Search results

[Bug]: FastIncrementalDetokenizer leaks full prompt text into streamed output when prompt ends with an incomplete UTF-8 token

Status: Open.

#48853 In vllm-project/vllm;

· spped2000 opened on Jul 16, 2026

[Bug]: Gemma4 MTP speculative decoding crashes at engine init on 0.25.1 — "a and b must have same reduction dim" (regression from 0.21.0)

Status: Open.

#48848 In vllm-project/vllm;

· pryingopen3rdi opened on Jul 16, 2026

[Bug]: AssertionError: graph_pool_id is not set under graph capture when VLLM_USE_NCCL_SYMM_MEM=1 with ModelRunner V2

bugSomething isn't workingSomething isn't working

Status: Open.

#48840 In vllm-project/vllm;

· ilmarkov opened on Jul 16, 2026

[Bug]: M-RoPE recomputation crashes or mis-associates media at a chunked-prefill boundary

bugSomething isn't workingSomething isn't working

Status: Open.

#48833 In vllm-project/vllm;

· parkilwoo opened on Jul 16, 2026

[Bug]: V1 Engine Produces Incorrect Scores for Qwen3-Reranker-0.6B on Long Sequences (>8K tokens)

bugSomething isn't workingSomething isn't working

Status: Open.

#48831 In vllm-project/vllm;

· yuejiyueren opened on Jul 16, 2026

[CI Failure]: Repo-wide job failures from HuggingFace Hub 504 outage (model downloads)

Status: Open.

#48830 In vllm-project/vllm;

· Change72 opened on Jul 16, 2026

[Bug]: enable_multithread_load silently disables EP weight filtering (local_expert_ids never reaches multi_thread_safetensors_weights_iterator)

Status: Open.

#48827 In vllm-project/vllm;

· yingguo-trt opened on Jul 16, 2026

[Feature]: Migration plan for ROCm 7.14 / TheRock production release (containers + wheels)

rocmRelated to AMD ROCmRelated to AMD ROCm

Status: Open.

#48826 In vllm-project/vllm;

· raravind007 opened on Jul 16, 2026

[Bug]: Failed to initialize FlashInfer All Reduce workspace: CUDA driver error: invalid device ordinal

bugSomething isn't workingSomething isn't working

Status: Open.

#48823 In vllm-project/vllm;

· timtimyim opened on Jul 16, 2026

[Bug] Optional Transformers model submodules can break vLLM model imports

Status: Open.

#48819 In vllm-project/vllm;

· host452b opened on Jul 16, 2026

[Bug]: The Deepseek v4 model compiles kenrel during the inference process

bugSomething isn't workingSomething isn't working

Status: Open.

#48809 In vllm-project/vllm;

· leihuang-sketch opened on Jul 16, 2026

[Bug]: DP request distribution becomes imbalanced under long-context workload on H20 GPU

bugSomething isn't workingSomething isn't working

Status: Open.

#48808 In vllm-project/vllm;

· Cancan0320 opened on Jul 16, 2026