snapshot-2024-01-28

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