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 264266 次贡献
更少更多
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 fix_nixl_triton_attn 分支3240a1e486 Merge branch 'main' into fix_nixl_triton_attn
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 fix_nixl_get_finished_handshake_failure 分支3245cc77c3 [Bugfix][NIXL] Skip heteroblocksize post-processing for failed recv requests in get_finished
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 fix_moe_test_flashinfer 分支05a556cf7b [Bugfix] Fix /moe/test_flashinfer.py
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 fix_fi_cutlass 分支比较 2 提交 »6a7b85d940 fix9f7ee8921a Merge branch 'main' into fix-qwen3-parse
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 fix_ds_eagle 分支9ffb182ba3 Fix config passed to deepseek_eagle
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 fix/topk-hash-indices-dtype 分支比较 7 提交 »b4dbbc7102 Fix dtype mismatch in topk_softplus_sqrt for DeepEP backends6d7a3fab28 Pass NCCL device group to DeepEP v2 ElasticBuffer7053a17886 Add barrier before mk.apply in MK test to sync ranks6e94d6f7d1 Use enforce_eager=True for MK tests (do_expand=True for DeepEP v2)6a2d75c993 Wrap MK test worker in set_current_vllm_config context
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 fix/flashinfer-nvfp4-cross-row-scale-corruption 分支比较 2 提交 »8bf46ccf97 tweaks0d92fc63fd [Bugfix] Fix FlashInfer NVFP4 cross-row scale corruption in MoE quant
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 fix/eplb-prometheus-metrics 分支比较 2 提交 »da742158e1 [Misc] Redesign EPLB Prometheus metrics for per-rank aggregation53c8d72f71 [Misc] Add EPLB balancedness to Prometheus metrics
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm 的 fix/eplb-nvfp4-modelopt 分支比较 4 提交 »36486dd057 EPLB: add post-rearrangement weight invariant checks1030f4c8b8 [BugFix] Fix EPLB balancedness metric using wrong dimension36b60baf60 Fix EPLB + NVFP4 modelopt: use direct Parameter assignment for g1/g2_alphas4f57aa6549 Fix EPLB + NVFP4: exclude broadcast scales and fix stale quant config