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 264209 次贡献
更少更多
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 ucx_oneshot_ar 分支b0ed553028 [Distributed] Add UCX one-shot AllReduce for DP metadata sync
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 tpu_v1_optimized 分支比较 7 提交 »70b4e46e70 compilation is fixed5fb9dbe6f6 fix capture model996b92ccb4 swap works!2b0526fa15 works!7be649256f fixes
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 topk_id_hack 分支比较 2 提交 »8209f9057d i honestly can't believe i spelled it that way19c51c3439 merge main, add environment variable, factor into function
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 tms/nvfp4-nan-contamination-test 分支比较 3 提交 »8cb4e2a378 [Test] Add NVFP4 scale padding contamination test for mm_fp46417976a66 [Test] Fix WorkspaceManager test imports0aa474a0e6 [Test] Add NVFP4 NaN contamination mechanism unit tests
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 tms/fix-nan 分支比较 10 提交 »dcbedb7661 [Bugfix] Add NaN masking to NVFP4 quantization to prevent output contamination0c534be7bf [Bugfix] Revert NVFP4 check_tensor calls (incompatible with fullgraph)f408ad2b73 [Bugfix] Make check_tensor fullgraph-safe and support FP8464b91bb42 [Bugfix] Fix check_tensor for torch.compile and FP8 dtypesd066cf30be [Bugfix] Fix torch.compile graph break in NaN detector check_tensor
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 tms/distributed_timeout 分支比较 2 提交 »2f86f710dd Fix precommitfeeb17303d Add VLLM_DISTRIBUTED_INIT_TIMEOUT_SECONDS