What's Changed
- Add prompt_lookup_num_tokens parameter by @oobabooga in #5296
- Fixed whisper readme Typo Hyperlink by @Katehuuh in #5281
- Fix error "Attempting to unscale FP16 gradients" when training LoRA by @optozorax in #5268
- Update n_gpu_layers to 256 for larger models. by @Ph0rk0z in #5262
- Update position of show-controls label to avoid line breaks due to font size by @Yiximail in #5256
- do_sample false when temperature 0 by @StefanDanielSchwarz in #5275
- Bump llama-cpp-python to 0.2.29 by @oobabooga in #5307
- Bump llama-cpp-python to 0.2.31 by @oobabooga in #5345
- Fix logprobs tokens in OpenAI API by @lmg-anon in #5339
- Properly handle Images with RGBA color format by @ercanozer in #5332
- Allow content arrays for multimodal OpenAI requests by @Cohee1207 in #5277
- Merge dev branch by @oobabooga in #5347
- Merge dev branch by @oobabooga in #5348
- Bump exllamav2 to 0.0.12 by @oobabooga in #5352
- Bump hqq from 0.1.2 to 0.1.2.post1 by @dependabot in #5349
- Added trust_remote_code to deepspeed when loading model by @sam-ngu in #5237
- Merge dev branch by @oobabooga in #5379
New Contributors
- @Katehuuh made their first contribution in #5281
- @optozorax made their first contribution in #5268
- @lmg-anon made their first contribution in #5339
- @ercanozer made their first contribution in #5332
- @sam-ngu made their first contribution in #5237
Full Changelog: snapshot-2024-01-14...snapshot-2024-01-28