feat(ml): rocm (#16613)

* feat(ml): introduce support of onnxruntime-rocm for AMD GPU

* try mutex for algo cache

use OrtMutex

* bump versions, run on mich

use 3.12

use 1.19.2

* acquire lock before any changes can be made

guard algo benchmark results

mark mutex as mutable

re-add /bin/sh (?)

use 3.10

use 6.1.2

* use composite cache key

1.19.2

fix variable name

fix variable reference

aaaaaaaaaaaaaaaaaaaa

* bump deps

* disable algo caching

* fix gha

* try ubuntu runner

* actually fix the gha

* update patch

* skip mimalloc preload for rocm

* increase build threads

* increase timeout for rocm

* Revert "increase timeout for rocm"

This reverts commit 2c4452f5d132198ed381a7b262b4a5cab5114b5f.

* attempt migraphx

* set migraphx_home

* Revert "set migraphx_home"

This reverts commit c121d3e48754b3bce100636f8d666deec58a44b7.

* Revert "attempt migraphx"

This reverts commit 521f9fb72dbe506dc6cb8faeb6494817d87265c6.

* migraphx, take two

* bump rocm

* allow cpu

* try only targeting migraphx

* skip tests

* migraph ❌

* known issues

* target gfx900 and gfx1102

* mention `HSA_USE_SVM`

* update lock

* set device id for rocm

---------

Co-authored-by: Mehdi GHESH <mehdi.ghesh@hotmail.fr>

machine-learning/Dockerfile

@@ -17,6 +17,34 @@ RUN mkdir /opt/armnn && \
 
 FROM builder-cpu AS builder-rknn
 
+# Warning: 25GiB+ disk space required to pull this image
+# TODO: find a way to reduce the image size
+FROM rocm/dev-ubuntu-22.04:6.3.4-complete AS builder-rocm
+
+WORKDIR /code
+
+RUN apt-get update && apt-get install -y --no-install-recommends wget git python3.10-venv
+RUN wget -nv https://github.com/Kitware/CMake/releases/download/v3.30.1/cmake-3.30.1-linux-x86_64.sh && \
+    chmod +x cmake-3.30.1-linux-x86_64.sh && \
+    mkdir -p /code/cmake-3.30.1-linux-x86_64 && \
+    ./cmake-3.30.1-linux-x86_64.sh --skip-license --prefix=/code/cmake-3.30.1-linux-x86_64 && \
+    rm cmake-3.30.1-linux-x86_64.sh
+
+ENV PATH=/code/cmake-3.30.1-linux-x86_64/bin:${PATH}
+
+RUN git clone --single-branch --branch v1.20.1 --recursive "https://github.com/Microsoft/onnxruntime" onnxruntime
+WORKDIR /code/onnxruntime
+# Fix for multi-threading based on comments in https://github.com/microsoft/onnxruntime/pull/19567
+# TODO: find a way to fix this without disabling algo caching
+COPY ./patches/* /tmp/
+RUN git apply /tmp/*.patch
+
+RUN /bin/sh ./dockerfiles/scripts/install_common_deps.sh
+# Note: the `parallel` setting uses a substantial amount of RAM
+RUN ./build.sh --allow_running_as_root --config Release --build_wheel --update --build --parallel 17 --cmake_extra_defines\
+    ONNXRUNTIME_VERSION=1.20.1 --skip_tests --use_rocm --rocm_home=/opt/rocm
+RUN mv /code/onnxruntime/build/Linux/Release/dist/*.whl /opt/
+
 FROM builder-${DEVICE} AS builder
 
 ARG DEVICE
@@ -32,6 +60,9 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     --mount=type=bind,source=uv.lock,target=uv.lock \
     --mount=type=bind,source=pyproject.toml,target=pyproject.toml \
     uv sync --frozen --extra ${DEVICE} --no-dev --no-editable --no-install-project --compile-bytecode --no-progress --active --link-mode copy
+RUN if [ "$DEVICE" = "rocm" ]; then \
+    uv pip install /opt/onnxruntime_rocm-*.whl; \
+    fi
 
 FROM python:3.11-slim-bookworm@sha256:614c8691ab74150465ec9123378cd4dde7a6e57be9e558c3108df40664667a4c AS prod-cpu
 
@@ -39,10 +70,10 @@ FROM prod-cpu AS prod-openvino
 
 RUN apt-get update && \
     apt-get install --no-install-recommends -yqq ocl-icd-libopencl1 wget && \
-    wget https://github.com/intel/intel-graphics-compiler/releases/download/igc-1.0.17384.11/intel-igc-core_1.0.17384.11_amd64.deb && \
-    wget https://github.com/intel/intel-graphics-compiler/releases/download/igc-1.0.17384.11/intel-igc-opencl_1.0.17384.11_amd64.deb && \
-    wget https://github.com/intel/compute-runtime/releases/download/24.31.30508.7/intel-opencl-icd_24.31.30508.7_amd64.deb && \
-    wget https://github.com/intel/compute-runtime/releases/download/24.31.30508.7/libigdgmm12_22.4.1_amd64.deb && \
+    wget -nv https://github.com/intel/intel-graphics-compiler/releases/download/igc-1.0.17384.11/intel-igc-core_1.0.17384.11_amd64.deb && \
+    wget -nv https://github.com/intel/intel-graphics-compiler/releases/download/igc-1.0.17384.11/intel-igc-opencl_1.0.17384.11_amd64.deb && \
+    wget -nv https://github.com/intel/compute-runtime/releases/download/24.31.30508.7/intel-opencl-icd_24.31.30508.7_amd64.deb && \
+    wget -nv https://github.com/intel/compute-runtime/releases/download/24.31.30508.7/libigdgmm12_22.4.1_amd64.deb && \
     dpkg -i *.deb && \
     rm *.deb && \
     apt-get remove wget -yqq && \
@@ -59,6 +90,8 @@ COPY --from=builder-cuda /usr/local/bin/python3 /usr/local/bin/python3
 COPY --from=builder-cuda /usr/local/lib/python3.11 /usr/local/lib/python3.11
 COPY --from=builder-cuda /usr/local/lib/libpython3.11.so /usr/local/lib/libpython3.11.so
 
+FROM rocm/dev-ubuntu-22.04:6.3.4-complete AS prod-rocm
+
 FROM prod-cpu AS prod-armnn
 
 ENV LD_LIBRARY_PATH=/opt/armnn
@@ -81,13 +114,12 @@ COPY --from=builder-armnn \
 
 FROM prod-cpu AS prod-rknn
 
-ADD --checksum=sha256:73993ed4b440460825f21611731564503cc1d5a0c123746477da6cd574f34885 https://github.com/airockchip/rknn-toolkit2/raw/refs/tags/v2.3.0/rknpu2/runtime/Linux/librknn_api/aarch64/librknnrt.so /usr/lib/
-
 FROM prod-${DEVICE} AS prod
+
 ARG DEVICE
 
 RUN apt-get update && \
-    apt-get install -y --no-install-recommends tini $(if ! [ "$DEVICE" = "openvino" ]; then echo "libmimalloc2.0"; fi) && \
+    apt-get install -y --no-install-recommends tini $(if ! [ "$DEVICE" = "openvino" ] && ! [ "$DEVICE" = "rocm" ]; then echo "libmimalloc2.0"; fi) && \
     apt-get autoremove -yqq && \
     apt-get clean && \
     rm -rf /var/lib/apt/lists/*