Merge pull request #7 from YakobusIP/production

core-model-prediction/cloudbuild.yaml → cloudbuild-endpoint.yaml

@@ -1,49 +1,53 @@
 steps:
-  - name: "docker"
-    dir: "core-model-prediction"
-    args:
-      [
-        "builds",
-        "submit",
-        "--tag",
-        "gcr.io/$PROJECT_ID/interview-ai-detector/model-prediction",
-      ]
-
   - name: "gcr.io/google.com/cloudsdktool/cloud-sdk"
     entrypoint: "bash"
+    id: upload-model
+    timeout: 600s
     args:
       - "-c"
       - |
-        MODEL_ID=$(gcloud ai models upload \
-          --region="$_GCP_VERTEX_AI_REGION" \
+        gcloud ai models upload \
+          --region="us-central1" \
           --container-ports=8080 \
-          --container-image-uri="gcr.io/$PROJECT_ID/interview-ai-detector/model-prediction:latest" \
+          --container-image-uri="us-central1-docker.pkg.dev/${PROJECT_ID}/interview-ai-detector/model-prediction:latest" \
           --container-predict-route="/predict" \
           --container-health-route="/health" \
-          --display-name="interview-ai-detector-model" \
-          --format="value(model)")
-        echo "MODEL_ID=${MODEL_ID}"
+          --display-name="interview-ai-detector-model"
+
+  - name: "gcr.io/google.com/cloudsdktool/cloud-sdk"
+    entrypoint: "bash"
+    id: fetch-model
+    waitFor: upload-model
+    timeout: 600s
+    args:
+      - "-c"
+      - |
+        MODEL_ID=$(gcloud ai models list --region=us-central1 --format="value(name)" | head -n 1)
+        echo MODEL_ID > /workspace/model_id.txt
 
   - name: "gcr.io/google.com/cloudsdktool/cloud-sdk"
     entrypoint: "bash"
+    id: create-endpoint
+    waitFor: fetch-model
     args:
       - "-c"
       - |
         ENDPOINT_ID=$(gcloud ai endpoints create \
-          --region="$_GCP_VERTEX_AI_REGION" \
+          --region="us-central1" \
           --display-name="interview-ai-detector-endpoint" \
           --format="value(name)")
-        echo "ENDPOINT_ID=${ENDPOINT_ID}"
+        echo ENDPOINT_ID > /workspace/endpoint_id.txt
 
   - name: "gcr.io/google.com/cloudsdktool/cloud-sdk"
     entrypoint: "bash"
+    waitFor: create-endpoint
     args:
       - "-c"
       - |
-        gcloud ai endpoints deploy-model "$ENDPOINT_ID" \
-          --region="$_GCP_VERTEX_AI_REGION" \
-          --model="$MODEL_ID" \
+        gcloud ai endpoints deploy-model "${$(cat /workspace/endpoint_id.txt)}" \
+          --region="us-central1" \
+          --model="${$(cat /workspace/model_id.txt)}" \
           --display-name="interview-ai-detector-deployment" \
           --machine-type="n1-standard-4" \
           --accelerator="count=1,type=nvidia-tesla-t4" \
-          --service-account="$_GCP_VERTEX_AI_SA_EMAIL"
+          --service-account="vertex-ai-user-managed-sa@steady-climate-416810.iam.gserviceaccount.com"

cloudbuild-model.yaml

@@ -0,0 +1,20 @@
+steps:
+  - name: "gcr.io/cloud-builders/docker"
+    dir: "core-model-prediction"
+    args:
+      [
+        "build",
+        "-t",
+        "us-central1-docker.pkg.dev/${PROJECT_ID}/interview-ai-detector/model-prediction:latest",
+        ".",
+      ]
+
+  - name: "gcr.io/cloud-builders/docker"
+    args:
+      [
+        "push",
+        "us-central1-docker.pkg.dev/${PROJECT_ID}/interview-ai-detector/model-prediction:latest",
+      ]
+
+images:
+  - "us-central1-docker.pkg.dev/${PROJECT_ID}/interview-ai-detector/model-prediction:latest"

core-model-prediction/prediction.py

@@ -64,7 +64,7 @@ def process_instance(data: PredictRequest):
     return {
         "prediction_class": "AI" if secondary_model_prediction == 1 else "HUMAN",
         "details": {
-            "main_model_probability": main_model_probability,
+            "main_model_probability": str(main_model_probability),
             "final_prediction": secondary_model_prediction
         }
     }

core-model-prediction/random_forest_dependencies.py

@@ -5,7 +5,6 @@ from collections import Counter
 class RandomForestDependencies:
     def __init__(self):
         self.gemma2bdependencies = Gemma2BDependencies()
-        self.random_forest_features = []
 
     def calculate_features(self, question: str, answer: str, probability: float, backspace_count: int, typing_duration: int, letter_click_counts: dict[str, int]):
         cosine_similarity = self.gemma2bdependencies.calculate_cosine_similarity(
@@ -15,7 +14,7 @@ class RandomForestDependencies:
         letter_discrepancy = self.calculate_letter_discrepancy(
             answer, letter_click_counts)
 
-        self.random_forest_features = [
+        return [
             cosine_similarity, probability, backspace_count_normalized,
             typing_duration_normalized, letter_discrepancy
         ]

core-model-prediction/random_forest_model.py

@@ -12,4 +12,4 @@ class RandomForestModel:
         return self.scaler.transform(np.array(secondary_model_features).astype(np.float32).reshape(1, -1))
 
     def predict(self, secondary_model_features: List[float]):
-        return self.model.predict(self.preprocess_input(secondary_model_features))[0]
+        return int(self.model.predict(self.preprocess_input(secondary_model_features))[0])