torchrec.datasets.scripts¶
torchrec.datasets.scripts.contiguous_preproc_criteo¶
- torchrec.datasets.scripts.contiguous_preproc_criteo.main(argv: List[str]) None ¶
This function processes the sparse features (.npy) to be contiguous and saves the result in a separate (.npy) file.
- Parameters:
argv (List[str]) – Command line args.
- Returns:
None.
- torchrec.datasets.scripts.contiguous_preproc_criteo.parse_args(argv: List[str]) Namespace ¶
torchrec.datasets.scripts.npy_preproc_criteo¶
- torchrec.datasets.scripts.npy_preproc_criteo.main(argv: List[str]) None ¶
This function preprocesses the raw Criteo tsvs into the format (npy binary) expected by InMemoryBinaryCriteoIterDataPipe.
- Parameters:
argv (List[str]) – Command line args.
- Returns:
None.
- torchrec.datasets.scripts.npy_preproc_criteo.parse_args(argv: List[str]) Namespace ¶