v0.8.3

What's new in 0.8.3 (2024-02-02)

These are the changes in inference v0.8.3.

New features

• FEAT: add whisper.small and belle distilwhisper model, fix parameter in rerank by @zhanghx0905 in #944
• FEAT: Support jina-embeddings-v2-base-zh by @aresnow1 in #948
• FEAT: Support Yi VL by @codingl2k1 in #946
• FEAT: Support more embedding and rerank models by @aresnow1 in #959

Enhancements

• ENH: Record gpu mem status in workers by @ChengjieLi28 in #941
• ENH: Allow chat max_tokens is None by @codingl2k1 in #960
• ENH: chatglm ggml format supports system_prompt by @ChengjieLi28 in #962

Bug fixes

• BUG: Fix roles in chat UI by @aresnow1 in #949
• BUG: Fix heartbeat by @codingl2k1 in #957
• BUG: Fix model's content length by @aresnow1 in #955

Documentation

• DOC: Update readme by @aresnow1 in #938
• DOC: Add image model doc by @codingl2k1 in #947
• DOC: Add audio model doc by @codingl2k1 in #954
• DOC: Reorge model related docs by @onesuper in #961

New Contributors

• @zhanghx0905 made their first contribution in #944

Full Changelog: v0.8.2...v0.8.3