Search results

Implement support for Qwen3.5

Status: Open.

#206 In GeeeekExplorer/nano-vllm;

· Wangbiao2 opened on Apr 20, 2026

CUDA graph replay can fail with block_tables shape mismatch

Status: Open.

#190 In GeeeekExplorer/nano-vllm;

· ilrewrite opened on Mar 24, 2026

Qwen models initialization fails with TypeError: unhashable type: 'dict' and AssertionError due to rope_scaling

Status: Open.

#189 In GeeeekExplorer/nano-vllm;

· junjue123 opened on Mar 23, 2026

Potential TP correctness issue: BlockManager capacity may rely on per-rank local KV-cache estimation

Status: Open.

#187 In GeeeekExplorer/nano-vllm;

· lllloowwoollll opened on Mar 20, 2026

5080 Laptop failed to going on😭😭😭

Status: Open.

#181 In GeeeekExplorer/nano-vllm;

· TeloySXH opened on Mar 8, 2026

Documentation: Outdated Hugging Face download command (--resume-download removed in new hf CLI)

Status: Open.

#179 In GeeeekExplorer/nano-vllm;

· junjue123 opened on Mar 6, 2026

Perf: CUDA graph decode still allocates tensors per step

Status: Open.

#175 In GeeeekExplorer/nano-vllm;

· MrAnayDongre opened on Feb 22, 2026

RMSNorm rms forward has a bug in the float32 case where the input matrix is modified, resulting in the incorrect modification of the residual connection values

Status: Open.

#170 In GeeeekExplorer/nano-vllm;

· philhuan opened on Feb 8, 2026

rope argument inconsistent

Status: Open.

#167 In GeeeekExplorer/nano-vllm;

· yfcAzul opened on Jan 28, 2026

🚀 Introducing Nano-vLLM-v1 – A Lightweight, Re-engineered Nano vLLM Implementation with vLLM v1 Scheduler and Chunked Prefill

Status: Open.

#165 In GeeeekExplorer/nano-vllm;

· slwang-ustc opened on Jan 25, 2026

Stale hash mapping remains in hash_to_block_id when reallocating blocks in BlockManager [Bug]

Status: Open.

#163 In GeeeekExplorer/nano-vllm;

· slwang-ustc opened on Jan 20, 2026

Two bugs

Status: Open.

#144 In GeeeekExplorer/nano-vllm;

· LiaoMengqi opened on Dec 17, 2025