jetbrains bot thinks we should do all this

Author: jiridanek

ci/cached-builds/gha_lvm_overlay.sh

@@ -8,9 +8,9 @@ set -Eeuo pipefail
 # https://github.com/easimon/maximize-build-space/blob/master/action.yml
 
 # root_reserve_mb=2048 was running out of disk space building cuda images
-root_reserve_mb=4096
+root_reserve_mb=2048
 temp_reserve_mb=100
-swap_size_mb=4096
+swap_size_mb=2048
 
 build_mount_path="${HOME}/.local/share/containers"
 build_mount_path_ownership="runner:runner"

jupyter/minimal/ubi9-python-3.11/Dockerfile.rocm

@@ -38,7 +38,7 @@ ARG AMDGPU_VERSION=6.2.4
 # Install the ROCm rpms
 # ref: https://github.com/ROCm/ROCm-docker/blob/master/dev/Dockerfile-centos-7-complete
 # Note: Based on 6.2 above new package mivisionx is a pre-requistes, which bring in more dependent packages
-# so we are only installing meta packages of rocm 
+# so we are only installing meta packages of rocm
 # ref: https://rocm.docs.amd.com/projects/install-on-linux/en/develop/reference/package-manager-integration.html#packages-in-rocm-programming-models
 RUN echo "[ROCm]" > /etc/yum.repos.d/rocm.repo && \
     echo "name=ROCm" >> /etc/yum.repos.d/rocm.repo && \
@@ -50,8 +50,21 @@ RUN echo "[ROCm]" > /etc/yum.repos.d/rocm.repo && \
     echo "baseurl=https://repo.radeon.com/amdgpu/$AMDGPU_VERSION/rhel/9.4/main/x86_64" >> /etc/yum.repos.d/amdgpu.repo && \
     echo "enabled=1" >> /etc/yum.repos.d/amdgpu.repo && \
     echo "gpgcheck=0" >> /etc/yum.repos.d/amdgpu.repo && \
-    yum install -y rocm-developer-tools rocm-ml-sdk rocm-opencl-sdk rocm-openmp-sdk rocm-utils && \
-    yum clean all && rm -rf /var/cache/yum
+    # Install ROCm packages in separate steps to better manage disk space
+    # First install core ROCm components
+    yum install -y --setopt=install_weak_deps=False rocm-core rocm-hip-runtime rocm-hip-sdk && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Then install ML components
+    yum install -y --setopt=install_weak_deps=False rocm-ml-sdk && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Finally install remaining components
+    yum install -y --setopt=install_weak_deps=False rocm-opencl-sdk rocm-utils && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Remove unnecessary files to save space
+    find /opt/rocm-$ROCM_VERSION -name "*.a" -delete && \
+    find /opt/rocm-$ROCM_VERSION -name "*.o" -delete && \
+    find /opt/rocm-$ROCM_VERSION -type d -name "samples" -exec rm -rf {} + 2>/dev/null || true && \
+    find /opt/rocm-$ROCM_VERSION -type d -name "doc" -exec rm -rf {} + 2>/dev/null || true
 
 # Restore notebook user workspace
 USER 1001
@@ -60,7 +73,7 @@ WORKDIR /opt/app-root/src
 ########################
 # rocm-jupyter-minimal #
 ########################
-FROM rocm-base AS rocm-jupyter-minimal  
+FROM rocm-base AS rocm-jupyter-minimal
 
 ARG JUPYTER_REUSABLE_UTILS=jupyter/utils
 ARG MINIMAL_SOURCE_CODE=jupyter/minimal/ubi9-python-3.11
@@ -94,7 +107,7 @@ RUN echo "Installing softwares and packages" && \
     fix-permissions /opt/app-root -P && \
     # Apply JupyterLab addons \
     /opt/app-root/bin/utils/addons/apply.sh
-    
+
 WORKDIR /opt/app-root/src
 
-ENTRYPOINT ["start-notebook.sh"]
+ENTRYPOINT ["start-notebook.sh"]

jupyter/rocm/pytorch/ubi9-python-3.11/Dockerfile.rocm

@@ -38,7 +38,7 @@ ARG AMDGPU_VERSION=6.2.4
 # Install the ROCm rpms
 # ref: https://github.com/ROCm/ROCm-docker/blob/master/dev/Dockerfile-centos-7-complete
 # Note: Based on 6.2 above new package mivisionx is a pre-requistes, which bring in more dependent packages
-# so we are only installing meta packages of rocm 
+# so we are only installing meta packages of rocm
 # ref: https://rocm.docs.amd.com/projects/install-on-linux/en/develop/reference/package-manager-integration.html#packages-in-rocm-programming-models
 RUN echo "[ROCm]" > /etc/yum.repos.d/rocm.repo && \
     echo "name=ROCm" >> /etc/yum.repos.d/rocm.repo && \
@@ -50,8 +50,21 @@ RUN echo "[ROCm]" > /etc/yum.repos.d/rocm.repo && \
     echo "baseurl=https://repo.radeon.com/amdgpu/$AMDGPU_VERSION/rhel/9.4/main/x86_64" >> /etc/yum.repos.d/amdgpu.repo && \
     echo "enabled=1" >> /etc/yum.repos.d/amdgpu.repo && \
     echo "gpgcheck=0" >> /etc/yum.repos.d/amdgpu.repo && \
-    yum install -y rocm-developer-tools rocm-ml-sdk rocm-opencl-sdk rocm-openmp-sdk rocm-utils && \
-    yum clean all && rm -rf /var/cache/yum
+    # Install ROCm packages in separate steps to better manage disk space
+    # First install core ROCm components
+    yum install -y --setopt=install_weak_deps=False rocm-core rocm-hip-runtime rocm-hip-sdk && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Then install ML components
+    yum install -y --setopt=install_weak_deps=False rocm-ml-sdk && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Finally install remaining components
+    yum install -y --setopt=install_weak_deps=False rocm-opencl-sdk rocm-utils && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Remove unnecessary files to save space
+    find /opt/rocm-$ROCM_VERSION -name "*.a" -delete && \
+    find /opt/rocm-$ROCM_VERSION -name "*.o" -delete && \
+    find /opt/rocm-$ROCM_VERSION -type d -name "samples" -exec rm -rf {} + 2>/dev/null || true && \
+    find /opt/rocm-$ROCM_VERSION -type d -name "doc" -exec rm -rf {} + 2>/dev/null || true
 
 # Restore notebook user workspace
 USER 1001
@@ -60,7 +73,7 @@ WORKDIR /opt/app-root/src
 ########################
 # rocm-jupyter-minimal #
 ########################
-FROM rocm-base AS rocm-jupyter-minimal  
+FROM rocm-base AS rocm-jupyter-minimal
 
 ARG JUPYTER_REUSABLE_UTILS=jupyter/utils
 ARG MINIMAL_SOURCE_CODE=jupyter/minimal/ubi9-python-3.11
@@ -70,15 +83,15 @@ WORKDIR /opt/app-root/bin
 COPY ${JUPYTER_REUSABLE_UTILS} utils/
 
 COPY ${MINIMAL_SOURCE_CODE}/start-notebook.sh ./
-    
+
 WORKDIR /opt/app-root/src
 
 ENTRYPOINT ["start-notebook.sh"]
 
 ############################
 # rocm-jupyter-datascience #
 ############################
-FROM rocm-jupyter-minimal AS rocm-jupyter-datascience  
+FROM rocm-jupyter-minimal AS rocm-jupyter-datascience
 
 ARG DATASCIENCE_SOURCE_CODE=jupyter/datascience/ubi9-python-3.11
 
@@ -114,7 +127,7 @@ WORKDIR /opt/app-root/src
 ########################
 # rocm-jupyter-pytorch #
 ########################
-FROM rocm-jupyter-datascience AS rocm-jupyter-pytorch  
+FROM rocm-jupyter-datascience AS rocm-jupyter-pytorch
 
 ARG DATASCIENCE_SOURCE_CODE=jupyter/datascience/ubi9-python-3.11
 ARG PYTORCH_SOURCE_CODE=jupyter/rocm/pytorch/ubi9-python-3.11
@@ -135,20 +148,20 @@ COPY ${PYTORCH_SOURCE_CODE}/Pipfile.lock ${PYTORCH_SOURCE_CODE}/de-vendor-torch.
 
 RUN echo "Installing softwares and packages" && \
     micropipenv install && \
-    rm -f ./Pipfile.lock && \    
+    rm -f ./Pipfile.lock && \
     # setup path for runtime configuration
     mkdir /opt/app-root/runtimes && \
     # Remove default Elyra runtime-images \
-    rm /opt/app-root/share/jupyter/metadata/runtime-images/*.json && \    
+    rm /opt/app-root/share/jupyter/metadata/runtime-images/*.json && \
     # Replace Notebook's launcher, "(ipykernel)" with Python's version 3.x.y \
     sed -i -e "s/Python.*/$(python --version | cut -d '.' -f-2)\",/" /opt/app-root/share/jupyter/kernels/python3/kernel.json && \
     # Disable announcement plugin of jupyterlab \
     jupyter labextension disable "@jupyterlab/apputils-extension:announcements" && \
     # Apply JupyterLab addons \
-    /opt/app-root/bin/utils/addons/apply.sh && \    
+    /opt/app-root/bin/utils/addons/apply.sh && \
     # De-vendor the ROCm libs that are embedded in Pytorch \
     ./de-vendor-torch.sh && \
-    rm ./de-vendor-torch.sh && \    
+    rm ./de-vendor-torch.sh && \
     # Fix permissions to support pip in Openshift environments \
     chmod -R g+w /opt/app-root/lib/python3.11/site-packages && \
     fix-permissions /opt/app-root -P
@@ -157,4 +170,4 @@ RUN echo "Installing softwares and packages" && \
 # Copy Elyra runtime-images definitions and set the version
 COPY ${DATASCIENCE_SOURCE_CODE}/runtime-images/ /opt/app-root/share/jupyter/metadata/runtime-images/
 
-WORKDIR /opt/app-root/src
+WORKDIR /opt/app-root/src

jupyter/rocm/tensorflow/ubi9-python-3.11/Dockerfile.rocm

@@ -38,7 +38,7 @@ ARG AMDGPU_VERSION=6.2.4
 # Install the ROCm rpms
 # ref: https://github.com/ROCm/ROCm-docker/blob/master/dev/Dockerfile-centos-7-complete
 # Note: Based on 6.2 above new package mivisionx is a pre-requistes, which bring in more dependent packages
-# so we are only installing meta packages of rocm 
+# so we are only installing meta packages of rocm
 # ref: https://rocm.docs.amd.com/projects/install-on-linux/en/develop/reference/package-manager-integration.html#packages-in-rocm-programming-models
 RUN echo "[ROCm]" > /etc/yum.repos.d/rocm.repo && \
     echo "name=ROCm" >> /etc/yum.repos.d/rocm.repo && \
@@ -50,8 +50,21 @@ RUN echo "[ROCm]" > /etc/yum.repos.d/rocm.repo && \
     echo "baseurl=https://repo.radeon.com/amdgpu/$AMDGPU_VERSION/rhel/9.4/main/x86_64" >> /etc/yum.repos.d/amdgpu.repo && \
     echo "enabled=1" >> /etc/yum.repos.d/amdgpu.repo && \
     echo "gpgcheck=0" >> /etc/yum.repos.d/amdgpu.repo && \
-    yum install -y rocm-developer-tools rocm-ml-sdk rocm-opencl-sdk rocm-openmp-sdk rocm-utils && \
-    yum clean all && rm -rf /var/cache/yum
+    # Install ROCm packages in separate steps to better manage disk space
+    # First install core ROCm components
+    yum install -y --setopt=install_weak_deps=False rocm-core rocm-hip-runtime rocm-hip-sdk && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Then install ML components
+    yum install -y --setopt=install_weak_deps=False rocm-ml-sdk && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Finally install remaining components
+    yum install -y --setopt=install_weak_deps=False rocm-opencl-sdk rocm-utils && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Remove unnecessary files to save space
+    find /opt/rocm-$ROCM_VERSION -name "*.a" -delete && \
+    find /opt/rocm-$ROCM_VERSION -name "*.o" -delete && \
+    find /opt/rocm-$ROCM_VERSION -type d -name "samples" -exec rm -rf {} + 2>/dev/null || true && \
+    find /opt/rocm-$ROCM_VERSION -type d -name "doc" -exec rm -rf {} + 2>/dev/null || true
 
 # Restore notebook user workspace
 USER 1001
@@ -60,7 +73,7 @@ WORKDIR /opt/app-root/src
 ########################
 # rocm-jupyter-minimal #
 ########################
-FROM rocm-base AS rocm-jupyter-minimal  
+FROM rocm-base AS rocm-jupyter-minimal
 
 ARG JUPYTER_REUSABLE_UTILS=jupyter/utils
 ARG MINIMAL_SOURCE_CODE=jupyter/minimal/ubi9-python-3.11
@@ -70,15 +83,15 @@ WORKDIR /opt/app-root/bin
 COPY ${JUPYTER_REUSABLE_UTILS} utils/
 
 COPY ${MINIMAL_SOURCE_CODE}/start-notebook.sh ./
-    
+
 WORKDIR /opt/app-root/src
 
 ENTRYPOINT ["start-notebook.sh"]
 
 ############################
 # rocm-jupyter-datascience #
 ############################
-FROM rocm-jupyter-minimal AS rocm-jupyter-datascience  
+FROM rocm-jupyter-minimal AS rocm-jupyter-datascience
 
 ARG DATASCIENCE_SOURCE_CODE=jupyter/datascience/ubi9-python-3.11
 
@@ -114,7 +127,7 @@ WORKDIR /opt/app-root/src
 ###########################
 # rocm-jupyter-tensorflow #
 ###########################
-FROM rocm-jupyter-datascience AS rocm-jupyter-tensorflow  
+FROM rocm-jupyter-datascience AS rocm-jupyter-tensorflow
 
 ARG DATASCIENCE_SOURCE_CODE=jupyter/datascience/ubi9-python-3.11
 ARG TENSORFLOW_SOURCE_CODE=jupyter/rocm/tensorflow/ubi9-python-3.11
@@ -135,17 +148,17 @@ COPY ${TENSORFLOW_SOURCE_CODE}/Pipfile.lock  ./
 
 RUN echo "Installing softwares and packages" && \
     micropipenv install --dev && \
-    rm -f ./Pipfile.lock && \    
+    rm -f ./Pipfile.lock && \
     # setup path for runtime configuration
     mkdir /opt/app-root/runtimes && \
     # Remove default Elyra runtime-images \
-    rm /opt/app-root/share/jupyter/metadata/runtime-images/*.json && \    
+    rm /opt/app-root/share/jupyter/metadata/runtime-images/*.json && \
     # Replace Notebook's launcher, "(ipykernel)" with Python's version 3.x.y \
     sed -i -e "s/Python.*/$(python --version | cut -d '.' -f-2)\",/" /opt/app-root/share/jupyter/kernels/python3/kernel.json && \
     # Disable announcement plugin of jupyterlab \
     jupyter labextension disable "@jupyterlab/apputils-extension:announcements" && \
     # Apply JupyterLab addons \
-    /opt/app-root/bin/utils/addons/apply.sh && \    
+    /opt/app-root/bin/utils/addons/apply.sh && \
     # Fix permissions to support pip in Openshift environments \
     chmod -R g+w /opt/app-root/lib/python3.11/site-packages && \
     fix-permissions /opt/app-root -P

runtimes/rocm-pytorch/ubi9-python-3.11/Dockerfile.rocm

@@ -38,7 +38,7 @@ ARG AMDGPU_VERSION=6.2.4
 # Install the ROCm rpms
 # ref: https://github.com/ROCm/ROCm-docker/blob/master/dev/Dockerfile-centos-7-complete
 # Note: Based on 6.2 above new package mivisionx is a pre-requistes, which bring in more dependent packages
-# so we are only installing meta packages of rocm 
+# so we are only installing meta packages of rocm
 # ref: https://rocm.docs.amd.com/projects/install-on-linux/en/develop/reference/package-manager-integration.html#packages-in-rocm-programming-models
 RUN echo "[ROCm]" > /etc/yum.repos.d/rocm.repo && \
     echo "name=ROCm" >> /etc/yum.repos.d/rocm.repo && \
@@ -50,8 +50,21 @@ RUN echo "[ROCm]" > /etc/yum.repos.d/rocm.repo && \
     echo "baseurl=https://repo.radeon.com/amdgpu/$AMDGPU_VERSION/rhel/9.4/main/x86_64" >> /etc/yum.repos.d/amdgpu.repo && \
     echo "enabled=1" >> /etc/yum.repos.d/amdgpu.repo && \
     echo "gpgcheck=0" >> /etc/yum.repos.d/amdgpu.repo && \
-    yum install -y rocm-developer-tools rocm-ml-sdk rocm-opencl-sdk rocm-openmp-sdk rocm-utils && \
-    yum clean all && rm -rf /var/cache/yum
+    # Install ROCm packages in separate steps to better manage disk space
+    # First install core ROCm components
+    yum install -y --setopt=install_weak_deps=False rocm-core rocm-hip-runtime rocm-hip-sdk && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Then install ML components
+    yum install -y --setopt=install_weak_deps=False rocm-ml-sdk && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Finally install remaining components
+    yum install -y --setopt=install_weak_deps=False rocm-opencl-sdk rocm-utils && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Remove unnecessary files to save space
+    find /opt/rocm-$ROCM_VERSION -name "*.a" -delete && \
+    find /opt/rocm-$ROCM_VERSION -name "*.o" -delete && \
+    find /opt/rocm-$ROCM_VERSION -type d -name "samples" -exec rm -rf {} + 2>/dev/null || true && \
+    find /opt/rocm-$ROCM_VERSION -type d -name "doc" -exec rm -rf {} + 2>/dev/null || true
 
 # Restore notebook user workspace
 USER 1001

runtimes/rocm-tensorflow/ubi9-python-3.11/Dockerfile.rocm

@@ -38,7 +38,7 @@ ARG AMDGPU_VERSION=6.2.4
 # Install the ROCm rpms
 # ref: https://github.com/ROCm/ROCm-docker/blob/master/dev/Dockerfile-centos-7-complete
 # Note: Based on 6.2 above new package mivisionx is a pre-requistes, which bring in more dependent packages
-# so we are only installing meta packages of rocm 
+# so we are only installing meta packages of rocm
 # ref: https://rocm.docs.amd.com/projects/install-on-linux/en/develop/reference/package-manager-integration.html#packages-in-rocm-programming-models
 RUN echo "[ROCm]" > /etc/yum.repos.d/rocm.repo && \
     echo "name=ROCm" >> /etc/yum.repos.d/rocm.repo && \
@@ -50,8 +50,21 @@ RUN echo "[ROCm]" > /etc/yum.repos.d/rocm.repo && \
     echo "baseurl=https://repo.radeon.com/amdgpu/$AMDGPU_VERSION/rhel/9.4/main/x86_64" >> /etc/yum.repos.d/amdgpu.repo && \
     echo "enabled=1" >> /etc/yum.repos.d/amdgpu.repo && \
     echo "gpgcheck=0" >> /etc/yum.repos.d/amdgpu.repo && \
-    yum install -y rocm-developer-tools rocm-ml-sdk rocm-opencl-sdk rocm-openmp-sdk rocm-utils && \
-    yum clean all && rm -rf /var/cache/yum
+    # Install ROCm packages in separate steps to better manage disk space
+    # First install core ROCm components
+    yum install -y --setopt=install_weak_deps=False rocm-core rocm-hip-runtime rocm-hip-sdk && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Then install ML components
+    yum install -y --setopt=install_weak_deps=False rocm-ml-sdk && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Finally install remaining components
+    yum install -y --setopt=install_weak_deps=False rocm-opencl-sdk rocm-utils && \
+    yum clean all && rm -rf /var/cache/yum && \
+    # Remove unnecessary files to save space
+    find /opt/rocm-$ROCM_VERSION -name "*.a" -delete && \
+    find /opt/rocm-$ROCM_VERSION -name "*.o" -delete && \
+    find /opt/rocm-$ROCM_VERSION -type d -name "samples" -exec rm -rf {} + 2>/dev/null || true && \
+    find /opt/rocm-$ROCM_VERSION -type d -name "doc" -exec rm -rf {} + 2>/dev/null || true
 
 # Restore notebook user workspace
 USER 1001