Merge pull request #1 from YakobusIP/main

.github/workflows/ci-production.yml

@@ -0,0 +1,61 @@
+name: Deploy to Google Vertex AI
+
+on:
+  push:
+    branches:
+      - production
+
+jobs:
+  setup-build-publish-deploy:
+    name: Setup, Build, Publish, and Deploy
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v2
+
+      - name: Set up Cloud SDK
+        uses: google-github-actions/[email protected]
+        with:
+          project_id: ${{ secrets.GCP_PROJECT_ID }}
+          service_account_key: ${{ secrets.GCP_DEPLOYER_SA_KEY }}
+          export_default_credentials: true
+
+      - name: Authenticate Docker
+        run: |
+          gcloud auth configure-docker ${{ secrets.GCP_REPO_REGION }}-docker.pkg.dev --quiet
+
+      - name: Build image
+        working-directory: ./
+        run: docker build . --file Dockerfile --tag ${{ secrets.GCP_REPO_REGION }}-docker.pkg.dev/${{ secrets.GCP_PROJECT_ID }}/interview-ai-detector/model-prediction
+
+      - name: Push image
+        run: docker push ${{ secrets.GCP_REPO_REGION }}-docker.pkg.dev/${{ secrets.GCP_PROJECT_ID }}/interview-ai-detector/model-prediction
+
+      - name: Create model
+        id: create_model
+        run: |
+          MODEL_ID=$(gcloud ai models upload \
+            --region=${{ secrets.GCP_VERTEX_AI_REGION }} \
+            --display-name="interview-ai-detector-model" \
+            --container-image-uri="${{ secrets.GCP_REPO_REGION }}-docker.pkg.dev/${{ secrets.GCP_PROJECT_ID }}/interview-ai-detector/model-prediction:latest" \
+            --format="value(model)")
+          echo "MODEL_ID=${MODEL_ID}" >> $GITHUB_ENV
+
+      - name: Create Vertex AI endpoint
+        id: create_endpoint
+        run: |
+          ENDPOINT_ID=$(gcloud ai endpoints create \
+            --region=${{ secrets.GCP_VERTEX_AI_REGION }} \
+            --display-name="interview-ai-detector-endpoint" \
+            --format="value(name)")
+          echo "ENDPOINT_ID=${ENDPOINT_ID}" >> $GITHUB_ENV
+
+      - name: Deploy model to endpoint
+        run: |
+          gcloud ai endpoints deploy-model ${{ env.ENDPOINT_ID }} \
+            --region ${{ secrets.GCP_VERTEX_AI_REGION }} \
+            --model ${{ env.MODEL_ID }} \
+            --display-name interview-ai-detector-deployment \
+            --machine-type n1-standard-4 \
+            --accelerator count=1,type=nvidia-tesla-t4

Dockerfile

@@ -0,0 +1,17 @@
+# Use an official Python runtime as a base image
+FROM python:3.12-slim
+
+# Set the working directory in the container
+WORKDIR /app
+
+# Copy the current directory contents into the container at /app
+COPY . /app
+
+# Install any needed packages specified in requirements.txt
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Make port 8080 available to the world outside this container
+EXPOSE 8080
+
+# Run gunicorn with Uvicorn workers
+CMD ["gunicorn", "-k", "uvicorn.workers.UvicornWorker", "-w", "4", "-b", "0.0.0.0:8080", "prediction:app"]

hypothesis.py

@@ -23,9 +23,9 @@ class BaseModelHypothesis:
         self.features_not_normalized = []
 
         self.scaler_normalized_text_length = joblib.load(
-            "scaler-normalized-text-length.joblib")
+            "scalers/scaler-normalized-text-length.joblib")
         self.scaler_not_normalized = joblib.load(
-            "scaler-not-normalized.joblib")
+            "scalers/scaler-not-normalized.joblib")
 
     def process_emotion_lexicon(self):
         emotion_lexicon = {}

main_model.py

@@ -50,7 +50,7 @@ class PredictMainModel:
         self.model = AlbertCustomClassificationHead(
             self.albert_model).to(self.device)
         # TODO : CHANGE MODEL STATE DICT PATH
-        self.model.load_state_dict(torch.load("best_model_fold_4.pth"))
+        self.model.load_state_dict(torch.load("models/albert_model.pth"))
 
     def preprocess_input(self, text: str, additional_features: np.ndarray):
         encoding = self.tokenizer.encode_plus(

models/albert_model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b203b54caba10e290830cf720c8cd35093a358792e996ee04ee7d8e5e341651
+size 70752187

random_forest_model.py

@@ -5,8 +5,8 @@ from typing import List
 
 class RandomForestModel:
     def __init__(self):
-        self.scaler = joblib.load("rf_scaler.joblib")
-        self.model = joblib.load("random_forest.joblib")
+        self.scaler = joblib.load("scalers/rf_scaler.joblib")
+        self.model = joblib.load("models/random_forest.joblib")
 
     def preprocess_input(self, secondary_model_features: List[float]) -> np.ndarray:
         return self.scaler.transform(np.array(secondary_model_features).astype(np.float32).reshape(1, -1))

requirements.txt

@@ -5,4 +5,6 @@ pandas
 textstat
 scikit-learn==1.4.1.post1
 transformers
-fastapi
+fastapi
+uvicorn
+gunicorn