predict and flask docker

api/app.py

@@ -1,14 +1,27 @@
 from flask import Flask, request, jsonify
+<<<<<<< Updated upstream
+=======
+from joblib import load
+import os
+>>>>>>> Stashed changes
 
 app = Flask(__name__)
 
 @app.route('/hello/<name>')
 def index(name):
     return "Hello, "+name+"!"
 
-@app.route('/model', methods=['POST'])
+@app.route('/predict', methods=['POST'])
 def pred_model():
     js = request.get_json()
-    x= js['x']
-    y = js['y']
-    return x+y
+    image1 = js['image']
+    #Assuming this is the path of our best trained model
+    dirname = os.path.dirname(__file__)
+    filename = os.path.join(dirname, '../models/svmgamma:0.001_C:1.joblib')
+    model = load(filename)
+    pred1 = model.predict(image1)
+    #reurn pred1 in json
+    return jsonify(prediction=pred1.tolist())
+
+if __name__ == '__main__':
+    app.run(debug=True)

docker/Dockerfile

@@ -2,4 +2,6 @@ FROM python:3.9.17
 COPY . /digits/
 RUN pip3 install --no-cache-dir -r /digits/requirements.txt
 WORKDIR /digits
-CMD ["python3","digits.py","--runs=1","--test_size=0.2","--dev_size=0.3","--models=svm,tree"]
+#CMD ["python3","digits.py","--runs=1","--test_size=0.2","--dev_size=0.3","--models=svm,tree"]
+ENV FLASK_APP=api/app
+CMD [ "python3", "-m" , "flask", "run", "--host=0.0.0.0"]

docker/docker_run.sh

@@ -3,5 +3,5 @@ docker build -t digits:v1 -f ./docker/Dockerfile .
 # Create out volume
 docker volume create mltrain
 # Mount our volume to models directory (where train data is stored)
-docker run -v mltrain:/digits/models digits:v1
+docker run -d  -p 5001:5000 -v mltrain:/digits/models digits:v1 
 

requirements.txt

@@ -2,3 +2,4 @@ matplotlib==3.7.2
 scikit-learn==1.3.0
 pytest==7.4.2
 pandas==2.1.1
+flask==3.0.0