`nemo_automodel.components.datasets.diffusion.meta_files_dataset`#

Module Contents#

Classes#

MetaFilesDataset

PyTorch dataset for WAN2.1 .meta files.

Functions#

`collate_fn`
`build_node_parallel_sampler`
`build_dataloader`
`create_dataloader`

Data#

API#

nemo_automodel.components.datasets.diffusion.meta_files_dataset.logger#: ‘getLogger(…)’

class nemo_automodel.components.datasets.diffusion.meta_files_dataset.MetaFilesDataset( meta_folder: str, transform_text: Optional[Callable[[torch.Tensor], torch.Tensor]] = None, transform_video: Optional[Callable[[torch.Tensor], torch.Tensor]] = None, filter_fn: Optional[Callable[[Dict], bool]] = None, device: str = 'cpu', max_files: Optional[int] = None, )#

Bases: torch.utils.data.Dataset

PyTorch dataset for WAN2.1 .meta files.

Initialization

_log_dataset_stats() → None#

__len__() → int#

__getitem__(index: int) → Dict[str, torch.Tensor]#

nemo_automodel.components.datasets.diffusion.meta_files_dataset.collate_fn( batch: List[Dict[str, torch.Tensor]], ) → Dict[str, torch.Tensor]#

nemo_automodel.components.datasets.diffusion.meta_files_dataset.build_node_parallel_sampler( dataset: torch.utils.data.Dataset, dp_rank: int, dp_world_size: int, shuffle: bool = True, ) → Optional[torch.utils.data.DistributedSampler]#

nemo_automodel.components.datasets.diffusion.meta_files_dataset.build_dataloader( *, meta_folder: str, batch_size: int, dp_rank: int, dp_world_size: int, shuffle: bool = True, num_workers: int = 2, device: str = 'cpu', transform_text: Optional[Callable[[torch.Tensor], torch.Tensor]] = None, transform_video: Optional[Callable[[torch.Tensor], torch.Tensor]] = None, filter_fn: Optional[Callable[[Dict], bool]] = None, max_files: Optional[int] = None, ) → Tuple[torch.utils.data.DataLoader, Optional[torch.utils.data.DistributedSampler]]#

nemo_automodel.components.datasets.diffusion.meta_files_dataset.create_dataloader( meta_folder: str, batch_size: int, num_nodes: int, ) → Tuple[torch.utils.data.DataLoader, Optional[torch.utils.data.DistributedSampler]]#