What's Changed

Release

• [release] update version (#5654) by Hongxin Liu
• [release] grok-1 inference benchmark (#5500) by binmakeswell
• [release] grok-1 314b inference (#5490) by binmakeswell

Hotfix

• [hotfix] add soft link to support required files (#5661) by Tong Li
• [hotfix] Fixed fused layernorm bug without apex (#5609) by Edenzzzz
• [hotfix] Fix examples no pad token & auto parallel codegen bug; (#5606) by Edenzzzz
• [hotfix] fix typo s/get_defualt_parser /get_default_parser (#5548) by digger yu
• [hotfix] quick fixes to make legacy tutorials runnable (#5559) by Edenzzzz
• [hotfix] set return_outputs=False in examples and polish code (#5404) by Wenhao Chen
• [hotfix] fix typo s/keywrods/keywords etc. (#5429) by digger yu

News

• [news] llama3 and open-sora v1.1 (#5655) by binmakeswell

Lazyinit

• [lazyinit] skip whisper test (#5653) by Hongxin Liu

Shardformer

• [shardformer] refactor pipeline grad ckpt config (#5646) by Hongxin Liu
• [shardformer] fix chatglm implementation (#5644) by Hongxin Liu
• [shardformer] remove useless code (#5645) by flybird11111
• [shardformer] update transformers (#5583) by Wang Binluo
• [shardformer] fix pipeline grad ckpt (#5620) by Hongxin Liu
• [shardformer] refactor embedding resize (#5603) by flybird11111
• [shardformer] Sequence Parallelism Optimization (#5533) by Zhongkai Zhao
• [shardformer] fix pipeline forward error if custom layer distribution is used (#5189) by Insu Jang
• [shardformer] update colo attention to support custom mask (#5510) by Hongxin Liu
• [shardformer]Fix lm parallel. (#5480) by flybird11111
• [shardformer] fix gathering output when using tensor parallelism (#5431) by flybird11111

Fix

• [Fix]: implement thread-safety singleton to avoid deadlock for very large-scale training scenarios (#5625) by Season
• [fix] fix typo s/muiti-node /multi-node etc. (#5448) by digger yu
• [Fix] Grok-1 use tokenizer from the same pretrained path (#5532) by Yuanheng Zhao
• [fix] fix grok-1 example typo (#5506) by Yuanheng Zhao

Coloattention

• [coloattention]modify coloattention (#5627) by flybird11111

Example

• [example] llama3 (#5631) by binmakeswell
• [example] update Grok-1 inference (#5495) by Yuanheng Zhao
• [example] add grok-1 inference (#5485) by Hongxin Liu

Exampe

• [exampe] update llama example (#5626) by Hongxin Liu

Feature

• [Feature] Support LLaMA-3 CPT and ST (#5619) by Tong Li

Zero

• [zero] support multiple (partial) backward passes (#5596) by Hongxin Liu

Doc

• [doc] fix ColossalMoE readme (#5599) by Camille Zhong
• [doc] update open-sora demo (#5479) by binmakeswell
• [doc] release Open-Sora 1.0 with model weights (#5468) by binmakeswell

Devops

• [devops] remove post commit ci (#5566) by Hongxin Liu
• [devops] fix example test ci (#5504) by Hongxin Liu
• [devops] fix compatibility (#5444) by Hongxin Liu

Shardformer, pipeline

• [shardformer, pipeline] add gradient_checkpointing_ratio and heterogenous shard policy for llama (#5508) by Wenhao Chen

Colossalchat

• [ColossalChat] Update RLHF V2 (#5286) by YeAnbang

Format

• [format] applied code formatting on changed files in pull request 5510 (#5517) by github-actions[bot]

Full Changelog: https://github.com/hpcaitech/ColossalAI/compare/v0.3.7...v0.3.6