test(workflow): add workflow for norm_weight_test (#70)

Co-authored-by: kkscilife <wangmengke@pjlab.org.cn>
InternLM · Mar 14, 2024 · e9fcf55 · e9fcf55
1 parent e82594c
commit e9fcf55
Show file tree

Hide file tree

Showing 4 changed files with 77 additions and 3 deletions.
diff --git a/.github/workflows/monthly_test.yaml b/.github/workflows/monthly_test.yaml
@@ -0,0 +1,74 @@
+name: monthly-tests
+on:
+  workflow_dispatch:
+  schedule:
+    - cron:  '32 17 9 * *'
+env:
+  WORKSPACE_PREFIX: $(echo $GITHUB_WORKSPACE |cut -d '/' -f 1-4)
+  SLURM_PARTITION: llm_s
+
+jobs:
+  training_tp_norm_layer_msp:
+    runs-on: [t_cluster]
+    steps:
+    - name: mask env
+      run: |
+        echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
+        echo "::add-mask::$path_prefix"
+    - uses: actions/checkout@v3
+      with:
+         ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
+
+    - name: training_tp_norm_layer_msp
+      run: |
+        source activate ${evo_env_torch21_flash2}
+        srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=monthly-test-${GITHUB_RUN_ID}-${GITHUB_JOB} -N 1 -n 8 --gres=gpu:8 pytest -s -v --color=yes -m "check_norm_msp" ./tests/test_training/test_norm_weight.py
+
+  training_tp_norm_layer_fsp:
+    runs-on: [t_cluster]
+    steps:
+    - name: mask env
+      run: |
+        echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
+        echo "::add-mask::$path_prefix"
+    - uses: actions/checkout@v3
+      with:
+         ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
+
+    - name: training_tp_norm_layer_fsp
+      run: |
+        source activate ${evo_env_torch21_flash2}
+        srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=monthly-test-${GITHUB_RUN_ID}-${GITHUB_JOB} -N 1 -n 8 --gres=gpu:8 pytest -s -v --color=yes -m "check_norm_fsp" ./tests/test_training/test_norm_weight.py
+
+  training_tp_norm_layer_isp:
+    runs-on: [t_cluster]
+    steps:
+    - name: mask env
+      run: |
+        echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
+        echo "::add-mask::$path_prefix"
+    - uses: actions/checkout@v3
+      with:
+         ref: ${{ github.event_name == 'schedule' && 'develop' || github.event_name == 'workflow_dispatch' && '' }}
+
+    - name: training_tp_norm_layer_isp
+      run: |
+        source activate ${evo_env_torch21_flash2}
+        srun -p ${SLURM_PARTITION} --kill-on-bad-exit=1 --job-name=monthly-test-${GITHUB_RUN_ID}-${GITHUB_JOB} -N 1 -n 8 --gres=gpu:8 pytest -s -v --color=yes -m "check_norm_isp" ./tests/test_training/test_norm_weight.py
+
+  notify_to_feishu:
+    if: ${{ always() && !cancelled() && contains(needs.*.result, 'failure') && (github.ref_name == 'develop' || github.ref_name == 'main') }}
+    needs: [
+      training_tp_norm_layer_msp,
+      training_tp_norm_layer_fsp,
+      training_tp_norm_layer_isp
+      ]
+    runs-on: [t_cluster]
+    steps:
+    - name: mask env
+      run: |
+        echo "::add-mask::${{env.WORKSPACE_PREFIX}}"
+        echo "::add-mask::$path_prefix"
+    - name: notify
+      run: |
+        curl -X POST -H "Content-Type: application/json" -d '{"msg_type":"post","content":{"post":{"zh_cn":{"title":"'${{ github.repository }}' GitHubAction Failed","content":[[{"tag":"text","text":""},{"tag":"a","text":"Please click here for details ","href":"https://github.com/'${{ github.repository }}'/actions/runs/'${GITHUB_RUN_ID}'"},{"tag":"at","user_id":"'${{ secrets.USER_ID }}'"}]]}}}}'  ${{ secrets.WEBHOOK_URL }}
diff --git a/.github/workflows/pr_merged.yaml b/.github/workflows/pr_merged.yaml
@@ -58,4 +58,4 @@ jobs:
         echo "::add-mask::$path_prefix"
     - name: notify
       run: |
-        curl -X POST -H "Content-Type: application/json" -d '{"msg_type":"post","content":{"post":{"zh_cn":{"title":"InternEvo GitHubAction Failed","content":[[{"tag":"text","text":""},{"tag":"a","text":"Please click here for details ","href":"https://github.com/'${{ github.repository }}'/actions/runs/'${GITHUB_RUN_ID}'"},{"tag":"at","user_id":"'${{ secrets.USER_ID }}'"}]]}}}}'  ${{ secrets.WEBHOOK_URL }}
+        curl -X POST -H "Content-Type: application/json" -d '{"msg_type":"post","content":{"post":{"zh_cn":{"title":"'${{ github.repository }}' GitHubAction Failed","content":[[{"tag":"text","text":""},{"tag":"a","text":"Please click here for details ","href":"https://github.com/'${{ github.repository }}'/actions/runs/'${GITHUB_RUN_ID}'"},{"tag":"at","user_id":"'${{ secrets.USER_ID }}'"}]]}}}}'  ${{ secrets.WEBHOOK_URL }}
diff --git a/.github/workflows/unit_tests.yaml b/.github/workflows/unit_tests.yaml
@@ -133,4 +133,4 @@ jobs:
         echo "::add-mask::$path_prefix"
     - name: notify
       run: |
-        curl -X POST -H "Content-Type: application/json" -d '{"msg_type":"post","content":{"post":{"zh_cn":{"title":"InternEvo GitHubAction Failed","content":[[{"tag":"text","text":""},{"tag":"a","text":"Please click here for details ","href":"https://github.com/'${{ github.repository }}'/actions/runs/'${GITHUB_RUN_ID}'"},{"tag":"at","user_id":"'${{ secrets.USER_ID }}'"}]]}}}}'  ${{ secrets.WEBHOOK_URL }}
+        curl -X POST -H "Content-Type: application/json" -d '{"msg_type":"post","content":{"post":{"zh_cn":{"title":"'${{ github.repository }}' GitHubAction Failed","content":[[{"tag":"text","text":""},{"tag":"a","text":"Please click here for details ","href":"https://github.com/'${{ github.repository }}'/actions/runs/'${GITHUB_RUN_ID}'"},{"tag":"at","user_id":"'${{ secrets.USER_ID }}'"}]]}}}}'  ${{ secrets.WEBHOOK_URL }}
diff --git a/.github/workflows/weekly_test.yaml b/.github/workflows/weekly_test.yaml
@@ -272,4 +272,4 @@ jobs:
         echo "::add-mask::$path_prefix"
     - name: notify
       run: |
-        curl -X POST -H "Content-Type: application/json" -d '{"msg_type":"post","content":{"post":{"zh_cn":{"title":"InternEvo GitHubAction Failed","content":[[{"tag":"text","text":""},{"tag":"a","text":"Please click here for details ","href":"https://github.com/'${{ github.repository }}'/actions/runs/'${GITHUB_RUN_ID}'"},{"tag":"at","user_id":"'${{ secrets.USER_ID }}'"}]]}}}}'  ${{ secrets.WEBHOOK_URL }}
+        curl -X POST -H "Content-Type: application/json" -d '{"msg_type":"post","content":{"post":{"zh_cn":{"title":"'${{ github.repository }}' GitHubAction Failed","content":[[{"tag":"text","text":""},{"tag":"a","text":"Please click here for details ","href":"https://github.com/'${{ github.repository }}'/actions/runs/'${GITHUB_RUN_ID}'"},{"tag":"at","user_id":"'${{ secrets.USER_ID }}'"}]]}}}}'  ${{ secrets.WEBHOOK_URL }}