[PyTorch][Training][EC2][SageMaker]PyTorch 2.8 Currency Release

dlc_developer_config.toml

@@ -37,12 +37,12 @@ deep_canary_mode = false
 [build]
 # Add in frameworks you would like to build. By default, builds are disabled unless you specify building an image.
 # available frameworks - ["base", "vllm", "autogluon", "huggingface_tensorflow", "huggingface_pytorch", "huggingface_tensorflow_trcomp", "huggingface_pytorch_trcomp", "pytorch_trcomp", "tensorflow", "pytorch", "stabilityai_pytorch"]
-build_frameworks = []
+build_frameworks = ["pytorch"]
 
 
 # By default we build both training and inference containers. Set true/false values to determine which to build.
 build_training = true
-build_inference = true
+build_inference = false
 
 # Set do_build to "false" to skip builds and test the latest image built by this PR
 # Note: at least one build is required to set do_build to "false"
@@ -65,13 +65,13 @@ ecs_tests = true
 eks_tests = true
 ec2_tests = true
 # Set it to true if you are preparing a Benchmark related PR
-ec2_benchmark_tests = false
+ec2_benchmark_tests = true
 
 ### Set ec2_tests_on_heavy_instances = true to be able to run any EC2 tests that use large/expensive instance types by
 ### default. If false, these types of tests will be skipped while other tests will run as usual.
 ### These tests are run in EC2 test jobs, so ec2_tests must be true if ec2_tests_on_heavy_instances is true.
 ### Off by default (set to false)
-ec2_tests_on_heavy_instances = false
+ec2_tests_on_heavy_instances = true
 ### SM specific tests
 ### On by default
 sagemaker_local_tests = true
@@ -119,7 +119,7 @@ use_scheduler = false
 ### TRAINING PR JOBS ###
 
 # Standard Framework Training
-dlc-pr-pytorch-training = ""
+dlc-pr-pytorch-training = "pytorch/training/buildspec-2-8-ec2.yml"
 dlc-pr-tensorflow-2-training = ""
 dlc-pr-autogluon-training = ""
 

pytorch/training/buildspec-2-8-ec2.yml

@@ -0,0 +1,72 @@
+account_id: &ACCOUNT_ID <set-$ACCOUNT_ID-in-environment>
+prod_account_id: &PROD_ACCOUNT_ID 763104351884
+region: &REGION <set-$REGION-in-environment>
+framework: &FRAMEWORK pytorch
+version: &VERSION 2.8.0
+short_version: &SHORT_VERSION "2.8"
+arch_type: x86
+# autopatch_build: "True"
+
+repository_info:
+  training_repository: &TRAINING_REPOSITORY
+    image_type: &TRAINING_IMAGE_TYPE training
+    root: !join [ *FRAMEWORK, "/", *TRAINING_IMAGE_TYPE ]
+    repository_name: &REPOSITORY_NAME !join [ pr, "-", *FRAMEWORK, "-", *TRAINING_IMAGE_TYPE ]
+    repository: &REPOSITORY !join [ *ACCOUNT_ID, .dkr.ecr., *REGION, .amazonaws.com/, *REPOSITORY_NAME ]
+    release_repository_name: &RELEASE_REPOSITORY_NAME !join [ *FRAMEWORK, "-", *TRAINING_IMAGE_TYPE ]
+    release_repository: &RELEASE_REPOSITORY !join [ *PROD_ACCOUNT_ID, .dkr.ecr., *REGION, .amazonaws.com/, *RELEASE_REPOSITORY_NAME ]
+
+context:
+  training_context: &TRAINING_CONTEXT
+    start_cuda_compat:
+      source: docker/build_artifacts/start_cuda_compat.sh
+      target: start_cuda_compat.sh
+    dockerd_entrypoint:
+      source: docker/build_artifacts/dockerd_entrypoint.sh
+      target: dockerd_entrypoint.sh
+    changehostname:
+      source: docker/build_artifacts/changehostname.c
+      target: changehostname.c
+    start_with_right_hostname:
+      source: docker/build_artifacts/start_with_right_hostname.sh
+      target: start_with_right_hostname.sh
+    example_mnist_file:
+      source: docker/build_artifacts/mnist.py
+      target: mnist.py
+    deep_learning_container:
+      source: ../../src/deep_learning_container.py
+      target: deep_learning_container.py
+
+images:
+  BuildEC2CPUPTTrainPy3DockerImage:
+    <<: *TRAINING_REPOSITORY
+    build: &PYTORCH_CPU_TRAINING_PY3 false
+    image_size_baseline: 7200
+    device_type: &DEVICE_TYPE cpu
+    python_version: &DOCKER_PYTHON_VERSION py3
+    tag_python_version: &TAG_PYTHON_VERSION py312
+    os_version: &OS_VERSION ubuntu22.04
+    tag: !join [ *VERSION, "-", *DEVICE_TYPE, "-", *TAG_PYTHON_VERSION, "-", *OS_VERSION, "-ec2" ]
+    latest_release_tag: !join [ *VERSION, "-", *DEVICE_TYPE, "-", *TAG_PYTHON_VERSION, "-", *OS_VERSION, "-ec2" ]
+    # skip_build: "False"
+    docker_file: !join [ docker/, *SHORT_VERSION, /, *DOCKER_PYTHON_VERSION, /Dockerfile., *DEVICE_TYPE ]
+    target: ec2
+    context:
+      <<: *TRAINING_CONTEXT
+  BuildEC2GPUPTTrainPy3cu129DockerImage:
+    <<: *TRAINING_REPOSITORY
+    build: &PYTORCH_GPU_TRAINING_PY3 false
+    image_size_baseline: 24000
+    device_type: &DEVICE_TYPE gpu
+    python_version: &DOCKER_PYTHON_VERSION py3
+    tag_python_version: &TAG_PYTHON_VERSION py312
+    cuda_version: &CUDA_VERSION cu129
+    os_version: &OS_VERSION ubuntu22.04
+    tag: !join [ *VERSION, "-", *DEVICE_TYPE, "-", *TAG_PYTHON_VERSION, "-", *CUDA_VERSION, "-", *OS_VERSION, "-ec2" ]
+    latest_release_tag: !join [ *VERSION, "-", *DEVICE_TYPE, "-", *TAG_PYTHON_VERSION, "-", *CUDA_VERSION, "-", *OS_VERSION, "-ec2" ]
+    # skip_build: "False"
+    docker_file: !join [ docker/, *SHORT_VERSION, /, *DOCKER_PYTHON_VERSION, /, *CUDA_VERSION, /Dockerfile.,
+                         *DEVICE_TYPE ]
+    target: ec2
+    context:
+      <<: *TRAINING_CONTEXT

pytorch/training/buildspec-2-8-sm.yml

@@ -0,0 +1,72 @@
+account_id: &ACCOUNT_ID <set-$ACCOUNT_ID-in-environment>
+prod_account_id: &PROD_ACCOUNT_ID 763104351884
+region: &REGION <set-$REGION-in-environment>
+framework: &FRAMEWORK pytorch
+version: &VERSION 2.8.0
+short_version: &SHORT_VERSION "2.8"
+arch_type: x86
+# autopatch_build: "True"
+
+repository_info:
+  training_repository: &TRAINING_REPOSITORY
+    image_type: &TRAINING_IMAGE_TYPE training
+    root: !join [ *FRAMEWORK, "/", *TRAINING_IMAGE_TYPE ]
+    repository_name: &REPOSITORY_NAME !join [ pr, "-", *FRAMEWORK, "-", *TRAINING_IMAGE_TYPE ]
+    repository: &REPOSITORY !join [ *ACCOUNT_ID, .dkr.ecr., *REGION, .amazonaws.com/, *REPOSITORY_NAME ]
+    release_repository_name: &RELEASE_REPOSITORY_NAME !join [ *FRAMEWORK, "-", *TRAINING_IMAGE_TYPE ]
+    release_repository: &RELEASE_REPOSITORY !join [ *PROD_ACCOUNT_ID, .dkr.ecr., *REGION, .amazonaws.com/, *RELEASE_REPOSITORY_NAME ]
+
+context:
+  training_context: &TRAINING_CONTEXT
+    start_cuda_compat:
+      source: docker/build_artifacts/start_cuda_compat.sh
+      target: start_cuda_compat.sh
+    dockerd_entrypoint:
+      source: docker/build_artifacts/dockerd_entrypoint.sh
+      target: dockerd_entrypoint.sh
+    changehostname:
+      source: docker/build_artifacts/changehostname.c
+      target: changehostname.c
+    start_with_right_hostname:
+      source: docker/build_artifacts/start_with_right_hostname.sh
+      target: start_with_right_hostname.sh
+    example_mnist_file:
+      source: docker/build_artifacts/mnist.py
+      target: mnist.py
+    deep_learning_container:
+      source: ../../src/deep_learning_container.py
+      target: deep_learning_container.py
+
+images:
+  BuildSageMakerCPUPTTrainPy3DockerImage:
+    <<: *TRAINING_REPOSITORY
+    build: &PYTORCH_CPU_TRAINING_PY3 false
+    image_size_baseline: 7200
+    device_type: &DEVICE_TYPE cpu
+    python_version: &DOCKER_PYTHON_VERSION py3
+    tag_python_version: &TAG_PYTHON_VERSION py312
+    os_version: &OS_VERSION ubuntu22.04
+    tag: !join [ *VERSION, "-", *DEVICE_TYPE, "-", *TAG_PYTHON_VERSION, "-", *OS_VERSION, "-sagemaker" ]
+    latest_release_tag: !join [ *VERSION, "-", *DEVICE_TYPE, "-", *TAG_PYTHON_VERSION, "-", *OS_VERSION, "-sagemaker" ]
+    # skip_build: "False"
+    docker_file: !join [ docker/, *SHORT_VERSION, /, *DOCKER_PYTHON_VERSION, /Dockerfile., *DEVICE_TYPE ]
+    target: sagemaker
+    context:
+      <<: *TRAINING_CONTEXT
+  BuildSageMakerGPUPTTrainPy3DockerImage:
+    <<: *TRAINING_REPOSITORY
+    build: &PYTORCH_GPU_TRAINING_PY3 false
+    image_size_baseline: 24000
+    device_type: &DEVICE_TYPE gpu
+    python_version: &DOCKER_PYTHON_VERSION py3
+    tag_python_version: &TAG_PYTHON_VERSION py312
+    cuda_version: &CUDA_VERSION cu129
+    os_version: &OS_VERSION ubuntu22.04
+    tag: !join [ *VERSION, "-", *DEVICE_TYPE, "-", *TAG_PYTHON_VERSION, "-", *CUDA_VERSION, "-", *OS_VERSION, "-sagemaker" ]
+    latest_release_tag: !join [ *VERSION, "-", *DEVICE_TYPE, "-", *TAG_PYTHON_VERSION, "-", *CUDA_VERSION, "-", *OS_VERSION, "-sagemaker" ]
+    # skip_build: "False"
+    docker_file: !join [ docker/, *SHORT_VERSION, /, *DOCKER_PYTHON_VERSION, /, *CUDA_VERSION, /Dockerfile.,
+                         *DEVICE_TYPE ]
+    target: sagemaker
+    context:
+      <<: *TRAINING_CONTEXT

pytorch/training/buildspec.yml

@@ -1 +1 @@
-buildspec_pointer: buildspec-2-7-sm.yml
+buildspec_pointer: buildspec-2-8-ec2.yml