Skip to content

Commit a98fda0

Browse files
author
Mark Saroufim
committed
Reapply "Update AMD Dockerfile: ROCm 7.2, latest aiter, remove multi-GPU deps"
This reverts commit e09a2cd.
1 parent b3dfd32 commit a98fda0

1 file changed

Lines changed: 7 additions & 63 deletions

File tree

docker/amd-docker.Dockerfile

Lines changed: 7 additions & 63 deletions
Original file line numberDiff line numberDiff line change
@@ -1,8 +1,6 @@
11
FROM ghcr.io/actions/actions-runner:latest
22

33
ENV CXX=clang++
4-
ENV UCX_CXX=g++
5-
ENV UCX_CC=gcc
64

75
RUN sudo apt-get update -y \
86
&& sudo apt-get install -y software-properties-common \
@@ -35,18 +33,20 @@ RUN sudo groupadd -g 109 render
3533
RUN sudo apt update -y \
3634
&& sudo apt install -y "linux-headers-$(uname -r)" "linux-modules-extra-$(uname -r)" \
3735
&& sudo usermod -a -G render,video runner \
38-
&& wget https://repo.radeon.com/amdgpu-install/6.3.1/ubuntu/jammy/amdgpu-install_6.3.60301-1_all.deb \
39-
&& sudo apt install -y ./amdgpu-install_6.3.60301-1_all.deb \
36+
&& wget https://repo.radeon.com/amdgpu-install/7.2/ubuntu/jammy/amdgpu-install_7.2.70200-1_all.deb \
37+
&& sudo apt install -y ./amdgpu-install_7.2.70200-1_all.deb \
4038
&& sudo apt update -y \
4139
&& sudo apt install -y rocm
4240

41+
ENV ROCM_PATH=/opt/rocm
42+
4343
RUN sudo pip install --upgrade pip
4444

45-
RUN sudo pip install --no-cache-dir torch==2.10.0.dev20250916+rocm6.3 pytorch-triton-rocm --index-url https://download.pytorch.org/whl/nightly/rocm6.3
45+
RUN sudo pip install --no-cache-dir torch --index-url https://download.pytorch.org/whl/nightly/rocm7.2
4646

4747
RUN git clone --recursive https://github.com/ROCm/aiter.git \
4848
&& cd aiter \
49-
&& git checkout 1d88633958236e942cba3c283864282f7af3ebc5 \
49+
&& git checkout f3be04a12a0cfd6b5e2c7a94edc774f1bc24460d \
5050
&& sudo pip install -r requirements.txt \
5151
&& sudo python3 setup.py develop
5252

@@ -62,60 +62,4 @@ RUN sudo pip install \
6262

6363
RUN sudo pip install git+https://github.com/ROCm/iris.git
6464

65-
RUN sudo apt-get update -y \
66-
&& sudo apt-get install -y --no-install-recommends \
67-
autoconf \
68-
automake \
69-
libtool \
70-
pkg-config \
71-
build-essential \
72-
gfortran \
73-
flex \
74-
bison \
75-
libomp-dev \
76-
libhwloc-dev \
77-
libnuma-dev \
78-
&& sudo rm -rf /var/lib/apt/lists/*
79-
80-
ENV UCX_INSTALL_DIR=/opt/ucx
81-
ENV OMPI_INSTALL_DIR=/opt/openmpi
82-
ENV ROCSHMEM_INSTALL_DIR=/opt/rocshmem
83-
ENV ROCM_PATH=/opt/rocm
84-
85-
RUN cd /tmp \
86-
&& git clone https://github.com/openucx/ucx.git -b v1.17.x \
87-
&& cd ucx \
88-
&& ./autogen.sh \
89-
&& CC=gcc CXX=g++ ./configure --prefix=${UCX_INSTALL_DIR} --with-rocm=${ROCM_PATH} --enable-mt --disable-optimizations \
90-
&& make -j$(nproc) \
91-
&& sudo make install \
92-
&& cd / \
93-
&& sudo rm -rf /tmp/ucx
94-
95-
RUN cd /tmp \
96-
&& git clone --recursive https://github.com/open-mpi/ompi.git -b v5.0.x \
97-
&& cd ompi \
98-
&& ./autogen.pl \
99-
&& ./configure --prefix=${OMPI_INSTALL_DIR} --with-rocm=${ROCM_PATH} --with-ucx=${UCX_INSTALL_DIR} \
100-
&& make -j$(nproc) \
101-
&& sudo make install \
102-
&& cd / \
103-
&& sudo rm -rf /tmp/ompi
104-
105-
ENV PATH="${OMPI_INSTALL_DIR}/bin:${PATH}"
106-
ENV LD_LIBRARY_PATH="${OMPI_INSTALL_DIR}/lib:${UCX_INSTALL_DIR}/lib:/opt/rocm/lib"
107-
108-
109-
RUN cd /tmp \
110-
&& git clone https://github.com/ROCm/rocSHMEM.git \
111-
&& cd rocSHMEM \
112-
&& mkdir build \
113-
&& cd build \
114-
&& MPI_ROOT=${OMPI_INSTALL_DIR} UCX_ROOT=${UCX_INSTALL_DIR} CMAKE_PREFIX_PATH="${ROCM_PATH}:$CMAKE_PREFIX_PATH" \
115-
sudo ../scripts/build_configs/ipc_single -DCMAKE_INSTALL_PREFIX=/opt/rocshmem \
116-
&& cd / \
117-
&& sudo rm -rf /tmp/rocSHMEM
118-
119-
120-
ENV ROCSHMEM_INSTALL_DIR=${ROCSHMEM_INSTALL_DIR}
121-
ENV LD_LIBRARY_PATH="${ROCSHMEM_INSTALL_DIR}/lib:${LD_LIBRARY_PATH}"
65+
ENV LD_LIBRARY_PATH="/opt/rocm/lib"

0 commit comments

Comments
 (0)