vllm-project/vllm 发布 v0.25.1

中文处理结果

本版本包含来自 2 位贡献者的 2 个提交（其中 1 位新贡献者）！

v0.25.1 是一个补丁版本，在 v0.25.0 基础上包含两个针对性的错误修复。

避免在系统缺少 FFmpeg 时阻止 TorchCodec 模型启动 (#47888)。此前，当系统缺少 FFmpeg 时，import torchcodec 会在导入时抛出 RuntimeError，即使未使用 TorchCodec 也会阻止启动（例如 vllm serve Qwen/Qwen3-VL-2B-Instruct）。该错误现在延迟到运行时出现，仅在真正需要 TorchCodec 时才会触发。
保护混合精度 allreduce RMSNorm 量化融合 (#48330)。融合的 FlashInfer allreduce + RMSNorm + 静态量化模式可能匹配激活与 RMSNorm 权重 dtype 不同的计算图（例如 NVFP4 模型中 BF16 残差流搭配 FP32 Gemma/Qwen 风格 RMSNorm 权重），从而损坏隐藏状态并产生垃圾输出，如重复的 !!!!! token。现在 dtype 匹配保护会将不兼容的混合精度计算图路由到安全路径，而相同 dtype 的模型保留完整的 allreduce + RMSNorm + 量化融合。

@Isotr0py, @hugo-cen

原始正文

vllm-project/vllm released v0.25.1

This release features 2 commits from 2 contributors (1 new)!

v0.25.1 is a patch release containing two targeted bug fixes on top of v0.25.0.

Avoid blocking model launching when no system FFmpeg is available for TorchCodec (#47888). Previously import torchcodec raised a RuntimeError at import time when system FFmpeg was missing, which blocked startup (e.g. vllm serve Qwen/Qwen3-VL-2B-Instruct) even when TorchCodec was not in use. The error is now deferred to runtime so it only surfaces if TorchCodec is actually needed.
Guard mixed-dtype allreduce RMSNorm quant fusions (#48330). The fused FlashInfer allreduce + RMSNorm + static-quantization patterns could match graphs where the activation and RMSNorm weight dtypes differ (e.g. a BF16 residual stream with an FP32 Gemma/Qwen-style RMSNorm weight in NVFP4 models), corrupting the hidden state and producing garbage output such as repeated !!!!! tokens. A dtype-match guard now routes incompatible mixed-dtype graphs to the safe path, while same-dtype models retain the full allreduce + RMSNorm + quant fusion.

@Isotr0py, @hugo-cen