LLM Digest
Subscribe

Story

vllm_releases · Jul 14, 2026 · release

Source brief

vllm v0.25.1

github.comJul 14, 2026
original source linked

Release highlights

  • Avoid blocking model launching when no system FFmpeg is available for TorchCodec . Previously import torchcodec raised a RuntimeError at import time when sys...
  • Guard mixed-dtype allreduce RMSNorm quant fusions . The fused FlashInfer allreduce + RMSNorm + static-quantization patterns could match graphs where the acti...
  • @hugo-cen made their first contribution in #48330

Continue reading

Read the original at github.com →Open in live feedRead that day’s brief

Earlier in this thread 4 items