SynapseAi 1.5.0 release

 * Update dockerfiles with 1.5.0 content
 * Update installation scripts for 1.5.0 install versions
 * Add check_habana_framework_env.py script

dockerfiles/Dockerfile_amzn2_base_installer

@@ -38,11 +38,11 @@ RUN wget -nv "https://efa-installer.amazonaws.com/aws-efa-installer-${EFA_INSTAL
     cd /tmp/aws-efa-installer && ./efa_installer.sh -y --skip-kmod --skip-limit-conf --no-verify && \
     cd - && rm -rf /tmp/*
 
-ENV LD_LIBRARY_PATH=/opt/amazon/openmpi/lib:/opt/amazon/efa/lib:$LD_LIBRARY_PATH
-ENV PATH=/opt/amazon/openmpi/bin:/opt/amazon/efa/bin:$PATH
-ENV OPAL_PREFIX=/opt/amazon/openmpi
 ENV MPI_ROOT=/opt/amazon/openmpi
-ENV MPICC=/opt/amazon/openmpi/bin/mpicc
+ENV LD_LIBRARY_PATH=${MPI_ROOT}/lib:/opt/amazon/efa/lib:$LD_LIBRARY_PATH
+ENV PATH=${MPI_ROOT}/bin:/opt/amazon/efa/bin:$PATH
+ENV OPAL_PREFIX=${MPI_ROOT}
+ENV MPICC=${MPI_ROOT}/bin/mpicc
 ENV FI_EFA_ENABLE_SHM_TRANSFER=0
 
 RUN echo "[habanalabs]" > /etc/yum.repos.d/habanalabs.repo && \
@@ -60,8 +60,6 @@ RUN update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.8 2
 RUN rm -f /etc/yum.repos.d/habanalabs.repo && \
     yum clean all && rm -rf /var/cache/yum
 
-RUN python3 -m pip install hpu_media_loader=="${VERSION}"."${REVISION}" --no-cache-dir \
-    --extra-index-url https://"${ARTIFACTORY_URL}"/artifactory/api/pypi/gaudi-python/simple
 # SSH configuration necessary to support mpi-operator v2
 RUN mkdir -p /var/run/sshd && \
     sed -i 's/[ #]\(.*StrictHostKeyChecking \).*/ \1no/g' /etc/ssh/ssh_config && \
@@ -71,10 +69,11 @@ RUN mkdir -p /var/run/sshd && \
 # There is no need to store pip installation files inside docker image
 ENV PIP_NO_CACHE_DIR=off
 
+RUN python3 -m pip install hpu_media_loader=="${VERSION}"."${REVISION}"
+
 RUN echo "export LANG=en_US.UTF-8" >> /root/.bashrc
 RUN export LANG=en_US.UTF-8
 ENV GC_KERNEL_PATH=/usr/lib/habanalabs/libtpc_kernels.so
 ENV HABANA_LOGS=/var/log/habana_logs/
-ENV SCAL_CFG_FILE_PATH=/opt/habanalabs/etc/scal/default.json
 ENV HABANA_SCAL_BIN_PATH=/opt/habanalabs/engines_fw
 ENV HABANA_PLUGINS_LIB_PATH=/opt/habanalabs/habana_plugins

dockerfiles/Dockerfile_amzn2_pytorch_installer

@@ -13,6 +13,7 @@ ARG REVISION
 ARG ARTIFACTORY_URL
 
 ENV LANG=en_US.UTF-8
+ENV PYTHONPATH=/root:/usr/lib/habanalabs/
 
 RUN yum install -y \
     unzip \
@@ -27,7 +28,8 @@ RUN yum install -y \
     zlib-devel \
     lapack-devel \
     blas-devel \
-    numactl && \
+    numactl \
+    gperftools && \
     yum clean all
 
 RUN amazon-linux-extras install epel -y
@@ -57,4 +59,6 @@ RUN wget "https://${ARTIFACTORY_URL}"/artifactory/gaudi-pt-modules/"${VERSION}"/
     rm -rf /root/habanalabs/pytorch_temp/ && \
     rm -rf pytorch_modules-v"${PT_VERSION}"_"${VERSION}"_"${REVISION}".tgz
 
+ENV LD_PRELOAD=/lib64/libtcmalloc.so
+
 RUN rm -rf /tmp/*

dockerfiles/Dockerfile_amzn2_tensorflow_installer

@@ -29,15 +29,15 @@ RUN wget https://github.com/protocolbuffers/protobuf/releases/download/v3.6.1/pr
     unzip protoc-3.6.1-linux-x86_64.zip -d /usr/local/protoc && \
     rm -rf protoc-3.6.1-linux-x86_64.zip
 
-COPY requirements-no-deps-tensorflow-cpu-"$TF_VERSION".txt requirements-no-deps-tensorflow-cpu-"$TF_VERSION".txt
+COPY requirements-tensorflow-cpu-"$TF_VERSION".txt requirements-tensorflow-cpu-"$TF_VERSION".txt
 
 RUN python3 -m pip install pip==21.0.1 && \
     pip3 install setuptools==60.5.0 && \
     rm -rf get-pip.py && \
     pip3 install tensorflow-cpu==${TF_VERSION} \
         tensorflow-model-optimization==0.7.0 && \
-    pip3 install --no-deps -r requirements-no-deps-tensorflow-cpu-"$TF_VERSION".txt && \
-    rm requirements-no-deps-tensorflow-cpu-"$TF_VERSION".txt
+    pip3 install -r requirements-tensorflow-cpu-"$TF_VERSION".txt && \
+    rm requirements-tensorflow-cpu-"$TF_VERSION".txt
 
 # For AML/CentOS/RHEL OS'es TFIO_DATAPATH have to be specified to import tensorflow_io lib correctly
 ENV TFIO_DATAPATH=/usr/local/lib64/python3.8/site-packages/

dockerfiles/Dockerfile_rhel8.3_base_installer

@@ -54,11 +54,11 @@ RUN wget -nv "https://efa-installer.amazonaws.com/aws-efa-installer-${EFA_INSTAL
     ./efa_installer.sh -y --skip-kmod --skip-limit-conf --no-verify && \
     cd - && rm -rf /tmp/*
 
-ENV LD_LIBRARY_PATH=/opt/amazon/openmpi/lib:/opt/amazon/efa/lib:$LD_LIBRARY_PATH
-ENV PATH=/opt/amazon/openmpi/bin:/opt/amazon/efa/bin:$PATH
-ENV OPAL_PREFIX=/opt/amazon/openmpi
 ENV MPI_ROOT=/opt/amazon/openmpi
-ENV MPICC=/opt/amazon/openmpi/bin/mpicc
+ENV LD_LIBRARY_PATH=${MPI_ROOT}/lib:/opt/amazon/efa/lib:$LD_LIBRARY_PATH
+ENV PATH=${MPI_ROOT}/bin:/opt/amazon/efa/bin:$PATH
+ENV OPAL_PREFIX=${MPI_ROOT}
+ENV MPICC=${MPI_ROOT}/bin/mpicc
 ENV FI_EFA_ENABLE_SHM_TRANSFER=0
 
 RUN echo "[habanalabs]" > /etc/yum.repos.d/habanalabs.repo && \
@@ -78,8 +78,8 @@ RUN dnf install -y habanalabs-thunk-"$VERSION"-"$REVISION".el8 \
 RUN rm -f /etc/yum.repos.d/habanalabs.repo && rm -rf /tmp/* && \
     dnf clean all && rm -rf /var/cache/yum
 
-RUN python3 -m pip install hpu_media_loader=="${VERSION}"."${REVISION}" --no-cache-dir \
-        --extra-index-url https://"${ARTIFACTORY_URL}"/artifactory/api/pypi/gaudi-python/simple
+RUN python3 -m pip install hpu_media_loader=="${VERSION}"."${REVISION}" --no-cache-dir
+
 # SSH configuration necessary to support mpi-operator v2
 RUN mkdir -p /var/run/sshd && \
     sed -i 's/[ #]\(.*StrictHostKeyChecking \).*/ \1no/g' /etc/ssh/ssh_config && \
@@ -90,6 +90,5 @@ RUN echo "export LANG=en_US.UTF-8" >> /root/.bashrc
 RUN export LANG=en_US.UTF-8
 ENV GC_KERNEL_PATH=/usr/lib/habanalabs/libtpc_kernels.so
 ENV HABANA_LOGS=/var/log/habana_logs/
-ENV SCAL_CFG_FILE_PATH=/opt/habanalabs/etc/scal/default.json
 ENV HABANA_SCAL_BIN_PATH=/opt/habanalabs/engines_fw
 ENV HABANA_PLUGINS_LIB_PATH=/opt/habanalabs/habana_plugins

dockerfiles/Dockerfile_rhel8.3_pytorch_installer

@@ -15,6 +15,7 @@ ARG HABANA_PIP_VERSION="21.0.1"
 
 ENV SETUPTOOLS_VERSION=41.0.0
 ENV LANG=en_US.UTF-8
+ENV PYTHONPATH=/root:/usr/lib/habanalabs/
 
 RUN dnf install -y \
     unzip \
@@ -31,7 +32,8 @@ RUN dnf install -y \
     cpupowerutils \
     lapack-devel \
     blas-devel \
-    numactl && \
+    numactl \
+    gperftools.x86_64 && \
     dnf clean all && rm -rf /var/cache/yum
 
 RUN python3 -m pip install pip=="${HABANA_PIP_VERSION}" && \
@@ -55,4 +57,6 @@ RUN wget "https://${ARTIFACTORY_URL}"/artifactory/gaudi-pt-modules/"${VERSION}"/
     rm -rf /root/habanalabs/pytorch_temp/ && \
     rm -rf pytorch_modules-v"${PT_VERSION}"_"${VERSION}"_"${REVISION}".tgz
 
+ENV LD_PRELOAD=/lib64/libtcmalloc.so
+
 RUN rm -rf /tmp/*

dockerfiles/Dockerfile_rhel8.3_tensorflow_installer

@@ -25,15 +25,15 @@ RUN dnf install -y libarchive unzip \
         protobuf-devel && \
     dnf clean all && rm -rf /var/cache/dnf
 
-COPY requirements-no-deps-tensorflow-cpu-"$TF_VERSION".txt requirements-no-deps-tensorflow-cpu-"$TF_VERSION".txt
+COPY requirements-tensorflow-cpu-"$TF_VERSION".txt requirements-tensorflow-cpu-"$TF_VERSION".txt
 
 RUN python3 -m pip install pip=="${HABANA_PIP_VERSION}" && \
     pip3 install setuptools==60.5.0
 
 RUN pip3 install tensorflow-cpu==${TF_VERSION} \
         tensorflow-model-optimization==0.7.0 && \
-    pip3 install --no-deps -r requirements-no-deps-tensorflow-cpu-"$TF_VERSION".txt && \
-    rm requirements-no-deps-tensorflow-cpu-"$TF_VERSION".txt
+    pip3 install -r requirements-tensorflow-cpu-"$TF_VERSION".txt && \
+    rm requirements-tensorflow-cpu-"$TF_VERSION".txt
 
 # For AML/CentOS/RHEL OS'es TFIO_DATAPATH have to be specified to import tensorflow_io lib correctly
 ENV TFIO_DATAPATH=/usr/local/lib64/python3.8/site-packages/
@@ -45,7 +45,6 @@ RUN ln -s /etc/ssl/certs/ca-bundle.crt /etc/ssl/certs/ca-certificates.crt
 RUN python3 -m pip install habana-tensorflow=="${VERSION}"."${REVISION}" && \
     python3 -m pip install habana-horovod=="${VERSION}"."${REVISION}"
 
-
 RUN /usr/bin/ssh-keygen -A && \
     echo "source /etc/profile.d/habanalabs.sh" >> ~/.bashrc && \
     echo "/usr/sbin/sshd -p 3022" >> ~/.bashrc && \