ML CI: Linux x86_64 (#34299)

* ML CI: Linux x86_64

* Update comments

* Rename again

* Rename comments

* Update to match other arches

* No compiler

* Compiler was wrong anyway

* Faster TF

share/spack/gitlab/cloud_pipelines/.gitlab-ci.yml

@@ -760,122 +760,122 @@ tutorial-protected-build:
     - artifacts: True
       job: tutorial-protected-generate
 
-########################################
+#######################################
-# Machine Learning (CPU)
+# Machine Learning - Linux x86_64 (CPU)
-########################################
+#######################################
-.ml-cpu:
+.ml-linux-x86_64-cpu:
   variables:
-    SPACK_CI_STACK_NAME: ml-cpu
+    SPACK_CI_STACK_NAME: ml-linux-x86_64-cpu
 
-.ml-cpu-generate:
+.ml-linux-x86_64-cpu-generate:
-  extends: .ml-cpu
+  extends: .ml-linux-x86_64-cpu
   image: ghcr.io/spack/e4s-amazonlinux-2:v2022-03-21
   tags: ["spack", "aws", "public", "medium", "x86_64_v4"]
 
-ml-cpu-pr-generate:
+ml-linux-x86_64-cpu-pr-generate:
-  extends: [ ".ml-cpu-generate", ".pr-generate"]
+  extends: [ ".ml-linux-x86_64-cpu-generate", ".pr-generate"]
 
-ml-cpu-protected-generate:
+ml-linux-x86_64-cpu-protected-generate:
-  extends: [ ".ml-cpu-generate", ".protected-generate"]
+  extends: [ ".ml-linux-x86_64-cpu-generate", ".protected-generate"]
 
-ml-cpu-pr-build:
+ml-linux-x86_64-cpu-pr-build:
-  extends: [ ".ml-cpu", ".pr-build" ]
+  extends: [ ".ml-linux-x86_64-cpu", ".pr-build" ]
   trigger:
     include:
       - artifact: jobs_scratch_dir/cloud-ci-pipeline.yml
-        job: ml-cpu-pr-generate
+        job: ml-linux-x86_64-cpu-pr-generate
     strategy: depend
   needs:
     - artifacts: True
-      job: ml-cpu-pr-generate
+      job: ml-linux-x86_64-cpu-pr-generate
 
-ml-cpu-protected-build:
+ml-linux-x86_64-cpu-protected-build:
-  extends: [ ".ml-cpu", ".protected-build" ]
+  extends: [ ".ml-linux-x86_64-cpu", ".protected-build" ]
   trigger:
     include:
       - artifact: jobs_scratch_dir/cloud-ci-pipeline.yml
-        job: ml-cpu-protected-generate
+        job: ml-linux-x86_64-cpu-protected-generate
     strategy: depend
   needs:
     - artifacts: True
-      job: ml-cpu-protected-generate
+      job: ml-linux-x86_64-cpu-protected-generate
 
 ########################################
-# Machine Learning (CUDA)
+# Machine Learning - Linux x86_64 (CUDA)
 ########################################
-.ml-cuda:
+.ml-linux-x86_64-cuda:
   variables:
-    SPACK_CI_STACK_NAME: ml-cuda
+    SPACK_CI_STACK_NAME: ml-linux-x86_64-cuda
 
-.ml-cuda-generate:
+.ml-linux-x86_64-cuda-generate:
-  extends: .ml-cuda
+  extends: .ml-linux-x86_64-cuda
   image: ghcr.io/spack/e4s-amazonlinux-2:v2022-03-21
   tags: ["spack", "aws", "public", "medium", "x86_64_v4"]
 
-ml-cuda-pr-generate:
+ml-linux-x86_64-cuda-pr-generate:
-  extends: [ ".ml-cuda-generate", ".pr-generate"]
+  extends: [ ".ml-linux-x86_64-cuda-generate", ".pr-generate"]
 
-ml-cuda-protected-generate:
+ml-linux-x86_64-cuda-protected-generate:
-  extends: [ ".ml-cuda-generate", ".protected-generate"]
+  extends: [ ".ml-linux-x86_64-cuda-generate", ".protected-generate"]
 
-ml-cuda-pr-build:
+ml-linux-x86_64-cuda-pr-build:
-  extends: [ ".ml-cuda", ".pr-build" ]
+  extends: [ ".ml-linux-x86_64-cuda", ".pr-build" ]
   trigger:
     include:
       - artifact: jobs_scratch_dir/cloud-ci-pipeline.yml
-        job: ml-cuda-pr-generate
+        job: ml-linux-x86_64-cuda-pr-generate
     strategy: depend
   needs:
     - artifacts: True
-      job: ml-cuda-pr-generate
+      job: ml-linux-x86_64-cuda-pr-generate
 
-ml-cuda-protected-build:
+ml-linux-x86_64-cuda-protected-build:
-  extends: [ ".ml-cuda", ".protected-build" ]
+  extends: [ ".ml-linux-x86_64-cuda", ".protected-build" ]
   trigger:
     include:
       - artifact: jobs_scratch_dir/cloud-ci-pipeline.yml
-        job: ml-cuda-protected-generate
+        job: ml-linux-x86_64-cuda-protected-generate
     strategy: depend
   needs:
     - artifacts: True
-      job: ml-cuda-protected-generate
+      job: ml-linux-x86_64-cuda-protected-generate
 
 ########################################
-# Machine Learning (ROCm)
+# Machine Learning - Linux x86_64 (ROCm)
 ########################################
-.ml-rocm:
+.ml-linux-x86_64-rocm:
   variables:
-    SPACK_CI_STACK_NAME: ml-rocm
+    SPACK_CI_STACK_NAME: ml-linux-x86_64-rocm
 
-.ml-rocm-generate:
+.ml-linux-x86_64-rocm-generate:
-  extends: .ml-rocm
+  extends: .ml-linux-x86_64-rocm
   image: ghcr.io/spack/e4s-amazonlinux-2:v2022-03-21
   tags: ["spack", "aws", "public", "medium", "x86_64_v4"]
 
-ml-rocm-pr-generate:
+ml-linux-x86_64-rocm-pr-generate:
-  extends: [ ".ml-rocm-generate", ".pr-generate"]
+  extends: [ ".ml-linux-x86_64-rocm-generate", ".pr-generate"]
 
-ml-rocm-protected-generate:
+ml-linux-x86_64-rocm-protected-generate:
-  extends: [ ".ml-rocm-generate", ".protected-generate"]
+  extends: [ ".ml-linux-x86_64-rocm-generate", ".protected-generate"]
 
-ml-rocm-pr-build:
+ml-linux-x86_64-rocm-pr-build:
-  extends: [ ".ml-rocm", ".pr-build" ]
+  extends: [ ".ml-linux-x86_64-rocm", ".pr-build" ]
   trigger:
     include:
       - artifact: jobs_scratch_dir/cloud-ci-pipeline.yml
-        job: ml-rocm-pr-generate
+        job: ml-linux-x86_64-rocm-pr-generate
     strategy: depend
   needs:
     - artifacts: True
-      job: ml-rocm-pr-generate
+      job: ml-linux-x86_64-rocm-pr-generate
 
-ml-rocm-protected-build:
+ml-linux-x86_64-rocm-protected-build:
-  extends: [ ".ml-rocm", ".protected-build" ]
+  extends: [ ".ml-linux-x86_64-rocm", ".protected-build" ]
   trigger:
     include:
       - artifact: jobs_scratch_dir/cloud-ci-pipeline.yml
-        job: ml-rocm-protected-generate
+        job: ml-linux-x86_64-rocm-protected-generate
     strategy: depend
   needs:
     - artifacts: True
-      job: ml-rocm-protected-generate
+      job: ml-linux-x86_64-rocm-protected-generate

share/spack/gitlab/cloud_pipelines/stacks/ml-linux-x86_64-cpu/spack.yaml

@@ -16,77 +16,85 @@ spack:
 
   packages:
     all:
-      compiler: [gcc@11.2.0]
       target: [x86_64_v3]
       variants: ~cuda~rocm
 
+  definitions:
+    - packages:
+      # Horovod
+      - py-horovod
+
+      # Hugging Face
+      - py-transformers
+
+      # JAX
+      - py-jax
+      - py-jaxlib
+
+      # Keras
+      - py-keras
+      - py-keras-applications
+      - py-keras-preprocessing
+      - py-keras2onnx
+
+      # PyTorch
+      - py-botorch
+      - py-efficientnet-pytorch
+      - py-gpytorch
+      - py-kornia
+      - py-pytorch-gradual-warmup-lr
+      - py-pytorch-lightning
+      - py-segmentation-models-pytorch
+      - py-timm
+      - py-torch
+      - py-torch-cluster
+      - py-torch-geometric
+      - py-torch-nvidia-apex
+      - py-torch-scatter
+      - py-torch-sparse
+      - py-torch-spline-conv
+      - py-torchaudio
+      - py-torchdata
+      - py-torchfile
+      - py-torchgeo
+      - py-torchmeta
+      - py-torchmetrics
+      - py-torchtext
+      - py-torchvision
+      - py-vector-quantize-pytorch
+
+      # scikit-learn
+      - py-scikit-learn
+      - py-scikit-learn-extra
+
+      # TensorBoard
+      - py-tensorboard
+      - py-tensorboard-data-server
+      - py-tensorboard-plugin-wit
+      - py-tensorboardx
+
+      # TensorFlow
+      - py-tensorflow
+      - py-tensorflow-datasets
+      - py-tensorflow-estimator
+      - py-tensorflow-hub
+      - py-tensorflow-metadata
+      - py-tensorflow-probability
+
+      # XGBoost
+      - py-xgboost
+      # - r-xgboost
+      - xgboost
+
+    - arch:
+      - target=x86_64_v3
+
   specs:
-    # Horovod
+    - matrix:
-    - py-horovod
+      - [$packages]
+      - [$arch]
 
-    # Hugging Face
+  mirrors: { "mirror": "s3://spack-binaries/develop/ml-linux-x86_64-cpu" }
-    - py-transformers
-
-    # JAX
-    - py-jax
-    - py-jaxlib
-
-    # Keras
-    - py-keras
-    - py-keras-applications
-    - py-keras-preprocessing
-    - py-keras2onnx
-
-    # PyTorch
-    - py-botorch
-    - py-efficientnet-pytorch
-    - py-gpytorch
-    - py-kornia
-    - py-pytorch-gradual-warmup-lr
-    - py-pytorch-lightning
-    - py-segmentation-models-pytorch
-    - py-timm
-    - py-torch
-    - py-torch-cluster
-    - py-torch-geometric
-    - py-torch-nvidia-apex
-    - py-torch-scatter
-    - py-torch-sparse
-    - py-torch-spline-conv
-    - py-torchaudio
-    - py-torchdata
-    - py-torchfile
-    - py-torchgeo
-    - py-torchmeta
-    - py-torchmetrics
-    - py-torchtext
-    - py-torchvision
-    - py-vector-quantize-pytorch
-
-    # scikit-learn
-    - py-scikit-learn
-    - py-scikit-learn-extra
-
-    # TensorBoard
-    - py-tensorboard
-    - py-tensorboard-data-server
-    - py-tensorboard-plugin-wit
-    - py-tensorboardx
-
-    # TensorFlow
-    - py-tensorflow
-    - py-tensorflow-datasets
-    - py-tensorflow-estimator
-    - py-tensorflow-hub
-    - py-tensorflow-metadata
-    - py-tensorflow-probability
-
-    # XGBoost
-    - py-xgboost
-    # - r-xgboost
-    - xgboost
-
-  mirrors: { "mirror": "s3://spack-binaries/develop/ml-cpu" }
 
   gitlab-ci:
     script:
@@ -113,6 +121,7 @@ spack:
     mappings:
       - match:
           - llvm
+          - py-tensorflow
           - py-torch
         runner-attributes:
           tags: [ "spack", "huge", "x86_64_v4" ]

share/spack/gitlab/cloud_pipelines/stacks/ml-linux-x86_64-cuda/spack.yaml

@@ -16,80 +16,88 @@ spack:
 
   packages:
     all:
-      compiler: [gcc@11.2.0]
       target: [x86_64_v3]
       variants: ~rocm+cuda cuda_arch=80
     llvm:
       # https://github.com/spack/spack/issues/27999
       require: ~cuda
 
+  definitions:
+    - packages:
+      # Horovod
+      - py-horovod
+
+      # Hugging Face
+      - py-transformers
+
+      # JAX
+      - py-jax
+      - py-jaxlib
+
+      # Keras
+      - py-keras
+      - py-keras-applications
+      - py-keras-preprocessing
+      - py-keras2onnx
+
+      # PyTorch
+      - py-botorch
+      - py-efficientnet-pytorch
+      - py-gpytorch
+      - py-kornia
+      - py-pytorch-gradual-warmup-lr
+      - py-pytorch-lightning
+      - py-segmentation-models-pytorch
+      - py-timm
+      - py-torch
+      - py-torch-cluster
+      - py-torch-geometric
+      - py-torch-nvidia-apex
+      - py-torch-scatter
+      - py-torch-sparse
+      - py-torch-spline-conv
+      - py-torchaudio
+      - py-torchdata
+      - py-torchfile
+      - py-torchgeo
+      - py-torchmeta
+      - py-torchmetrics
+      - py-torchtext
+      - py-torchvision
+      - py-vector-quantize-pytorch
+
+      # scikit-learn
+      - py-scikit-learn
+      - py-scikit-learn-extra
+
+      # TensorBoard
+      - py-tensorboard
+      - py-tensorboard-data-server
+      - py-tensorboard-plugin-wit
+      - py-tensorboardx
+
+      # TensorFlow
+      - py-tensorflow
+      - py-tensorflow-datasets
+      - py-tensorflow-estimator
+      - py-tensorflow-hub
+      - py-tensorflow-metadata
+      - py-tensorflow-probability
+
+      # XGBoost
+      - py-xgboost
+      # - r-xgboost
+      - xgboost
+
+    - arch:
+      - target=x86_64_v3
+
   specs:
-    # Horovod
+    - matrix:
-    - py-horovod
+      - [$packages]
+      - [$arch]
 
-    # Hugging Face
+  mirrors: { "mirror": "s3://spack-binaries/develop/ml-linux-x86_64-cuda" }
-    - py-transformers
-
-    # JAX
-    - py-jax
-    - py-jaxlib
-
-    # Keras
-    - py-keras
-    - py-keras-applications
-    - py-keras-preprocessing
-    - py-keras2onnx
-
-    # PyTorch
-    - py-botorch
-    - py-efficientnet-pytorch
-    - py-gpytorch
-    - py-kornia
-    - py-pytorch-gradual-warmup-lr
-    - py-pytorch-lightning
-    - py-segmentation-models-pytorch
-    - py-timm
-    - py-torch
-    - py-torch-cluster
-    - py-torch-geometric
-    - py-torch-nvidia-apex
-    - py-torch-scatter
-    - py-torch-sparse
-    - py-torch-spline-conv
-    - py-torchaudio
-    - py-torchdata
-    - py-torchfile
-    - py-torchgeo
-    - py-torchmeta
-    - py-torchmetrics
-    - py-torchtext
-    - py-torchvision
-    - py-vector-quantize-pytorch
-
-    # scikit-learn
-    - py-scikit-learn
-    - py-scikit-learn-extra
-
-    # TensorBoard
-    - py-tensorboard
-    - py-tensorboard-data-server
-    - py-tensorboard-plugin-wit
-    - py-tensorboardx
-
-    # TensorFlow
-    - py-tensorflow
-    - py-tensorflow-datasets
-    - py-tensorflow-estimator
-    - py-tensorflow-hub
-    - py-tensorflow-metadata
-    - py-tensorflow-probability
-
-    # XGBoost
-    - py-xgboost
-    # - r-xgboost
-    - xgboost
-
-  mirrors: { "mirror": "s3://spack-binaries/develop/ml-cuda" }
 
   gitlab-ci:
     script:
@@ -116,6 +124,7 @@ spack:
     mappings:
       - match:
           - llvm
+          - py-tensorflow
           - py-torch
         runner-attributes:
           tags: [ "spack", "huge", "x86_64_v4" ]

share/spack/gitlab/cloud_pipelines/stacks/ml-linux-x86_64-rocm/spack.yaml

@@ -16,7 +16,6 @@ spack:
 
   packages:
     all:
-      compiler: [gcc@11.2.0]
       target: [x86_64_v3]
       variants: ~cuda+rocm amdgpu_target=gfx90a
     gl:
@@ -25,74 +24,83 @@ spack:
       # Does not yet support Spack-installed ROCm
       require: ~rocm
 
+  definitions:
+    - packages:
+      # Horovod
+      - py-horovod
+
+      # Hugging Face
+      - py-transformers
+
+      # JAX
+      - py-jax
+      - py-jaxlib
+
+      # Keras
+      - py-keras
+      - py-keras-applications
+      - py-keras-preprocessing
+      - py-keras2onnx
+
+      # PyTorch
+      # Does not yet support Spack-install ROCm
+      # - py-botorch
+      # - py-efficientnet-pytorch
+      # - py-gpytorch
+      # - py-kornia
+      # - py-pytorch-gradual-warmup-lr
+      # - py-pytorch-lightning
+      # - py-segmentation-models-pytorch
+      # - py-timm
+      # - py-torch
+      # - py-torch-cluster
+      # - py-torch-geometric
+      # - py-torch-nvidia-apex
+      # - py-torch-scatter
+      # - py-torch-sparse
+      # - py-torch-spline-conv
+      # - py-torchaudio
+      # - py-torchdata
+      # - py-torchfile
+      # - py-torchgeo
+      # - py-torchmeta
+      # - py-torchmetrics
+      # - py-torchtext
+      # - py-torchvision
+      # - py-vector-quantize-pytorch
+
+      # scikit-learn
+      - py-scikit-learn
+      - py-scikit-learn-extra
+
+      # TensorBoard
+      - py-tensorboard
+      - py-tensorboard-data-server
+      - py-tensorboard-plugin-wit
+      - py-tensorboardx
+
+      # TensorFlow
+      - py-tensorflow
+      - py-tensorflow-datasets
+      - py-tensorflow-estimator
+      - py-tensorflow-hub
+      - py-tensorflow-metadata
+      - py-tensorflow-probability
+
+      # XGBoost
+      - py-xgboost
+      # - r-xgboost
+      - xgboost
+
+    - arch:
+      - target=x86_64_v3
+
   specs:
-    # Horovod
+    - matrix:
-    - py-horovod
+      - [$packages]
+      - [$arch]
 
-    # Hugging Face
+  mirrors: { "mirror": "s3://spack-binaries/develop/ml-linux-x86_64-rocm" }
-    - py-transformers
-
-    # JAX
-    - py-jax
-    - py-jaxlib
-
-    # Keras
-    - py-keras
-    - py-keras-applications
-    - py-keras-preprocessing
-    - py-keras2onnx
-
-    # PyTorch
-    # Does not yet support Spack-install ROCm
-    # - py-botorch
-    # - py-efficientnet-pytorch
-    # - py-gpytorch
-    # - py-kornia
-    # - py-pytorch-gradual-warmup-lr
-    # - py-pytorch-lightning
-    # - py-segmentation-models-pytorch
-    # - py-timm
-    # - py-torch
-    # - py-torch-cluster
-    # - py-torch-geometric
-    # - py-torch-nvidia-apex
-    # - py-torch-scatter
-    # - py-torch-sparse
-    # - py-torch-spline-conv
-    # - py-torchaudio
-    # - py-torchdata
-    # - py-torchfile
-    # - py-torchgeo
-    # - py-torchmeta
-    # - py-torchmetrics
-    # - py-torchtext
-    # - py-torchvision
-    # - py-vector-quantize-pytorch
-
-    # scikit-learn
-    - py-scikit-learn
-    - py-scikit-learn-extra
-
-    # TensorBoard
-    - py-tensorboard
-    - py-tensorboard-data-server
-    - py-tensorboard-plugin-wit
-    - py-tensorboardx
-
-    # TensorFlow
-    - py-tensorflow
-    - py-tensorflow-datasets
-    - py-tensorflow-estimator
-    - py-tensorflow-hub
-    - py-tensorflow-metadata
-    - py-tensorflow-probability
-
-    # XGBoost
-    - py-xgboost
-    # - r-xgboost
-    - xgboost
-
-  mirrors: { "mirror": "s3://spack-binaries/develop/ml-rocm" }
 
   gitlab-ci:
     script:
@@ -118,8 +126,9 @@ spack:
     match_behavior: first
     mappings:
       - match:
-          - llvm-amdgpu
           - llvm
+          - llvm-amdgpu
+          - py-tensorflow
           - py-torch
           - rocblas
         runner-attributes: