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 13, 04:30 UTC
[ROCm][CI/Build] Upgrade to ROCm 7.1 and AITER main (#28753)
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Gregory Shtrasberg committed
Nov 14, 2025 at 18:53 UTC
75f01b9d3c3a40e52e2fa4a2c9efc92cf45a88fc
1 file changed
+5
-2
docker/Dockerfile.rocm_base
+5
-2
@@ -1,4 +1,4 @@
1
-ARG BASE_IMAGE=rocm/dev-ubuntu-22.04:7.0-complete
1
+ARG BASE_IMAGE=rocm/dev-ubuntu-22.04:7.1-complete
2
ARG TRITON_BRANCH="57c693b6"
3
ARG TRITON_REPO="https://github.com/ROCm/triton.git"
4
ARG PYTORCH_BRANCH="1c57644d"
@@ -7,7 +7,7 @@ ARG PYTORCH_REPO="https://github.com/ROCm/pytorch.git"
7
ARG PYTORCH_VISION_REPO="https://github.com/pytorch/vision.git"
8
ARG FA_BRANCH="0e60e394"
9
ARG FA_REPO="https://github.com/Dao-AILab/flash-attention.git"
10
-ARG AITER_BRANCH="9716b1b8"
10
+ARG AITER_BRANCH="59bd8ff2"
11
ARG AITER_REPO="https://github.com/ROCm/aiter.git"
12
13
FROM ${BASE_IMAGE} AS base
@@ -19,6 +19,9 @@ ARG PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201;gfx11
19
ENV PYTORCH_ROCM_ARCH=${PYTORCH_ROCM_ARCH}
20
ENV AITER_ROCM_ARCH=gfx942;gfx950
21
22
+# Required for RCCL in ROCm7.1
23
+ENV HSA_NO_SCRATCH_RECLAIM=1
24
+
25
ARG PYTHON_VERSION=3.12
26
27
RUN mkdir -p /app