pytorch · vfdev-5 · Oct 18, 2023 · Oct 18, 2023 · Oct 18, 2023
diff --git a/.github/workflows/docker-build.yml b/.github/workflows/docker-build.yml
diff --git a/.github/workflows/docker-publish.yml b/.github/workflows/docker-publish.yml
diff --git a/docker/build.sh b/docker/build.sh
@@ -91,5 +91,5 @@ echo "Show installed packages:"
 docker run --rm -i pytorchignite/${image_name}:${image_tag} pip list
 
 echo "Test pytorchignite/${image_name}:${image_tag}"
-python test_image.py pytorchignite/${image_name}:${image_tag}
+docker run --rm -i -v $PWD:/ws -w /ws -e HVD_VERSION=${HVD_VERSION:-} -e MSDP_VERSION=${MSDP_VERSION:-} pytorchignite/${image_name}:${image_tag} /bin/bash -c "python test_image.py pytorchignite/${image_name}:${image_tag}"
 echo "OK"
diff --git a/docker/docker.cfg b/docker/docker.cfg
@@ -1,4 +1,4 @@
 [DEFAULT]
-build_docker_image_pytorch_version = 2.0.0-cuda11.7-cudnn8
-build_docker_image_hvd_version = v0.27.0
+build_docker_image_pytorch_version = 2.1.0-cuda12.1-cudnn8
+build_docker_image_hvd_version = v0.28.1
 build_docker_image_msdp_version = v0.8.1
diff --git a/docker/hvd/Dockerfile.hvd-apex b/docker/hvd/Dockerfile.hvd-apex
@@ -6,8 +6,6 @@ ARG PTH_VERSION
 # 1/Building apex with pytorch:*-devel
 FROM pytorch/pytorch:${PTH_VERSION}-devel AS apex-hvd-builder
 
-ARG ARG_TORCH_CUDA_ARCH_LIST="3.5 5.2 6.0 6.1 7.0+PTX 8.0 8.6"
-ENV TORCH_CUDA_ARCH_LIST=$ARG_TORCH_CUDA_ARCH_LIST
 ENV CUDA_HOME=/usr/local/cuda
 
 # Install git
@@ -21,7 +19,7 @@ RUN echo "Setup NVIDIA Apex" && \
     git clone https://github.com/NVIDIA/apex $tmp_apex_path && \
     cd $tmp_apex_path && \
     pip install packaging && \
-    pip wheel -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" .
+    pip wheel -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" .
 
 ARG HVD_VERSION
 
@@ -30,6 +28,9 @@ RUN apt-get update && apt-get install -y git && \
         git clone --recursive --depth 1 --branch ${HVD_VERSION} https://github.com/horovod/horovod.git /horovod && \
         conda install -y cmake nccl -c conda-forge && \
         cd /horovod && \
+        # temporary -std=c++17 fix
+        sed -i "s/CMAKE_CXX_STANDARD 14/CMAKE_CXX_STANDARD 17/g" CMakeLists.txt && \
+        sed -i "s/CMAKE_CXX_STANDARD 14/CMAKE_CXX_STANDARD 17/g" horovod/torch/CMakeLists.txt && \
         HOROVOD_GPU_OPERATIONS=NCCL HOROVOD_NCCL_LINK=SHARED HOROVOD_WITHOUT_MPI=1 HOROVOD_WITH_PYTORCH=1 pip wheel --no-cache-dir . && \
         rm -rf /var/lib/apt/lists/*
 

diff --git a/docker/hvd/Dockerfile.hvd-apex-nlp b/docker/hvd/Dockerfile.hvd-apex-nlp
@@ -4,4 +4,5 @@ FROM pytorchignite/hvd-apex:latest
 # Ignite NLP dependencies
 RUN pip install --upgrade --no-cache-dir transformers \
                                          spacy \
-                                         nltk
+                                         nltk \
+                                         torchtext
diff --git a/docker/hvd/Dockerfile.hvd-apex-vision b/docker/hvd/Dockerfile.hvd-apex-vision
@@ -1,15 +1,6 @@
 # Dockerfile.hvd-apex-vision
 FROM pytorchignite/hvd-apex:latest
 
-# Install opencv dependencies
-RUN apt-get update && \
-    apt-get -y install --no-install-recommends libglib2.0 \
-                                               libsm6 \
-                                               libxext6 \
-                                               libxrender-dev \
-                                               libgl1-mesa-glx && \
-    rm -rf /var/lib/apt/lists/*
-
 # Ignite vision dependencies
 RUN pip install --upgrade --no-cache-dir albumentations \
                                          image-dataset-viz \

diff --git a/docker/hvd/Dockerfile.hvd-base b/docker/hvd/Dockerfile.hvd-base
@@ -12,6 +12,9 @@ RUN apt-get update && apt-get install -y git && \
         git clone --recursive --depth 1 --branch ${HVD_VERSION} https://github.com/horovod/horovod.git /horovod && \
         conda install -y cmake nccl -c conda-forge && \
         cd /horovod && \
+        # temporary -std=c++17 fix
+        sed -i "s/CMAKE_CXX_STANDARD 14/CMAKE_CXX_STANDARD 17/g" CMakeLists.txt && \
+        sed -i "s/CMAKE_CXX_STANDARD 14/CMAKE_CXX_STANDARD 17/g" horovod/torch/CMakeLists.txt && \
         HOROVOD_GPU_OPERATIONS=NCCL HOROVOD_NCCL_LINK=SHARED HOROVOD_WITHOUT_MPI=1 HOROVOD_WITH_PYTORCH=1 pip wheel --no-cache-dir . && \
         rm -rf /var/lib/apt/lists/*
 

diff --git a/docker/hvd/Dockerfile.hvd-nlp b/docker/hvd/Dockerfile.hvd-nlp
@@ -4,4 +4,5 @@ FROM pytorchignite/hvd-base:latest
 # Ignite NLP dependencies
 RUN pip install --upgrade --no-cache-dir transformers \
                                          spacy \
-                                         nltk
+                                         nltk \
+                                         torchtext
diff --git a/docker/hvd/Dockerfile.hvd-vision b/docker/hvd/Dockerfile.hvd-vision
@@ -1,15 +1,6 @@
 # Dockerfile.hvd-vision
 FROM pytorchignite/hvd-base:latest
 
-# Install opencv dependencies
-RUN apt-get update && \
-    apt-get -y install --no-install-recommends libglib2.0 \
-                                               libsm6 \
-                                               libxext6 \
-                                               libxrender-dev \
-                                               libgl1-mesa-glx && \
-    rm -rf /var/lib/apt/lists/*
-
 # Ignite vision dependencies
 RUN pip install --upgrade --no-cache-dir albumentations \
                                          image-dataset-viz \

diff --git a/docker/main/Dockerfile.apex b/docker/main/Dockerfile.apex
@@ -6,8 +6,6 @@ ARG PTH_VERSION
 # 1/Building apex with pytorch:*-devel
 FROM pytorch/pytorch:${PTH_VERSION}-devel AS apex-builder
 
-ARG ARG_TORCH_CUDA_ARCH_LIST="3.5 5.2 6.0 6.1 7.0+PTX 8.0 8.6"
-ENV TORCH_CUDA_ARCH_LIST=$ARG_TORCH_CUDA_ARCH_LIST
 ENV CUDA_HOME=/usr/local/cuda
 
 # Install git
@@ -21,7 +19,7 @@ RUN echo "Setup NVIDIA Apex" && \
     git clone https://github.com/NVIDIA/apex $tmp_apex_path && \
     cd $tmp_apex_path && \
     pip install packaging && \
-    pip wheel -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" .
+    pip wheel -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" .
 
 # 2/ Build the runtime image
 FROM pytorch/pytorch:${PTH_VERSION}-runtime

diff --git a/docker/main/Dockerfile.apex-nlp b/docker/main/Dockerfile.apex-nlp
@@ -4,4 +4,5 @@ FROM pytorchignite/apex:latest
 # Ignite NLP dependencies
 RUN pip install --upgrade --no-cache-dir transformers \
                                          spacy \
-                                         nltk
+                                         nltk \
+                                         torchtext
diff --git a/docker/main/Dockerfile.apex-vision b/docker/main/Dockerfile.apex-vision
@@ -1,15 +1,6 @@
 # Dockerfile.apex-vision
 FROM pytorchignite/apex:latest
 
-# Install opencv dependencies
-RUN apt-get update && \
-    apt-get -y install --no-install-recommends libglib2.0 \
-                                               libsm6 \
-                                               libxext6 \
-                                               libxrender-dev \
-                                               libgl1-mesa-glx && \
-    rm -rf /var/lib/apt/lists/*
-
 # Ignite vision dependencies
 RUN pip install --upgrade --no-cache-dir albumentations \
                                          image-dataset-viz \

diff --git a/docker/main/Dockerfile.nlp b/docker/main/Dockerfile.nlp
@@ -4,4 +4,5 @@ FROM pytorchignite/base:latest
 # Ignite NLP dependencies
 RUN pip install --upgrade --no-cache-dir transformers \
                                          spacy \
-                                         nltk
+                                         nltk \
+                                         torchtext
diff --git a/docker/main/Dockerfile.vision b/docker/main/Dockerfile.vision
@@ -1,15 +1,6 @@
 # Dockerfile.vision
 FROM pytorchignite/base:latest
 
-# Install opencv dependencies
-RUN apt-get update && \
-    apt-get -y install --no-install-recommends libglib2.0 \
-                                               libsm6 \
-                                               libxext6 \
-                                               libxrender-dev \
-                                               libgl1-mesa-glx && \
-    rm -rf /var/lib/apt/lists/*
-
 # Ignite vision dependencies
 RUN pip install --upgrade --no-cache-dir albumentations \
                                          image-dataset-viz \

diff --git a/docker/msdp/Dockerfile.msdp-apex b/docker/msdp/Dockerfile.msdp-apex
@@ -6,8 +6,6 @@ ARG PTH_VERSION
 # 1/Building apex with pytorch:*-devel
 FROM pytorch/pytorch:${PTH_VERSION}-devel AS apex-msdp-builder
 
-ARG ARG_TORCH_CUDA_ARCH_LIST="3.5 5.2 6.0 6.1 7.0+PTX 8.0 8.6"
-ENV TORCH_CUDA_ARCH_LIST=$ARG_TORCH_CUDA_ARCH_LIST
 ENV CUDA_HOME=/usr/local/cuda
 
 # Install git
@@ -21,7 +19,7 @@ RUN echo "Setup NVIDIA Apex" && \
     git clone https://github.com/NVIDIA/apex $tmp_apex_path && \
     cd $tmp_apex_path && \
     pip install packaging && \
-    pip wheel -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" .
+    pip wheel -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" .
 
 # For pip --use-feature option
 RUN python -m pip install --upgrade pip

diff --git a/docker/msdp/Dockerfile.msdp-apex-nlp b/docker/msdp/Dockerfile.msdp-apex-nlp
@@ -4,4 +4,5 @@ FROM pytorchignite/msdp-apex:latest
 # Ignite NLP dependencies
 RUN pip install --upgrade --no-cache-dir transformers \
                                          spacy \
-                                         nltk
+                                         nltk \
+                                         torchtext
diff --git a/docker/msdp/Dockerfile.msdp-apex-vision b/docker/msdp/Dockerfile.msdp-apex-vision
@@ -1,15 +1,6 @@
 # Dockerfile.msdp-apex-vision
 FROM pytorchignite/msdp-apex:latest
 
-# Install opencv dependencies
-RUN apt-get update && \
-    apt-get -y install --no-install-recommends libglib2.0 \
-                                               libsm6 \
-                                               libxext6 \
-                                               libxrender-dev \
-                                               libgl1-mesa-glx && \
-    rm -rf /var/lib/apt/lists/*
-
 # Ignite vision dependencies
 RUN pip install --upgrade --no-cache-dir albumentations \
                                          image-dataset-viz \

diff --git a/docker/push_all.sh b/docker/push_all.sh
@@ -14,46 +14,66 @@ if [ -z $DOCKER_TOKEN ]; then
     exit 1
 fi
 
+if [ -z "$1" ]; then
+    push_selected_image="all"
+else
+    push_selected_image="$1"
+fi
+
 set -eu
 
 echo $DOCKER_TOKEN | docker login --username=$DOCKER_USER --password-stdin
 
 set -xeu
 
-image_name="base"
-image_tag=`docker run --rm -i pytorchignite/${image_name}:latest python -c "import torch; import ignite; print(torch.__version__ + \"-\" + ignite.__version__, end=\"\")"`
 
-for image_name in "base" "vision" "nlp" "apex" "apex-vision" "apex-nlp"
-do
+if [ ${folder_name} == "all" ]; then
 
-    docker push pytorchignite/${image_name}:latest
-    docker push pytorchignite/${image_name}:${image_tag}
+    image_name="base"
+    image_tag=`docker run --rm -i pytorchignite/${image_name}:latest python -c "import torch; import ignite; print(torch.__version__ + \"-\" + ignite.__version__, end=\"\")"`
 
-done
+    for image_name in "base" "vision" "nlp" "apex" "apex-vision" "apex-nlp"
+    do
 
+        docker push pytorchignite/${image_name}:latest
+        docker push pytorchignite/${image_name}:${image_tag}
 
-image_name="hvd-base"
-image_tag=`docker run --rm -i pytorchignite/${image_name}:latest python -c "import torch; import ignite; print(torch.__version__ + \"-\" + ignite.__version__, end=\"\")"`
+    done
 
-for image_name in "hvd-base" "hvd-vision" "hvd-nlp" "hvd-apex" "hvd-apex-vision" "hvd-apex-nlp"
-do
+    image_name="hvd-base"
+    image_tag=`docker run --rm -i pytorchignite/${image_name}:latest python -c "import torch; import ignite; print(torch.__version__ + \"-\" + ignite.__version__, end=\"\")"`
 
-    docker push pytorchignite/${image_name}:latest
-    docker push pytorchignite/${image_name}:${image_tag}
+    for image_name in "hvd-base" "hvd-vision" "hvd-nlp" "hvd-apex" "hvd-apex-vision" "hvd-apex-nlp"
+    do
+
+        docker push pytorchignite/${image_name}:latest
+        docker push pytorchignite/${image_name}:${image_tag}
+
+    done
 
-done
+    # DEPRECATED due to no activity
+    # image_name="msdp-apex"
+    # image_tag=`docker run --rm -i pytorchignite/${image_name}:latest python -c "import torch; import ignite; print(torch.__version__ + \"-\" + ignite.__version__, end=\"\")"`
 
-# DEPRECATED due to no activity
-# image_name="msdp-apex"
-# image_tag=`docker run --rm -i pytorchignite/${image_name}:latest python -c "import torch; import ignite; print(torch.__version__ + \"-\" + ignite.__version__, end=\"\")"`
+    # for image_name in "msdp-apex" "msdp-apex-vision" "msdp-apex-nlp"
+    # do
 
-# for image_name in "msdp-apex" "msdp-apex-vision" "msdp-apex-nlp"
-# do
+    #     docker push pytorchignite/${image_name}:latest
+    #     docker push pytorchignite/${image_name}:${image_tag}
+
+    # done
+
+else
+
+    image_name=${push_selected_image}
+    image_tag=`docker run --rm -i pytorchignite/${image_name}:latest python -c "import torch; import ignite; print(torch.__version__ + \"-\" + ignite.__version__, end=\"\")"`
+
+    docker push pytorchignite/${image_name}:latest
+    docker push pytorchignite/${image_name}:${image_tag}
+
+fi
 
-#     docker push pytorchignite/${image_name}:latest
-#     docker push pytorchignite/${image_name}:${image_tag}
 
-# done
 
 # If use locally, mind to clean dangling images
 # docker images | grep 'pytorchignite\|<none>' | awk '{print $3}' | xargs docker rmi -f