Actually, why not just use our official images

Author: malfet

.circleci/config.yml

@@ -92,8 +92,6 @@ pytorch_tutorial_build_defaults: &pytorch_tutorial_build_defaults
 
         # for some reason, pip installs it in a different place than what is looked at in the py file
         sudo pip3 install requests --target=/opt/circleci/.pyenv/versions/3.9.4/lib/python3.9/site-packages
-        export pyTorchDockerImageTag=$(python3 .jenkins/get_docker_tag.py)
-        echo "PyTorchDockerImageTag: "${pyTorchDockerImageTag}
 
         cat >/home/circleci/project/ci_build_script.sh \<<EOL
         # =================== The following code will be executed inside Docker container ===================
@@ -104,7 +102,6 @@ pytorch_tutorial_build_defaults: &pytorch_tutorial_build_defaults
         EOL
         chmod +x /home/circleci/project/ci_build_script.sh
 
-        export DOCKER_IMAGE=${DOCKER_IMAGE}:${pyTorchDockerImageTag}
         echo "DOCKER_IMAGE: "${DOCKER_IMAGE}
         docker pull ${DOCKER_IMAGE} >/dev/null
         if [ -n "${CUDA_VERSION}" ]; then
@@ -141,14 +138,14 @@ pytorch_tutorial_build_defaults: &pytorch_tutorial_build_defaults
 
 pytorch_tutorial_build_worker_defaults: &pytorch_tutorial_build_worker_defaults
   environment:
-    DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-bionic-cuda11.6-cudnn8-py3-gcc7"
+    DOCKER_IMAGE: "pytorch/pytorch:1.12.1-cuda11.3-cudnn8-devel"
     CUDA_VERSION: "9"
   resource_class: gpu.nvidia.small
   <<: *pytorch_tutorial_build_defaults
 
 pytorch_tutorial_build_manager_defaults: &pytorch_tutorial_build_manager_defaults
   environment:
-    DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-bionic-cuda11.6-cudnn8-py3-gcc7"
+    DOCKER_IMAGE: "pytorch/pytorch:1.12.1-cuda11.3-cudnn8-devel"
   resource_class: medium
   <<: *pytorch_tutorial_build_defaults
 

.circleci/config.yml.in

@@ -92,8 +92,6 @@ pytorch_tutorial_build_defaults: &pytorch_tutorial_build_defaults
 
         # for some reason, pip installs it in a different place than what is looked at in the py file
         sudo pip3 install requests --target=/opt/circleci/.pyenv/versions/3.9.4/lib/python3.9/site-packages
-        export pyTorchDockerImageTag=$(python3 .jenkins/get_docker_tag.py)
-        echo "PyTorchDockerImageTag: "${pyTorchDockerImageTag}
 
         cat >/home/circleci/project/ci_build_script.sh \<<EOL
         # =================== The following code will be executed inside Docker container ===================
@@ -104,7 +102,6 @@ pytorch_tutorial_build_defaults: &pytorch_tutorial_build_defaults
         EOL
         chmod +x /home/circleci/project/ci_build_script.sh
 
-        export DOCKER_IMAGE=${DOCKER_IMAGE}:${pyTorchDockerImageTag}
         echo "DOCKER_IMAGE: "${DOCKER_IMAGE}
         docker pull ${DOCKER_IMAGE} >/dev/null
         if [ -n "${CUDA_VERSION}" ]; then
@@ -141,14 +138,14 @@ pytorch_tutorial_build_defaults: &pytorch_tutorial_build_defaults
 
 pytorch_tutorial_build_worker_defaults: &pytorch_tutorial_build_worker_defaults
   environment:
-    DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-bionic-cuda11.6-cudnn8-py3-gcc7"
+    DOCKER_IMAGE: "pytorch/pytorch:1.12.1-cuda11.3-cudnn8-devel"
     CUDA_VERSION: "9"
   resource_class: gpu.nvidia.small
   <<: *pytorch_tutorial_build_defaults
 
 pytorch_tutorial_build_manager_defaults: &pytorch_tutorial_build_manager_defaults
   environment:
-    DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-bionic-cuda11.6-cudnn8-py3-gcc7"
+    DOCKER_IMAGE: "pytorch/pytorch:1.12.1-cuda11.3-cudnn8-devel"
   resource_class: medium
   <<: *pytorch_tutorial_build_defaults
 {% raw %}