Restore CUDA 13.0.1 + patch vLLM for cuMemcpyBatchAsync API change

CUDA 13 removed the fail_idx parameter from cuMemcpyBatchAsync.
Patch cache_kernels.cu to match new API signature instead of downgrading.

- Restore CUDA 13.0.1, PyTorch 2.9.0+cu130, flashinfer cu130
- Patch: remove fail_idx variable and parameter from cuMemcpyBatchAsync call
- Simplify error message to not reference fail_idx

vllm/Dockerfile

@@ -81,7 +81,7 @@ RUN mkdir -p /wheels && \
 FROM build-base AS build-flashinfer
 ARG FLASHINFER_ENABLE_AOT=1
 ARG FLASHINFER_REF=v0.6.6
-ARG FLASHINFER_BUILD_SUFFIX=cu128
+ARG FLASHINFER_BUILD_SUFFIX=cu130
 ENV FLASHINFER_LOCAL_VERSION=${FLASHINFER_BUILD_SUFFIX:-}
 RUN git clone https://github.com/flashinfer-ai/flashinfer.git
 RUN cd flashinfer && \
@@ -134,6 +134,9 @@ RUN cd vllm && \
     git submodule sync && \
     git submodule update --init --recursive -j 8 && \
     sed -i 's/GIT_TAG [a-f0-9]\{40\}/GIT_TAG main/' cmake/external_projects/vllm_flash_attn.cmake && \
+    sed -i '/size_t fail_idx = 0;/d' csrc/cache_kernels.cu && \
+    sed -i 's/, \&fail_idx,/,/' csrc/cache_kernels.cu && \
+    sed -i 's/"cuMemcpyBatchAsync failed at index ",\s*fail_idx, " with error "/"cuMemcpyBatchAsync failed with error "/' csrc/cache_kernels.cu && \
     export MAX_JOBS=8 && \
     export CMAKE_BUILD_PARALLEL_LEVEL=$MAX_JOBS && \
     python use_existing_torch.py && \