[Lab5] 周昱名-2200012718 Docker容器化

lab5-doc/周昱名-2200012718/docker-compose.yml

@@ -0,0 +1,29 @@
+version: '3.8'
+
+services:
+  training:
+    build:
+      context: .
+      dockerfile: docker/training/Dockerfile # TODO: Build the container using the training Dockerfile
+    volumes:
+      - model_storage:/app/models # TODO: Use the shared volume to store the trained model
+    # Set it to restart when it fails
+    restart: on-failure
+
+  inference:
+    build:
+      context: .
+      dockerfile: docker/inference/Dockerfile # TODO: Build the container using the inference Dockerfile
+    volumes:
+      - model_storage:/app/models # TODO: Use the shared volume to load the trained model
+    ports:
+      - "8080:8080" # TODO: Expose port 8080 for the Flask app
+    # Ensure that the inference service starts after the training service is complete
+    depends_on:
+      - training
+    restart: on-failure
+
+#Define a shared volume for the model file
+volumes:
+  model_storage:
+    driver: local

lab5-doc/周昱名-2200012718/docker/inference/Dockerfile

@@ -0,0 +1,16 @@
+# Base image
+FROM python:3.9-slim
+
+WORKDIR /app
+
+COPY docker/training/requirements.txt .
+RUN pip install -r requirements.txt
+
+COPY docker/inference/server.py ./server.py
+
+VOLUME ["/app/models"]
+
+EXPOSE 8080
+
+CMD ["python", "server.py"]
+

lab5-doc/周昱名-2200012718/docker/inference/server.py

@@ -0,0 +1,30 @@
+from flask import Flask, request, jsonify
+import numpy as np
+import joblib
+
+app = Flask(__name__)
+
+# TODO: Load the trained model from the shared volume (use the correct path)
+model = joblib.load('/app/models/iris_model.pkl')
+
+# TODO: Add request method to predict
+@app.route('/predict', methods=['POST'])
+def predict():
+    # TODO: Get the input array from the request body and make prediction using the model
+    get_json = request.get_json()
+    try:
+        iris_input = get_json['input']
+        # HINT: use np.array().reshape(1, -1) to convert input to 2D array
+        prediction = model.predict(np.array(iris_input, dtype=float).reshape(1, -1))
+
+        return jsonify({'prediction': prediction[0]})
+    except:
+        return jsonify({'error': 'something is wrong'}), 400
+
+@app.route('/')
+def hello():
+    return 'Welcome to Docker Lab'
+
+if __name__ == '__main__':
+    #Run the Flask app (bind it to port 8080 or any other port)
+    app.run(debug=True, port=8080, host='0.0.0.0')

lab5-doc/周昱名-2200012718/docker/training/Dockerfile

@@ -0,0 +1,15 @@
+# Base image
+FROM python:3.9-slim
+
+# Setup working directory
+WORKDIR /app
+# Copy requirements.txt and install dependencies
+COPY docker/training/requirements.txt .
+RUN pip install -r requirements.txt
+# Copy the training script to the working directory
+COPY docker/training/train.py ./train.py
+# Store model in a shared volume
+VOLUME ["/app/models"]
+# TODO: Set the command to run the training script
+CMD ["python", "train.py"]
+

lab5-doc/周昱名-2200012718/docker/training/requirements.txt

@@ -0,0 +1,4 @@
+Flask>=2.2.2
+scikit-learn==1.3.1
+joblib==1.3.2
+numpy==1.23.5

lab5-doc/周昱名-2200012718/docker/training/train.py

@@ -0,0 +1,21 @@
+from sklearn.svm import SVC
+from sklearn import datasets
+import joblib
+import os
+
+#Load the Iris dataset
+iris = datasets.load_iris()
+
+#Create an SVM classifier
+clf = SVC()
+
+#Train the model using the iris dataset
+model = clf.fit(iris.data, iris.target_names[iris.target])
+
+# TODO: Save the trained model to the shared volume (make sure to use the correct path)
+os.makedirs("/app/models", exist_ok=True)
+joblib.dump(model, '/app/models/iris_model.pkl')
+
+print("Model training complete and saved as iris_model.pkl")
+
+