Auto build moderator (#717)

Author: Dreamsorcerer

.github/workflows/post-dependabot.yml

@@ -0,0 +1,51 @@
+name: Dependabot post-update (rebuild moderator model)
+on:
+  pull_request_target:
+    action: [opened, synchronize, reopened]
+    branches:
+      - 'master'
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number }}
+  cancel-in-progress: true
+
+permissions: {}
+jobs:
+  post-update:
+    permissions:
+      pull-requests: read # for gh pr checkout
+      contents: write # to push code in repo (stefanzweifel/git-auto-commit-action)
+
+    if: github.actor == 'dependabot[bot]' && contains(github.event.pull_request.title, 'scikit-learn')
+    runs-on: ubuntu-latest
+    steps:
+    - name: Generate token
+      id: generate_token
+      uses: tibdex/github-app-token@v1
+      with:
+        app_id: ${{ secrets.BOT_APP_ID }}
+        private_key: ${{ secrets.BOT_PRIVATE_KEY }}
+    - name: Checkout
+      uses: actions/checkout@v3
+      with:
+        token: ${{ steps.generate_token.outputs.token }}
+    - name: Login
+      run: |
+        echo "${{ secrets.GITHUB_TOKEN }}" | gh auth login --with-token
+    - name: Checkout
+      run: |
+        gh pr checkout ${{ github.event.pull_request.number }}
+    - name: Setup Python
+      uses: actions/setup-python@v4
+      cache: 'pip'
+      cache-dependency-path: '**/requirements*.txt'
+    - name: Download train.csv
+      # Original file can be found at: https://www.kaggle.com/code/piumallick/toxic-comments-sentiment-analysis/input?select=train.csv.zip
+      run: curl "$MODERATOR_TRAINING_URL" > train.csv
+    - name: Rebuild model
+      run: PYTHONPATH='demos/moderator' python -m moderator.build_model train.csv
+    - name: Commit and push if needed
+      uses: stefanzweifel/git-auto-commit-action@v4
+      with:
+        commit_message: Rebuild moderator model
+        file_pattern: 'demos/moderator/model/pipeline.dat'

demos/moderator/moderator/build_model.py

@@ -1,8 +1,11 @@
+import sys
+from pathlib import Path
+
 from moderator.consts import PROJ_ROOT
 from moderator.model.pipeline import build_model
 
 
 if __name__ == '__main__':
-    dataset_path = PROJ_ROOT / 'moderator' / 'model' / 'data' / 'train.csv'
+    dataset_path = Path(sys.argv[1])
     model_path = PROJ_ROOT / 'model'
     build_model(dataset_path, model_path)

demos/moderator/moderator/model/pipeline.py

@@ -1,6 +1,8 @@
+import pickle
+from pathlib import Path
+
 import pandas as pd
 
-import pickle
 from sklearn.feature_extraction.text import TfidfVectorizer
 from sklearn.linear_model import LogisticRegression
 from sklearn.multioutput import MultiOutputClassifier
@@ -38,11 +40,12 @@ def build_pipeline():
     return pipeline
 
 
-def build_model(dataset_path, model_path):
+def build_model(dataset_path: Path, model_path: Path) -> None:
     train, targets = read_data(dataset_path)
 
     pipeline = build_pipeline()
     pipeline.fit(train, targets)
 
-    with open("pipeline.dat", "wb") as f:
+    output_path = model_path / "pipeline.dat"
+    with output_path.open("wb") as f:
         pickle.dump(pipeline, f)

demos/moderator_bot/model/pipeline.dat

@@ -0,0 +1 @@
+../../moderator/model/pipeline.dat