Build and Deploy Tensorflow Model and its integration with a Flask application from scratch
In this article, we will cover what needs to be done to deploy the TensorFlow model with the TensorFlow model server and related commands. I have extended one of my previous projects with new topics covering early stopping and model evaluation techniques. Also, will be building a flask application with a form in it, to provide new data to make a prediction.
Points that are covered:
- Build regression model
- Train model with Early Stopping callback
- Evaluate model with the test dataset
- Save model
- Save the stats of the dataset for normalization
- Build a Flask application from scratch
- Start TensorFlow production-ready server to serve model prediction
- Make prediction from a flask
Complete video for the above step is as below:
Useful commands are as below:
Deploying TensorFlow model:
# Saving tensorflow modelsave_path = '/tmp/regression_model/1'
tf.keras.models.save_model(model, save_path)
# model_name: Any name for the your model
# model_base_path: location of the saved modeltensorflow_model_server \
--rest_api_port="8501" \
--model_name="regression_model" \
--model_base_path="/tmp/regression_model/"API request to TensorFlow model server:curl -X POST http://localhost:8501/v1/models/regression_model:predict -d '{"instances": [ [ 0.3, -1.4, 0.8, -0.1, -0.9, -0.3 ] ] }'
Github link for the project is here. Video is pretty lengthy but it is done everything from scratch to understand integration better. Hope it is useful. Enjoy coding!!!
