.Tensor, patch_num: int)
olmo/nn/image_vit.py:502
↓ 1 callersMethodadd_pos_emb
(self, x: torch.Tensor, patch_num: int)
olmo/nn/image_vit.py:719
↓ 1 callersMethodadd_wandb
(self, wandb_url)
olmo/train/trainer.py:129
↓ 1 callersFunctionall_to_all_4D
all-to-all for QKV Args: input (torch.tensor): a tensor sharded along dim scatter dim scatter_idx (int): default 1 g
olmo/nn/ulysses.py:29
↓ 1 callersFunctionall_to_all_5D
all-to-all for QKV forward (bs, seqlen/N, 3, hc, hs) -> (bs, seqlen, 3, hc/N, hs) Args: input (torch.tensor): a tensor sharded a
olmo/nn/ulysses.py:151
↓ 1 callersFunctionanls_metric
Calculates ANLS for DocVQA. There does not seem to be an official evaluation script. Public implementation on which this implementation is ba
olmo/eval/vqa.py:158
↓ 1 callersMethodapply
( self, state: ConstraintStateType, class_log_probabilities: torch.Tensor, )
olmo/nn/beam_search.py:856
↓ 1 callersMethodapply_activation_checkpointing
(self)
olmo/nn/llm.py:827
↓ 1 callersMethodapply_compile
(self, **kwargs)
olmo/nn/llm.py:835
↓ 1 callersMethodapply_cp
( self, cp_mesh: DeviceMesh, load_balancer: CPLoadBalancerType, head_stride: i
olmo/nn/llm.py:1995
↓ 1 callersMethodapply_fsdp2_v2
Apply FSDP(2) to the model when using context parallelism (CP). .. warning:: This should generally be called last if usi
olmo/models/molmo2/molmo2.py:214
↓ 1 callersFunctionapply_keywords
(prompt, example, keywords)
olmo/preprocessing/data_formatter.py:635
↓ 1 callersMethodapply_llama3_scaling_factor
(self, inv_freq: torch.Tensor)
olmo/nn/llm.py:1111
↓ 1 callersFunctionapply_rotary_pos_emb
Applies Rotary Position Embedding to the query and key tensors. Args: q (`torch.Tensor`): The query tensor. k (`torch.Tensor`): T
olmo/hf_model/modeling_molmo2.py:484
↓ 1 callersFunctionarange_for_pooling
( idx_arr: np.ndarray, pool_h: int, pool_w: int, )
olmo/hf_model/video_processing_molmo2.py:144
↓ 1 callersMethodbackward
(ctx: Any, *grad_output: Tensor)
olmo/nn/ulysses.py:139
↓ 1 callersFunctionbatch_pixels_to_patches
Reshape images of [n_images, h, w, 3] -> [n_images, n_patches, pixels_per_patch]
olmo/hf_model/image_processing_molmo2.py:218
↓ 1 callersFunctionbatch_pixels_to_patches
Reshape images of [n_images, h, w, 3] -> [n_images, n_patches, pixels_per_patch]
olmo/hf_model/video_processing_molmo2.py:124
↓ 1 callersMethodbatch_start
(self, global_total_tokens: int, device_batch_num_tokens: int, device_batch_num_loss_tokens: int, record: bool
olmo/train/trainer.py:219
↓ 1 callersMethodbuild
(cls, data)
olmo/preprocessing/preprocessor_utils.py:68
↓ 1 callersMethodbuild
(self, default_save_dir=None)
olmo/eval/inf_evaluator.py:249
↓ 1 callersMethodbuild_batched_images
( self, input_ids: torch.LongTensor, pixel_values: torch.Tensor, image_token_p
olmo/hf_model/modeling_molmo_point.py:1076
↓ 1 callersMethodbuild_batched_images
( self, input_ids: torch.LongTensor, pixel_values: torch.Tensor, image_token_p
olmo/hf_model/modeling_molmo2.py:1148
↓ 1 callersMethodbuild_batched_videos
( self, input_ids: torch.LongTensor, pixel_values_videos: torch.Tensor, video_
olmo/hf_model/modeling_molmo_point.py:1223
↓ 1 callersMethodbuild_batched_videos
( self, input_ids: torch.LongTensor, pixel_values_videos: torch.Tensor, video_
olmo/hf_model/modeling_molmo2.py:1295
↓ 1 callersMethodbuild_connector
(self, llm_config, device)
olmo/nn/vision_backbone.py:238
↓ 1 callersMethodbuild_from_legacy_config
(cls, config: D)
olmo/preprocessing/video_preprocessor.py:38
↓ 1 callersMethodbuild_input_embeddings
( self, input_ids: torch.LongTensor, images: Optional[torch.FloatTensor] = None, # im
olmo/hf_model/modeling_molmo2.py:1430
↓ 1 callersFunctionbuild_instruction_templates
(question: str, options: str, option_type)
olmo/preprocessing/multiple_choice_templates.py:88
← previousnext →501–600 of 2,270, ranked by callers