MCPcopy Create free account

hub / github.com/AlmondGod/tinyworlds / functions

Functions154 in github.com/AlmondGod/tinyworlds

⨍ Functions 154◇ Types & classes 36
↓ 17 callersFunctionunwrap_model
(model: torch.nn.Module)
utils/distributed.py:82
↓ 7 callersMethodget_latents_from_indices
(self, indices, dim=-1)
models/fsq.py:60
↓ 6 callersMethod__init__
(self, embed_dim, hidden_dim, conditioning_dim=None)
models/st_transformer.py:99
↓ 6 callersFunctionfind_latest_checkpoint
Find latest checkpoint. If run_root_dir (and optional stage_name) are provided, search only under <run_root_dir>/<stage_name>/checkpoints (or
utils/utils.py:26
↓ 6 callersFunctionmissing
(path: Optional[str])
scripts/run_inference.py:29
↓ 6 callersFunctionshard_layers
(layers: Iterable[torch.nn.Module])
utils/distributed.py:48
↓ 6 callersMethodstep
(self, closure=None)
models/muon.py:33
↓ 5 callersFunction_load_video_dataset_pair
(dataset_cls, video_rel_path, h5_rel_path, num_frames, transform=None, fps=30, preload_ratio=1, **kwargs)
datasets/data_utils.py:27
↓ 5 callersFunctionload_data_and_data_loaders
(dataset, batch_size, num_frames=1, distributed=False, rank=0, world_size=1, fps=15, preload_ratio=1)
datasets/data_utils.py:144
↓ 5 callersFunctionreadable_timestamp
Generate a sortable timestamp for filenames (no weekday).
utils/utils.py:22
↓ 4 callersMethod__init__
(self, video_path, transform=None, save_path=None, train=True, num_frames=1, resolution=(64, 64), fps=30, prel
datasets/datasets.py:143
↓ 4 callersFunction_validate_amp_fsdp
(amp: bool, distributed: DistributedConfig)
utils/config.py:60
↓ 4 callersFunction_validate_distibuted_training
(nproc_per_node: int, distributed: DistributedConfig)
utils/config.py:67
↓ 4 callersFunction_validate_distributed_device
(device: DeviceType, distributed: DistributedConfig)
utils/config.py:74
↓ 4 callersMethodencode
(self, frames)
models/latent_actions.py:132
↓ 4 callersFunctionprepare_pipeline_run_root
Create a top-level run root directory results/<timestamp_or_name>
utils/utils.py:312
↓ 3 callersMethod__init__
(self, frame_size=(128, 128), patch_size=8, embed_dim=128, num_heads=8, hidden_dim=256, num_b
models/video_tokenizer.py:80
↓ 3 callersMethod_resolve_dtype
(self, value: str | torch.dtype)
utils/config.py:23
↓ 3 callersFunctionbuild_spatial_only_pe
(frame_size, patch_size, embed_dim, device='cpu', dtype=torch.float32)
models/positional_encoding.py:31
↓ 3 callersFunctioncleanup_distributed
(is_distributed: bool)
utils/distributed.py:96
↓ 3 callersFunctioncollect_checkpoint_paths
(roots, model_name)
utils/utils.py:34
↓ 3 callersFunctioncreate_cosine_scheduler
(optimizer, total_steps, warmup_fraction=0.05, min_lr_fraction=0.01)
utils/scheduler_utils.py:14
↓ 3 callersFunctioncreate_optimizer
(model, args)
utils/optimizer_utils.py:7
↓ 3 callersFunctionfinish_wandb
Finish the W&B run
utils/wandb_utils.py:69
↓ 3 callersMethodget_indices_from_latents
(self, latents, dim=-1)
models/fsq.py:47
↓ 3 callersFunctioninit_distributed_from_env
Initialize DeviceMesh from torchrun env vars. Returns a context dict with is_distributed, world_size, is_main, device_mesh.
utils/distributed.py:14
↓ 3 callersFunctioninit_wandb
(project_name: str, config: Dict[str, Any], run_name: Optional[str] = None)
utils/wandb_utils.py:8
↓ 3 callersFunctionload_latent_actions_from_checkpoint
Instantiate LatentActionModel from a checkpoint's saved config and load weights.
utils/utils.py:235
↓ 3 callersFunctionload_stage_config_merged
Load a stage config YAML, then overlay values from training_config.yaml (priority), restricted to keys that exist in the stage schema. CLI dotlist ov
utils/config.py:344
↓ 3 callersFunctionload_videotokenizer_from_checkpoint
Instantiate VideoTokenizer from a checkpoint's saved config and load weights.
utils/utils.py:203
↓ 3 callersFunctionlog_learning_rate
(optimizer: torch.optim.Optimizer, step: int)
utils/wandb_utils.py:35
↓ 3 callersFunctionlog_system_metrics
(step: int)
utils/wandb_utils.py:61
↓ 3 callersFunctionprepare_model_for_distributed
(model: torch.nn.Module, config: DistributedConfig, model_type: ModelType, device_mesh: DeviceMesh)
utils/distributed.py:37
↓ 3 callersFunctionprepare_stage_dirs
Create stage subdirectories under the given run root. Structure: <run_root_dir>/<stage_name>/checkpoints <run_root_dir>/<stage_name>/
utils/utils.py:322
↓ 3 callersFunctionprint_param_count_if_main
(model: torch.nn.Module, model_name: str, is_main: bool)
utils/distributed.py:86
↓ 3 callersFunctionrun_command
(cmd, description)
utils/utils.py:144
↓ 3 callersFunctionsave_training_state
Save a checkpoint with model/optimizer/scheduler and the exact config. The filename includes the global step and a timestamp for uniqueness.
utils/utils.py:162
↓ 3 callersFunctionsincos_1d
(L, D, device, dtype)
models/positional_encoding.py:5
↓ 3 callersMethodtokenize
(self, frames)
models/video_tokenizer.py:96
↓ 3 callersFunctionvisualize_reconstruction
(original, reconstruction, save_path=None)
datasets/data_utils.py:172
↓ 2 callersMethod__init__
(self, frame_size=(128, 128), n_actions=8, patch_size=8, embed_dim=128, num_heads=8, hidden_
models/latent_actions.py:100
↓ 2 callersMethod__init__
(self, embed_dim, eps=1e-5)
models/norms.py:6
↓ 2 callersFunctiondownload_pairs
(pairs: List[Tuple[str, str]], output_dir: Path, resume: bool = True, repo_type: str = "model")
scripts/download_assets.py:29
↓ 2 callersMethodforward
(self, discrete_latents, training=True, conditioning=None, targets=None)
models/dynamics.py:34
↓ 2 callersFunctionload_config
(config_cls, default_config_path: Optional[str] = None)
utils/config.py:316
↓ 2 callersFunctionload_dynamics_from_checkpoint
Instantiate DynamicsModel from a checkpoint's saved config and load weights.
utils/utils.py:266
↓ 2 callersFunctionlog_action_distribution
(action_indices: torch.Tensor, step: int, n_actions: int)
utils/wandb_utils.py:47
↓ 2 callersFunctionproject_wide_search
()
utils/utils.py:88
↓ 2 callersFunctionrepo_root
()
scripts/download_assets.py:13
↓ 2 callersFunctionrun_dir_of
(path: str)
utils/utils.py:73
↓ 2 callersFunctionsample_random_action
(n_actions)
utils/inference_utils.py:122
↓ 2 callersMethodscale_and_shift
(self, z)
models/fsq.py:18
↓ 2 callersMethodunscale_and_unshift
(self, z)
models/fsq.py:22
↓ 2 callersFunctionvisualize_batch
(frames, save_path=None, title="Video Sequences Batch", max_batch_size=8, max_seq_length=8)
scripts/visualize_batch.py:9
↓ 1 callersFunction_create_adamw
(model, args)
utils/optimizer_utils.py:19
↓ 1 callersFunction_create_muon_split
(model, args)
utils/optimizer_utils.py:28
↓ 1 callersFunction_default_video_transform
()
datasets/data_utils.py:20
↓ 1 callersMethod_preprocess_video
( self, video_path: str, resize_to: Tuple[int, int], read_step: int = 1,
datasets/datasets.py:71
↓ 1 callersFunction_split_decay_params
(model)
utils/optimizer_utils.py:72
↓ 1 callersFunctionadd_candidate
(path: str)
utils/utils.py:44
↓ 1 callersFunctionbuild_parser
()
scripts/download_assets.py:111
↓ 1 callersFunctioncosine_with_warmup
(step, *, warmup_steps, total_steps, min_lr=0.0)
utils/scheduler_utils.py:4
↓ 1 callersFunctiondata_loaders
(train_data, val_data, batch_size, distributed=False, rank=0, world_size=1)
datasets/data_utils.py:111
↓ 1 callersMethoddetokenize
(self, quantized_z)
models/video_tokenizer.py:103
↓ 1 callersMethodexp_schedule_torch
(self, t, T, P_total, k, device)
models/dynamics.py:84
↓ 1 callersFunctionexpand_patterns
(repo_id: str, patterns: List[str], repo_type: str)
scripts/download_assets.py:17
↓ 1 callersFunctionextract_step
(path: str)
utils/utils.py:136
↓ 1 callersMethodforward_inference
(self, context_latents, prediction_horizon, num_steps, index_to_latents_fn, conditioning=None, schedule_k=5.0,
models/dynamics.py:95
↓ 1 callersFunctionget_action_latent
(args, inferred_actions, n_actions, context_frames, latent_action_model, step)
utils/inference_utils.py:127
↓ 1 callersMethodget_mixed_precision_policy
(self)
utils/config.py:52
↓ 1 callersFunctionload_models
(video_tokenizer_path, latent_actions_path, dynamics_path, device, use_actions=True)
utils/inference_utils.py:10
↓ 1 callersFunctionload_picodoom
(num_frames=4, fps=30, preload_ratio=1)
datasets/data_utils.py:89
↓ 1 callersFunctionload_pole_position
(num_frames=4, fps=15, preload_ratio=1)
datasets/data_utils.py:78
↓ 1 callersFunctionload_pong
(num_frames=1, fps=15, preload_ratio=1)
datasets/data_utils.py:56
↓ 1 callersFunctionload_sonic
(num_frames=4, fps=15, preload_ratio=1)
datasets/data_utils.py:67
↓ 1 callersFunctionload_zelda
(num_frames=4, fps=15, preload_ratio=1)
datasets/data_utils.py:100
↓ 1 callersFunctionlog_training_metrics
(step: int, metrics: Dict[str, float], prefix: str = "train")
utils/wandb_utils.py:29
↓ 1 callersFunctionmain
()
scripts/run_inference.py:16
↓ 1 callersFunctionmain
()
scripts/full_train.py:8
↓ 1 callersFunctionmain
()
scripts/visualize_batch.py:78
↓ 1 callersFunctionmain
()
scripts/train_latent_actions.py:17
↓ 1 callersFunctionmain
()
scripts/train_video_tokenizer.py:17
↓ 1 callersFunctionmain
(argv=None)
scripts/download_assets.py:136
↓ 1 callersFunctionmain
()
scripts/train_dynamics.py:29
↓ 1 callersMethodmoe_aux_loss
(self)
models/st_transformer.py:260
↓ 1 callersMethodmoe_expert_utilization
Per-block expert token fractions. Returns dict of block_idx -> [num_experts] tensor.
models/st_transformer.py:268
↓ 1 callersFunctionsave_frames_as_mp4
(frames, output_path, fps=2)
utils/inference_utils.py:97
↓ 1 callersFunctionsincos_time
(T, D, device, dtype)
models/positional_encoding.py:26
↓ 1 callersMethodto_policy
(self)
utils/config.py:31
↓ 1 callersFunctionvisualize_batch_with_stats
(frames, save_path=None, title="Video Sequences Batch with Statistics")
scripts/visualize_batch.py:50
↓ 1 callersFunctionvisualize_inference
(predicted_frames, ground_truth_frames, inferred_actions, fps, use_actions=True)
utils/inference_utils.py:23
↓ 1 callersFunctionzeropower_via_newtonschulz5
(G: Tensor, steps: int = 10, eps: float = 1e-7)
models/muon.py:8
Method__del__
(self)
datasets/datasets.py:137
Method__getitem__
(self, index: int)
datasets/datasets.py:116
Method__init__
( self, video_path: str, transform=None, # postnormalization save_path: Option
datasets/datasets.py:12
Method__init__
(self, video_path, transform=None, save_path=None, train=True, num_frames=4, resolution=(64, 64), fps=30, prel
datasets/datasets.py:160
Method__init__
(self, video_path, transform=None, save_path=None, train=True, num_frames=4, resolution=(128, 128), fps=15, pr
datasets/datasets.py:178
Method__init__
(self, video_path, transform=None, save_path=None, train=True, num_frames=4, resolution=(128, 128), fps=30, pr
datasets/datasets.py:196
Method__init__
(self, video_path, transform=None, save_path=None, train=True, num_frames=4, resolution=(128, 128), fps=15, pr
datasets/datasets.py:214
Method__init__
(self, frame_size=(128, 128), patch_size=8, embed_dim=128, num_heads=8, hidden_dim=256, num_
models/video_tokenizer.py:12
next →1–100 of 154, ranked by callers