Sockeye

Latest version: v3.1.34

Safety actively analyzes 629004 Python packages for vulnerabilities to keep your Python projects secure.

Page 14 of 45

Fixed
- Avoid a crash due to an edge case when no model improvement has been observed by the time the learning rate gets reduced for the first time.

Fixed
- Enforce sentence batching for sockeye score tool, set default batch size to 56

Changed
- Use softmax with length in DotAttentionCell.
- Use `contrib.arange_like` in AutoRegressiveBias block to reduce number of ops.

Added

- Log the absolute number of `<unk>` tokens in source and target data

Fixed

- Fix: Guard against null division for small batch sizes.

Fixed

- Fixes a corner case bug by which the beam decoder can wrongly return a best hypothesis with -infinite score.

Has known vulnerabilities

Previous Next