@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / aac622e0cd
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 14, 05:30 UTC

[ROCm][Build] Add support for AMD Ryzen AI MAX / AI 300 Series (#25908)

Signed-off-by: Hosang Yoon <hosang.yoon@amd.com>

Hosang committed Oct 1, 2025 at 17:39 UTC aac622e0cd002ee91263667f27c3f9d23951acd2
3 files changed +5 -4
CMakeLists.txt
+1 -1
@@ -37,7 +37,7 @@ install(CODE "set(CMAKE_INSTALL_LOCAL_ONLY TRUE)" ALL_COMPONENTS)
37 set(PYTHON_SUPPORTED_VERSIONS "3.9" "3.10" "3.11" "3.12" "3.13")
38
39 # Supported AMD GPU architectures.
40 -set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1101;gfx1200;gfx1201")
40 +set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx1151")
41
42 #
43 # Supported/expected torch versions for CUDA/ROCm.
csrc/rocm/attention.cu
+2 -1
@@ -40,7 +40,8 @@ using __hip_fp8_e5m2 = __hip_fp8_e5m2_fnuz;
40 #define __HIP__FP8MFMA__
41 #endif
42
43 -#if defined(__HIPCC__) && (defined(__gfx1100__) || defined(__gfx1101__))
43 +#if defined(__HIPCC__) && (defined(__gfx1100__) || defined(__gfx1101__) || \
44 + defined(__gfx1150__) || defined(__gfx1151__))
45 #define __HIP__GFX11__
46 #endif
47
docker/Dockerfile.rocm_base
+2 -2
@@ -15,7 +15,7 @@ FROM ${BASE_IMAGE} AS base
15 ENV PATH=/opt/rocm/llvm/bin:/opt/rocm/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
16 ENV ROCM_PATH=/opt/rocm
17 ENV LD_LIBRARY_PATH=/opt/rocm/lib:/usr/local/lib:
18 -ARG PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201
18 +ARG PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx1151
19 ENV PYTORCH_ROCM_ARCH=${PYTORCH_ROCM_ARCH}
20 ENV AITER_ROCM_ARCH=gfx942;gfx950
21
@@ -141,4 +141,4 @@ RUN echo "BASE_IMAGE: ${BASE_IMAGE}" > /app/versions.txt \
141 && echo "FA_BRANCH: ${FA_BRANCH}" >> /app/versions.txt \
142 && echo "FA_REPO: ${FA_REPO}" >> /app/versions.txt \
143 && echo "AITER_BRANCH: ${AITER_BRANCH}" >> /app/versions.txt \
144 - && echo "AITER_REPO: ${AITER_REPO}" >> /app/versions.txt
\ No newline at end of file
144 + && echo "AITER_REPO: ${AITER_REPO}" >> /app/versions.txt