12月1月2月3月4月5月6月7月8月9月10月11月
周一周三周五
一年内 260013 次贡献
更少更多
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 gma/muon_triton 分支比较 4 提交 »87f1979d9f feat: Add Triton-optimized Muon optimizer implementation632ab6b2b0 make torch.compile more adaptive to old pytorch versione02c0ec98b apply torch.compile to Muon optimizerbbb4bfcf6e make muon optimizer totally running on GPU
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 pr-7806 分支比较 5 提交 »040316e973 show warning when multiple matches founda518a27d08 update blogf2b8c4984c add links to preset code and messages to welcome PRsc592949a3e fix codeblock lang0cf73fed65 Support custom partitioning patterns for AutoTP
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 pr_moe_tutorial 分支7b24585fdb Merge branch 'master' into pr_moe_tutorial
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 gma/investigate_7334 分支比较 10 提交 »d5791b999b use h100 instead68b675156f remove cuda version restriction4849e5b245 use a6000 instead of v100eb59f2ef15 update nv-a6000 workflow to test test_paralle.py343820be05 add a temp test for investigation
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 pr-7798-clean 分支比较 2 提交 »da9b77384a Address review feedback for Muon ZeRO Stage 3 support56451ea0fe Extending Muon Optimizer Support for ZeRO Stage 3
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 patch-z1-cont-grad 分支b42f1a3aec Merge branch 'master' into patch-z1-cont-grad
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 paper 分支比较 2 提交 »638e67e4bd Update index.mdf693e42554 Update README.md
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 gma/fix_cpu_train 分支比较 3 提交 »7f8aff5120 test(zero): add training step coverage for accelerator flatten path0bfe45e3aa remove unnecessary detach commentb70e8629ee fix(zero): detach flat buffer to prevent autograd inplace error on CPU accelerator
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 olruwase/zero_stage1_elastic_checkpoint 分支比较 2 提交 »b0cba77652 Merge with saksham-zero1-fixes branch6a4db4bf1f Fix incorrect communication interval schema in checkpointing logic
- frostbyte_neo 推送了仓库 github-featured/deepspeedai--deepspeed 的 olruwase/zinf_none_swapper 分支a0842da4fb Merge branch 'master' into olruwase/zinf_none_swapper
