peft
https://github.com/huggingface/peft
Triage Issues!
When you volunteer to triage issues, you'll receive an email each day with a link to an open issue that needs help in this project. You'll also receive instructions on how to triage issues.
Triage Docs!
Receive a documented method or class from your favorite GitHub repos in your inbox every day. If you're really pro, receive undocumented methods or classes and supercharge your commit history.
not yet supported1 Subscribers
Add a CodeTriage badge to peft
Help out
- Issues
- merge_adapter() / merge_and_unload() silently drop adapter when base layer is parametrized (e.g. weight_norm)
- FIX Keep LoRA weight tying after adapter dtype cast
- HRA forward crashes for Conv2d padding='same' and padding='valid'
- BOFT fails on Conv2d layers with non-square kernels
- CorDA and LoRA-GA preprocessing fail with a cache filename in the current directory
- FIX Match IA3 feedforward modules on a dotted suffix
- Preserve source layer targeting in weighted LoRA and IA³ adapters
- add_weighted_adapter silently drops source layers when targeting filters differ (LoRA and IA3)
- Add Bayesian-LoRA as a LoRA variant
- FSDP checkpoint resume deadlocks across ranks during distributed optimizer state sync (Llama-405B, 8x H200)
- Docs
- not yet supported