2.0.1

What's Changed

• Evaluation updates: add RICES + prompt ensembling by @i-gao in #220
• Pass **kwargs in "generate" function instead of passing each argument by name by @DavidMChan in #236
• update setup and requirements by @loftusa in #241

New Contributors

• @ElegantLin made their first contribution in #229
• @DavidMChan made their first contribution in #236
• @loftusa made their first contribution in #241

Full Changelog: 2.0.0...2.0.1

Second release

Major codebase updates, along with five new models.

Changes:

• Add gradient checkpointing, FullyShardedDataParallel
• Remove color jitter when training
• Fix cross-attention bug when calling generate()
• Model releases
  • (CLIP ViT-L-14 / MPT-1B)
  • (CLIP ViT-L-14 / MPT-1B Dolly)
  • (CLIP ViT-L-14 / RedPajama-3B)
  • (CLIP ViT-L-14 / RedPajama-3B Instruct)
  • (CLIP ViT-L-14 / MPT-7B)

Initial version w/ __init__ fix

Thanks to @chris-alexiuk for fixing!

First release

Merge pull request #128 from chris-alexiuk/main

Adding __init__.py to avoid module not found errors