Add auto retries for Captum OSS GitHub Actions (#1408)

Summary:

We frequently see sporadic failures in Captum GitHub actions test workflows, often related to package download, http errors, conda environment setup, etc.

We add auto-retries to automatically retry failed workflows rather than needing to do this manually.

Differential Revision: D64693773

Author: Vivek Miglani

.github/workflows/retry.yml

@@ -0,0 +1,19 @@
+name: Retry Test
+on:
+  workflow_dispatch:
+    inputs:
+      run_id:
+        required: true
+jobs:
+  rerun-on-failure:
+    permissions: write-all
+    runs-on: ubuntu-latest
+    steps:
+      - name: rerun ${{ inputs.run_id }}
+        env:
+          GH_REPO: ${{ github.repository }}
+          GH_TOKEN: ${{ github.token }}
+          GH_DEBUG: api
+        run: |
+          gh run watch ${{ inputs.run_id }} > /dev/null 2>&1
+          gh run rerun ${{ inputs.run_id }} --failed

.github/workflows/test-conda-cpu.yml

@@ -1,15 +1,15 @@
 name: Unit-tests for Conda install
 
 on:
-  pull_request:
-  push:
-    branches:
-      - master
+pull_request:
+push:
+  branches:
+    - master
 
-  workflow_dispatch:
+workflow_dispatch:
 
 env:
-  CHANNEL: "nightly"
+CHANNEL: "nightly"
 
 jobs:
   tests:
@@ -32,3 +32,18 @@ jobs:
 
         # Run Tests
         python3 -m pytest -ra --cov=. --cov-report term-missing
+
+  auto-retry:
+    name: Auto retry on failure
+    needs: tests
+    if: failure() && fromJSON(github.run_attempt) < 2
+    runs-on: ubuntu-latest
+    steps:
+      - name: Start rerun workflow
+        env:
+          GH_REPO: ${{ github.repository }}
+          GH_TOKEN: ${{ github.token }}
+          GH_DEBUG: api
+        run: |
+          gh workflow run retry.yml \
+            -F run_id=${{ github.run_id }}

.github/workflows/test-pip-cpu-with-mypy.yml

@@ -25,3 +25,18 @@ jobs:
         ./scripts/run_mypy.sh
         # Run Tests
         python3 -m pytest -ra --cov=. --cov-report term-missing
+
+  auto-retry:
+    name: Auto retry on failure
+    needs: tests
+    if: failure() && fromJSON(github.run_attempt) < 2
+    runs-on: ubuntu-latest
+    steps:
+      - name: Start rerun workflow
+        env:
+          GH_REPO: ${{ github.repository }}
+          GH_TOKEN: ${{ github.token }}
+          GH_DEBUG: api
+        run: |
+          gh workflow run retry.yml \
+            -F run_id=${{ github.run_id }}

.github/workflows/test-pip-cpu.yml

@@ -35,3 +35,18 @@ jobs:
         ./scripts/install_via_pip.sh ${{ matrix.pytorch_args }} ${{ matrix.transformers_args }}
         # Run Tests
         python3 -m pytest -ra --cov=. --cov-report term-missing
+
+  auto-retry:
+    name: Auto retry on failure
+    needs: tests
+    if: failure() && fromJSON(github.run_attempt) < 2
+    runs-on: ubuntu-latest
+    steps:
+      - name: Start rerun workflow
+        env:
+          GH_REPO: ${{ github.repository }}
+          GH_TOKEN: ${{ github.token }}
+          GH_DEBUG: api
+        run: |
+          gh workflow run retry.yml \
+            -F run_id=${{ github.run_id }}

.github/workflows/test-pip-gpu.yml

@@ -30,3 +30,18 @@ jobs:
 
         # Run Tests
         python3 -m pytest -ra --cov=. --cov-report term-missing
+
+  auto-retry:
+    name: Auto retry on failure
+    needs: tests
+    if: failure() && fromJSON(github.run_attempt) < 2
+    runs-on: ubuntu-latest
+    steps:
+      - name: Start rerun workflow
+        env:
+          GH_REPO: ${{ github.repository }}
+          GH_TOKEN: ${{ github.token }}
+          GH_DEBUG: api
+        run: |
+          gh workflow run retry.yml \
+            -F run_id=${{ github.run_id }}