v0.1.3

The last version that feature transform by pandas.

Full Changelog: v0.1.2...v0.1.3

v0.1.2

• Fixed the OOM problem of DataFrameDataset by replacing list with numpy, and process all data to numpy before __get_item__
• Default value changes of DataFrameDataset:
  • is_raw parameter default to True
  • list_padding_value parameter default to None. Therefore, you need to set the list_padding_value when init DataFrameDataset or using collate_fn in DataFrameDataset when data loading if you have var-length sequence features
• Support string input type of sequence features (comma split)

Full Changelog: v0.1.1...v0.1.2

v0.1.1

Full Changelog: v0.1.0...v0.1.1

Fixed a minor problem when input sequence features are the same length.

v0.1.0

Full Changelog: https://github.com/xiahouzuoxin/torchrec/commits/v0.1.0

The first release with all features implemented as README.md.