开放[Usage]: How to get query embeddings from ColBERT?good first issueusagevllm-project/vllm #42,234 创建于 16天前 · Python · 80,034 stars9 条评论0 个反应1 负责人
开放[Docs] Document NIXL KV connector metrics aggregation semanticsgood first issuevllm-project/vllm #41,230 创建于 27天前 · Python · 80,034 stars4 条评论1 个反应1 负责人
开放[Feature]: Integrate fused `kMoEFinalizeARResidualRMSNorm` from FlashInferfeature requesthelp wantedvllm-project/vllm #40,544 创建于 上个月 · Python · 80,034 stars3 条评论1 个反应0 负责人
开放[Feature]: Priority scheduling supports preemption of requests in the running queue by requests in the waiting queuefeature requesthelp wantedvllm-project/vllm #40,004 创建于 上个月 · Python · 80,034 stars5 条评论0 个反应0 负责人
开放[torch.compile] config hashing refactor follow-upsfeature requestgood first issuehelp wantedvllm-project/vllm #39,479 创建于 2个月前 · Python · 80,034 stars15 条评论0 个反应3 负责人
开放[torch.compile] E2E correctness testing for fusionshelp wantedtorch.compilevllm-project/vllm #39,428 创建于 2个月前 · Python · 80,034 stars6 条评论0 个反应0 负责人
开放[Bug]: Certain Ranks Take a Look Time to Load Weightsbughelp wantedvllm-project/vllm #39,030 创建于 2个月前 · Python · 80,034 stars3 条评论0 个反应0 负责人
开放[Transformers v5] Tarsier2ForConditionalGenerationgood first issuehelp wantedvllm-project/vllm #38,736 创建于 2个月前 · Python · 80,034 stars3 条评论0 个反应0 负责人
开放[Transformers v5] SarvamMLAForCausalLMgood first issuehelp wantedvllm-project/vllm #38,734 创建于 2个月前 · Python · 80,034 stars2 条评论0 个反应1 负责人
开放[Transformers v5] InternVL2good first issuehelp wantedvllm-project/vllm #38,425 创建于 2个月前 · Python · 80,034 stars4 条评论0 个反应0 负责人
开放[Transformers v5] IsaacForConditionalGenerationgood first issuehelp wantedvllm-project/vllm #38,389 创建于 2个月前 · Python · 80,034 stars4 条评论0 个反应0 负责人
开放[Transformers v5] Base model and LoRA used in test has incorrect `tokenizer_config.json`good first issuehelp wantedvllm-project/vllm #38,386 创建于 2个月前 · Python · 80,034 stars8 条评论0 个反应1 负责人
开放[Transformers v5] MiniCPMV cannot apply processorgood first issuehelp wantedvllm-project/vllm #38,385 创建于 2个月前 · Python · 80,034 stars8 条评论0 个反应1 负责人
开放Upgrade to Transformers v5help wantedvllm-project/vllm #38,379 创建于 2个月前 · Python · 80,034 stars1 条评论10 个反应1 负责人
开放[Feature]: Better Flashinfer compilation loggingfeature requesthelp wantedvllm-project/vllm #38,246 创建于 2个月前 · Python · 80,034 stars8 条评论0 个反应0 负责人
开放[RFC]: Support ViT Full CUDA Graph (Tracker)RFChelp wantedmulti-modalityvllm-project/vllm #38,175 创建于 2个月前 · Python · 80,034 stars14 条评论1 个反应0 负责人
开放[Feature]: Unify MoE "Oracles" with Class Structurefeature requestgood first issuehelp wantedvllm-project/vllm #37,753 创建于 2个月前 · Python · 80,034 stars6 条评论0 个反应1 负责人
开放[Feature]: Upstream DGX spark improvements from Avarok-Cybersecurity/dgx-vllmfeature requesthelp wantednvidiaquantizationvllm-project/vllm #37,141 创建于 2个月前 · Python · 80,034 stars13 条评论1 个反应0 负责人
开放[Performance]: qknorm+rope fusion slower than unfused on H100help wantedperformancetorch.compilevllm-project/vllm #34,391 创建于 3个月前 · Python · 80,034 stars12 条评论1 个反应1 负责人
开放[Roadmap]: PD Disaggregation with `NixlConnector` Roadmapfeature requesthelp wantedvllm-project/vllm #33,702 创建于 4个月前 · Python · 80,034 stars5 条评论15 个反应0 负责人
开放[Usage]: How to get query embeddings from ColBERT?good first issueusagevllm-project/vllm #42,234 创建于 16天前 · Python · 80,034 stars9 条评论0 个反应1 负责人
开放[Docs] Document NIXL KV connector metrics aggregation semanticsgood first issuevllm-project/vllm #41,230 创建于 27天前 · Python · 80,034 stars4 条评论1 个反应1 负责人
开放[Feature]: Integrate fused `kMoEFinalizeARResidualRMSNorm` from FlashInferfeature requesthelp wantedvllm-project/vllm #40,544 创建于 上个月 · Python · 80,034 stars3 条评论1 个反应0 负责人
开放[Feature]: Priority scheduling supports preemption of requests in the running queue by requests in the waiting queuefeature requesthelp wantedvllm-project/vllm #40,004 创建于 上个月 · Python · 80,034 stars5 条评论0 个反应0 负责人
开放[torch.compile] config hashing refactor follow-upsfeature requestgood first issuehelp wantedvllm-project/vllm #39,479 创建于 2个月前 · Python · 80,034 stars15 条评论0 个反应3 负责人
开放[torch.compile] E2E correctness testing for fusionshelp wantedtorch.compilevllm-project/vllm #39,428 创建于 2个月前 · Python · 80,034 stars6 条评论0 个反应0 负责人
开放[Bug]: Certain Ranks Take a Look Time to Load Weightsbughelp wantedvllm-project/vllm #39,030 创建于 2个月前 · Python · 80,034 stars3 条评论0 个反应0 负责人
开放[Transformers v5] Tarsier2ForConditionalGenerationgood first issuehelp wantedvllm-project/vllm #38,736 创建于 2个月前 · Python · 80,034 stars3 条评论0 个反应0 负责人
开放[Transformers v5] SarvamMLAForCausalLMgood first issuehelp wantedvllm-project/vllm #38,734 创建于 2个月前 · Python · 80,034 stars2 条评论0 个反应1 负责人
开放[Transformers v5] InternVL2good first issuehelp wantedvllm-project/vllm #38,425 创建于 2个月前 · Python · 80,034 stars4 条评论0 个反应0 负责人
开放[Transformers v5] IsaacForConditionalGenerationgood first issuehelp wantedvllm-project/vllm #38,389 创建于 2个月前 · Python · 80,034 stars4 条评论0 个反应0 负责人
开放[Transformers v5] Base model and LoRA used in test has incorrect `tokenizer_config.json`good first issuehelp wantedvllm-project/vllm #38,386 创建于 2个月前 · Python · 80,034 stars8 条评论0 个反应1 负责人
开放[Transformers v5] MiniCPMV cannot apply processorgood first issuehelp wantedvllm-project/vllm #38,385 创建于 2个月前 · Python · 80,034 stars8 条评论0 个反应1 负责人
开放Upgrade to Transformers v5help wantedvllm-project/vllm #38,379 创建于 2个月前 · Python · 80,034 stars1 条评论10 个反应1 负责人
开放[Feature]: Better Flashinfer compilation loggingfeature requesthelp wantedvllm-project/vllm #38,246 创建于 2个月前 · Python · 80,034 stars8 条评论0 个反应0 负责人
开放[RFC]: Support ViT Full CUDA Graph (Tracker)RFChelp wantedmulti-modalityvllm-project/vllm #38,175 创建于 2个月前 · Python · 80,034 stars14 条评论1 个反应0 负责人
开放[Feature]: Unify MoE "Oracles" with Class Structurefeature requestgood first issuehelp wantedvllm-project/vllm #37,753 创建于 2个月前 · Python · 80,034 stars6 条评论0 个反应1 负责人
开放[Feature]: Upstream DGX spark improvements from Avarok-Cybersecurity/dgx-vllmfeature requesthelp wantednvidiaquantizationvllm-project/vllm #37,141 创建于 2个月前 · Python · 80,034 stars13 条评论1 个反应0 负责人
开放[Performance]: qknorm+rope fusion slower than unfused on H100help wantedperformancetorch.compilevllm-project/vllm #34,391 创建于 3个月前 · Python · 80,034 stars12 条评论1 个反应1 负责人
开放[Roadmap]: PD Disaggregation with `NixlConnector` Roadmapfeature requesthelp wantedvllm-project/vllm #33,702 创建于 4个月前 · Python · 80,034 stars5 条评论15 个反应0 负责人