From 2aea5fd33145e82bef9aa02ab3a3819a4125e052 Mon Sep 17 00:00:00 2001 From: Guillaume LEGENDRE Date: Thu, 26 Sep 2024 17:47:19 +0200 Subject: [PATCH 1/5] change default runner --- .github/workflows/integration-test-action.yaml | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/.github/workflows/integration-test-action.yaml b/.github/workflows/integration-test-action.yaml index 2e3479fc..9b871d6a 100644 --- a/.github/workflows/integration-test-action.yaml +++ b/.github/workflows/integration-test-action.yaml @@ -39,11 +39,12 @@ on: runs_on: type: string required: false - default: '["single-gpu", "nvidia-gpu", "t4", "ci"]' + default: 'aws-g4dn-2xlarge' jobs: pytorch-integration-tests: - runs-on: ${{ fromJson(inputs.runs_on) }} + runs-on: + group: ${{ inputs.runs_on }} env: AWS_REGION: ${{ inputs.region }} HF_HOME: ${{ inputs.hf_home }} @@ -66,4 +67,4 @@ jobs: python -m pytest \ ${{ inputs.test_path }} -n ${{ inputs.test_parallelism }} \ --log-cli-level='${{ inputs.log_level }}' \ - --log-format='${{ inputs.log_format }}' \ No newline at end of file + --log-format='${{ inputs.log_format }}' From 6764955e1c52a070be09dfe34e5a6eab5b7c6f53 Mon Sep 17 00:00:00 2001 From: Guillaume LEGENDRE Date: Thu, 26 Sep 2024 17:48:09 +0200 Subject: [PATCH 2/5] remove run on param --- .github/workflows/integration-test.yaml | 2 -- 1 file changed, 2 deletions(-) diff --git a/.github/workflows/integration-test.yaml b/.github/workflows/integration-test.yaml index 5343b571..f4afbea7 100644 --- a/.github/workflows/integration-test.yaml +++ b/.github/workflows/integration-test.yaml @@ -35,7 +35,6 @@ jobs: with: test_path: "tests/integ/test_pytorch_remote_cpu.py" build_img_cmd: "make inference-pytorch-cpu" - runs_on: "['ci']" pytorch-integration-local-cpu: name: Local Integration Tests - CPU uses: ./.github/workflows/integration-test-action.yaml @@ -43,4 +42,3 @@ jobs: test_path: "tests/integ/test_pytorch_local_cpu.py" build_img_cmd: "make inference-pytorch-cpu" test_parallelism: "1" - runs_on: "['ci']" \ No newline at end of file From 799a5994a05d5bc431da9967b5cc5d588cebe37f Mon Sep 17 00:00:00 2001 From: Guillaume LEGENDRE Date: Thu, 26 Sep 2024 17:50:01 +0200 Subject: [PATCH 3/5] add cache --- .github/workflows/integration-test-action.yaml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.github/workflows/integration-test-action.yaml b/.github/workflows/integration-test-action.yaml index 9b871d6a..c6665d6b 100644 --- a/.github/workflows/integration-test-action.yaml +++ b/.github/workflows/integration-test-action.yaml @@ -39,7 +39,7 @@ on: runs_on: type: string required: false - default: 'aws-g4dn-2xlarge' + default: 'aws-g4dn-2xlarge-cache' jobs: pytorch-integration-tests: From dd7e01da3b5338abc1c8ee915688b7c8fed4a6e0 Mon Sep 17 00:00:00 2001 From: Guillaume LEGENDRE Date: Thu, 26 Sep 2024 17:51:58 +0200 Subject: [PATCH 4/5] change unit test runner --- .github/workflows/unit-test.yaml | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/.github/workflows/unit-test.yaml b/.github/workflows/unit-test.yaml index 0c8db252..5a77c035 100644 --- a/.github/workflows/unit-test.yaml +++ b/.github/workflows/unit-test.yaml @@ -19,7 +19,8 @@ concurrency: jobs: pytorch-unit-test: - runs-on: [single-gpu, nvidia-gpu, t4, ci] + runs-on: + group: aws-g4dn-2xlarge-cache env: AWS_REGION: us-east-1 CACHE_TEST_DIR: /mnt/hf_cache/hf-inference-toolkit-tests From 507831e054a65410552105bdf590f8abf6de5c6a Mon Sep 17 00:00:00 2001 From: Guillaume LEGENDRE Date: Thu, 26 Sep 2024 18:11:21 +0200 Subject: [PATCH 5/5] fix hf cache permissions --- .github/workflows/unit-test.yaml | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/.github/workflows/unit-test.yaml b/.github/workflows/unit-test.yaml index 5a77c035..8b929cba 100644 --- a/.github/workflows/unit-test.yaml +++ b/.github/workflows/unit-test.yaml @@ -29,8 +29,9 @@ jobs: - uses: actions/checkout@v4.1.1 - name: Copy unit tests to cache mount run: | - rm -rf ${{ env.CACHE_TEST_DIR }} && \ - mkdir ${{ env.CACHE_TEST_DIR }} && \ + sudo rm -rf ${{ env.CACHE_TEST_DIR }} && \ + sudo mkdir ${{ env.CACHE_TEST_DIR }} && \ + sudo chown -R runner ${{ env.CACHE_TEST_DIR }} cp -r tests ${{ env.CACHE_TEST_DIR }} - name: Docker Setup Buildx uses: docker/setup-buildx-action@v3.0.0