Ludwig v0.10.1
·
34 commits
to master
since this release
What's Changed
- Fixed a critical bug in Gemma model fine-tuning that prevented the model from learning when to stop generation. This is accomplished by using eos token in target tensor for instruction-tuning by @geoffreyangus in #3945
Full Changelog: v0.10.0...v0.10.1
Contributors
geoffreyangus