diff --git a/Dockerfile b/Dockerfile index 4258f4f..e0ababb 100644 --- a/Dockerfile +++ b/Dockerfile @@ -335,7 +335,9 @@ RUN --mount=type=cache,target=/root/.cache/uv bash - <<'BASH' TORCH_CUDA_ARCH_LIST="${FI_TORCH_CUDA_ARCH_LIST}" \ uv pip install --system --no-build-isolation . popd - rm -rf flashinfer + rm -rf flashinfer \ + ~/.cache/flashinfer/aot/*/unused* + BASH COPY ./vllm_v0.10.0/examples examples COPY ./vllm_v0.10.0/benchmarks benchmarks