espnetez.trainer.check_argument

Less than 1 minute

espnetez.trainer.check_argument

espnetez.trainer.check_argument(train_dump_dir, valid_dump_dir, train_dataset, valid_dataset, train_dataloader, valid_dataloader)

Validate the arguments for training and validation data sources.

This function checks the consistency of the input arguments used for specifying training and validation data sources. It ensures that the user adheres to the rules of specifying either dump directories, datasets, or dataloaders, but not a mix of them. The function will raise a ValueError if the conditions are not met.

Parameters:
- train_dump_dir (str or None) – The directory containing training dump files. Should be None if using datasets or dataloaders.
- valid_dump_dir (str or None) – The directory containing validation dump files. Should be None if using datasets or dataloaders.
- train_dataset (Dataset or None) – The training dataset. Should be None if using dump files or dataloaders.
- valid_dataset (Dataset or None) – The validation dataset. Should be None if using dump files or dataloaders.
- train_dataloader (DataLoader or None) – The training dataloader. Should be None if using dump files or datasets.
- valid_dataloader (DataLoader or None) – The validation dataloader. Should be None if using dump files or datasets.
Returns: Returns True if all checks pass.
Return type: bool
Raises:ValueError – If any of the argument conditions are violated.

Examples

Example of valid usage with dump files

check_argument(‘/path/to/train_dump’, ‘/path/to/valid_dump’, None, None, None,

None)

Example of valid usage with datasets

check_argument(None, None, train_dataset, valid_dataset, None, None)

Example of invalid usage - mix of dump files and datasets

check_argument(‘/path/to/train_dump’, None, train_dataset, None, None, None)

Raises ValueError: If you try to use dump file, dataset or dataloader should

be None.

NOTE

Ensure to specify at least one of the arguments: dump directories, datasets, or dataloaders. The function enforces exclusive use of each data source type.