add lambda application

Author: Javier Poveda-Panter

online-machine-learning-aws-lambda/app/lambda_inference/Dockerfile

@@ -0,0 +1,9 @@
+FROM public.ecr.aws/lambda/python:3.8
+
+COPY requirements.txt ./
+
+RUN python3.8 -m pip install -r requirements.txt -t .
+
+COPY app.py ./
+
+CMD ["app.lambda_handler"]

online-machine-learning-aws-lambda/app/lambda_inference/__init__.py

@@ -0,0 +1,50 @@
+import json
+import boto3
+import os
+from aws_lambda_powertools import Logger
+from aws_lambda_powertools.utilities.data_classes import APIGatewayProxyEvent
+import tempfile
+import joblib
+
+s3 = boto3.client('s3')
+logger = Logger(service="inference")
+
+BUCKET_NAME = os.environ['BUCKET_NAME']
+model_name = 'model.joblib'
+
+
+def _download_model_from_s3(model_key):
+    with tempfile.TemporaryFile() as fp:
+        s3.download_fileobj(BUCKET_NAME, model_key, fp)
+        fp.seek(0)
+        model = joblib.load(fp)
+    return model
+
+
+# Lambda handler code
+@logger.inject_lambda_context
+def lambda_handler(event, _):
+    event = APIGatewayProxyEvent(event)
+    logger.info(event.__dict__)
+    # parse input event
+    data = event.get('body')
+    data = json.loads(data)
+    data = data.get("data")
+    # download current model from s3
+    regr = _download_model_from_s3(model_name)
+    # make prediction
+    pred = regr.predict(data)
+    # log prediction
+    logger.info({
+        "data": data,
+        "prediction": pred,
+    })
+
+    return {
+        'statusCode': 200,
+        'body': json.dumps(
+            {
+                "prediction": json.dumps(pred.tolist()),
+            }
+        )
+    }

online-machine-learning-aws-lambda/app/lambda_inference/app.py

@@ -0,0 +1,2 @@
+aws-lambda-powertools==1.18.1
+scikit-learn==0.24.2

online-machine-learning-aws-lambda/app/lambda_inference/requirements.txt

@@ -0,0 +1,9 @@
+FROM public.ecr.aws/lambda/python:3.8
+
+COPY requirements.txt ./
+
+RUN python3.8 -m pip install -r requirements.txt -t .
+
+COPY app.py ./
+
+CMD ["app.lambda_handler"]

online-machine-learning-aws-lambda/app/lambda_training/Dockerfile

@@ -0,0 +1,95 @@
+import json
+import boto3
+import os
+from aws_lambda_powertools import Logger
+from aws_lambda_powertools.utilities.data_classes import APIGatewayProxyEvent
+from sklearn import linear_model
+from sklearn.metrics import mean_squared_error, r2_score
+
+import tempfile
+import joblib
+
+s3 = boto3.client('s3')
+logger = Logger(service="training")
+
+BUCKET_NAME = os.environ['BUCKET_NAME']
+model_name = 'model.joblib'
+
+
+def _upload_model_to_s3(model, model_key):
+    with tempfile.TemporaryFile() as fp:
+        joblib.dump(model, fp)
+        fp.seek(0)
+        s3.upload_fileobj(fp, BUCKET_NAME, model_key)
+
+
+def _train_regression_model(X_train, y_train):
+    # Create linear regression object
+    regr = linear_model.LinearRegression()
+    # Train the model using the training sets
+    regr.fit(X_train, y_train)
+    return regr
+
+
+def _test_model(model, X_test, y_test):
+    # Make predictions using the testing set
+    y_pred = model.predict(X_test)
+    # calculate quality coefficients
+    mse = mean_squared_error(y_test, y_pred)
+    r2 = r2_score(y_test, y_pred)
+    return mse, r2
+
+
+def _parse_input(event):
+    # extract input data from event
+    data = event.get('body')
+    d = json.loads(data)
+    print('data', d.get("data"))
+    logger.info(d)
+    X = d["data"]['X']
+    y = d["data"]['y']
+    # Split the data into training/testing sets
+    X_train = X[:-20]
+    X_test = X[-20:]
+    # Split the targets into training/testing sets
+    y_train = y[:-20]
+    y_test = y[-20:]
+    return X_train, X_test, y_train, y_test
+
+
+# Lambda handler code
+@logger.inject_lambda_context
+def lambda_handler(event, _):
+    event = APIGatewayProxyEvent(event)
+    print(f'bucketname: {BUCKET_NAME}')
+    logger.info(event.__dict__)
+    # parse input event and split dataset
+    X_train, X_test, y_train, y_test = _parse_input(event)
+    # train regression model
+    regr = _train_regression_model(X_train, y_train)
+    # test model
+    mse, r2 = _test_model(regr, X_test, y_test)
+    logger.info({
+        "message": "model training successful",
+        'mean_squared_error': mse,
+        'r_squared': r2
+    })
+
+    # save trained model to s3
+    _upload_model_to_s3(regr, model_name)
+    logger.info({
+        "message": "model saved to s3",
+        'bucket_name': BUCKET_NAME,
+        'model_name': model_name
+    })
+
+    return {
+        'statusCode': 200,
+        'body': json.dumps(
+            {
+                'training': 'success',
+                'mean_squared_error': mse,
+                'r_squared': r2
+            }
+        )
+    }

online-machine-learning-aws-lambda/app/lambda_training/__init__.py

@@ -0,0 +1,2 @@
+aws-lambda-powertools==1.18.1
+scikit-learn==0.24.2