Update .github/workflows to be more granular

In most of the cases, splitting those is for the best and to reduce execution time, assuming we tend to update the DLCs one at a time, so it's not really probable for all the containers to change at once. Pros: easier to manage, more granular, no need for extra `docker pull`s, just runs what's modified Cons: when modifying a bunch of tests it will be slower as `docker pull` needs to be done per each test as instances are ephemeral
huggingface · Sep 9, 2024 · cb7ddb6 · cb7ddb6
1 parent b71a392
commit cb7ddb6
Show file tree

Hide file tree

Showing 5 changed files with 133 additions and 13 deletions.
diff --git a/.github/workflows/run-tests-reusable.yml b/.github/workflows/run-tests-reusable.yml
@@ -10,6 +10,15 @@ on:
         options:
           - aws-general-8-plus
           - aws-g4dn-2xlarge
+      tests-path:
+        description: "The path of the tests to run inside `tests`."
+        required: true
+        type: string
+        options:
+          - pytorch/training
+          - pytorch/inference
+          - tgi
+          - tei
       training-dlc:
         description: "The URI of the Hugging Face PyTorch DLC for Training (GPU only)."
         required: false
@@ -54,7 +63,7 @@ jobs:
 
       - name: Run Hugging Face DLC Tests
         if:
-        run: uv run pytest -s tests/ --basetemp=${{ runner.temp }}
+        run: uv run pytest -s tests/${{ inputs.tests-path }} --basetemp=${{ runner.temp }}
         env:
           TRAINING_DLC: ${{ inputs.training-dlc }}
           INFERENCE_DLC: ${{ inputs.inference-dlc }}

diff --git a/.github/workflows/test-huggingface-dlcs.yml → ...workflows/test-pytorch-inference-dlcs.yml b/.github/workflows/test-huggingface-dlcs.yml → ...workflows/test-pytorch-inference-dlcs.yml
@@ -1,41 +1,42 @@
-name: Test Hugging Face DLCs
+name: Test Hugging Face PyTorch DLCs for Inference
 
 on:
   push:
     branches:
       - main
+    paths:
+      - tests/pytorch/inference/*
+      - .github/workflows/run-tests-reusable.yml
+      - .github/workflows/test-pytorch-inference-dlcs.yml
   pull_request:
     types:
       - synchronize
       - ready_for_review
     branches:
       - main
     paths:
-      - tests/*
-      - pytest.ini
+      - tests/pytorch/inference/*
       - .github/workflows/run-tests-reusable.yml
-      - .github/workflows/test-huggingface-dlcs.yml
+      - .github/workflows/test-pytorch-inference-dlcs.yml
   workflow_dispatch:
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
   cancel-in-progress: true
 
 jobs:
-  dlcs-on-cpu:
-    name: Run Hugging Face DLCs Tests on CPU
+  inference-on-cpu:
+    name: Test Hugging Face PyTorch DLCs for Inference on CPU
     uses: huggingface/Google-Cloud-Containers/.github/workflows/run-tests-reusable.yml@add-integration-tests
     with:
       group: aws-general-8-plus
+      tests-path: pytorch/inference
       inference-dlc: us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-pytorch-inference-cpu.2-2.transformers.4-44.ubuntu2204.py311
-      tei-dlc: us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-embeddings-inference-cpu.1-2
 
-  dlcs-on-gpu:
-    name: Run Hugging Face DLCs Tests on GPU
+  inference-on-gpu:
+    name: Test Hugging Face PyTorch DLCs for Inference on GPU
     uses: huggingface/Google-Cloud-Containers/.github/workflows/run-tests-reusable.yml@add-integration-tests
     with:
       group: aws-g4dn-2xlarge
-      training-dlc: us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-pytorch-training-cu121.transformers.4-42.ubuntu2204.py310
+      tests-path: pytorch/inference
       inference-dlc: us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-pytorch-inference-cu121.2-2.transformers.4-44.ubuntu2204.py311
-      tgi-dlc: us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-generation-inference-cu121.2-2.ubuntu2204.py310
-      tei-dlc: us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-embeddings-inference-cu122.1-4.ubuntu2204
diff --git a/.github/workflows/test-pytorch-training-dlcs.yml b/.github/workflows/test-pytorch-training-dlcs.yml
@@ -0,0 +1,34 @@
+name: Test Hugging Face PyTorch DLCs for Training
+
+on:
+  push:
+    branches:
+      - main
+    paths:
+      - tests/pytorch/training/*
+      - .github/workflows/run-tests-reusable.yml
+      - .github/workflows/test-pytorch-training-dlcs.yml
+  pull_request:
+    types:
+      - synchronize
+      - ready_for_review
+    branches:
+      - main
+    paths:
+      - tests/pytorch/training/*
+      - .github/workflows/run-tests-reusable.yml
+      - .github/workflows/test-pytorch-training-dlcs.yml
+  workflow_dispatch:
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
+jobs:
+  training-on-gpu:
+    name: Test Hugging Face PyTorch DLCs for Training on GPU
+    uses: huggingface/Google-Cloud-Containers/.github/workflows/run-tests-reusable.yml@add-integration-tests
+    with:
+      group: aws-g4dn-2xlarge
+      tests-path: pytorch/training
+      training-dlc: us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-pytorch-training-cu121.transformers.4-42.ubuntu2204.py310
diff --git a/.github/workflows/test-text-embeddings-inference-dlcs.yml b/.github/workflows/test-text-embeddings-inference-dlcs.yml
@@ -0,0 +1,42 @@
+name: Test Hugging Face DLCs for TEI (CPU and GPU)
+
+on:
+  push:
+    branches:
+      - main
+    paths:
+      - tests/tei/*
+      - .github/workflows/run-tests-reusable.yml
+      - .github/workflows/test-text-embeddings-inference-dlcs.yml
+  pull_request:
+    types:
+      - synchronize
+      - ready_for_review
+    branches:
+      - main
+    paths:
+      - tests/tei/*
+      - .github/workflows/run-tests-reusable.yml
+      - .github/workflows/test-text-embeddings-inference-dlcs.yml
+  workflow_dispatch:
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
+jobs:
+  tei-on-cpu:
+    name: Test Hugging Face DLCs for TEI on CPU
+    uses: huggingface/Google-Cloud-Containers/.github/workflows/run-tests-reusable.yml@add-integration-tests
+    with:
+      group: aws-general-8-plus
+      tests-path: tei
+      tei-dlc: us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-embeddings-inference-cpu.1-2
+
+  tei-on-gpu:
+    name: Test Hugging Face DLCs for TEI on GPU
+    uses: huggingface/Google-Cloud-Containers/.github/workflows/run-tests-reusable.yml@add-integration-tests
+    with:
+      group: aws-g4dn-2xlarge
+      tests-path: tei
+      tei-dlc: us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-embeddings-inference-cu122.1-4.ubuntu2204
diff --git a/.github/workflows/test-text-generation-inference-dlcs.yml b/.github/workflows/test-text-generation-inference-dlcs.yml
@@ -0,0 +1,34 @@
+name: Test Hugging Face DLCs for TGI (GPU)
+
+on:
+  push:
+    branches:
+      - main
+    paths:
+      - tests/tgi/*
+      - .github/workflows/run-tests-reusable.yml
+      - .github/workflows/test-text-generation-inference-dlcs.yml
+  pull_request:
+    types:
+      - synchronize
+      - ready_for_review
+    branches:
+      - main
+    paths:
+      - tests/tgi/*
+      - .github/workflows/run-tests-reusable.yml
+      - .github/workflows/test-text-generation-inference-dlcs.yml
+  workflow_dispatch:
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
+jobs:
+  tgi-on-gpu:
+    name: Test Hugging Face DLCs for TGI on GPU
+    uses: huggingface/Google-Cloud-Containers/.github/workflows/run-tests-reusable.yml@add-integration-tests
+    with:
+      group: aws-g4dn-2xlarge
+      tests-path: tgi
+      tgi-dlc: us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-generation-inference-cu121.2-2.ubuntu2204.py310