From 4063db33ffa4fdac7b9b4e485e3d293753aabfb7 Mon Sep 17 00:00:00 2001 From: tzhouam Date: Fri, 3 Jul 2026 06:32:10 +0000 Subject: [PATCH] Rebase adaptation: sync Dockerfile.ci/deps with upstream vLLM before rebase --- docker/Dockerfile.ci | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/docker/Dockerfile.ci b/docker/Dockerfile.ci index 47464426b..70848371c 100644 --- a/docker/Dockerfile.ci +++ b/docker/Dockerfile.ci @@ -20,7 +20,7 @@ RUN apt-get update && \ # (e.g. +gaeee7ef93.cu130) which pip/uv refuse to install from a PEP 503 package index. # The cu130 wheel index lives at //cu130/vllm/ but the actual wheel files # are stored at the top level //. -ENV VLLM_PRECOMPILED_WHEEL_COMMIT=ee0da84ab9e04ac7610e28580af62c365e898389 +ENV VLLM_PRECOMPILED_WHEEL_COMMIT=2dfaae752b4db0d43cfc0715c780e33be030d0f1 # Ensure CUDA library path is in LD_LIBRARY_PATH so bitsandbytes' precompiled # C extension can find libnvJitLink.so.13 (CUDA 13.x JIT linker) at import time. # The base vLLM image has bitsandbytes pre-installed; without this env var, @@ -42,12 +42,12 @@ RUN VLLM_WHEEL_URL=$(python3 -c "import urllib.request,re; \ RUN uv pip install --system ".[dev]" RUN uv pip install --system --upgrade \ - "flashinfer-python==0.6.12" \ - "flashinfer-cubin==0.6.12" \ + "flashinfer-python==0.6.13" \ + "flashinfer-cubin==0.6.13" \ "numpy==2.2.6" RUN uv pip install --system --upgrade \ - "flashinfer-jit-cache==0.6.12" \ + "flashinfer-jit-cache==0.6.13" \ --index-url https://flashinfer.ai/whl/cu130 # Force-reinstall torch AFTER flashinfer, because flashinfer --upgrade pulls in