How to load a keras model (.h5 file) from local to Azure databricks workspace

Question

I have a keras model created on my local machine and I saved it as a model.h5 format. Now how do I load this model into my workspace on Azure databricks and import inside a databricks notebook and use the model?

trying the below URL but not successful, seems like its useful if and only if you save the model from databricks notebook using mlFlow and load it back under databricks using mlFlow:

https://www.mlflow.org/docs/latest/python_api/mlflow.keras.html

what if I have a keras model created in my local machine, how do I go ahead for importing?, please help.

Answer 1

If you just copied over the .h5 to your azure blog location and tried to read it using the MLflow Keras Load model, it may not work as you may need additional model meta-data info stored as part of the MLmodel file.

One way is to export and import all your meta-data and artifacts over to Azure using these tools. Another way is to just use the Keras native library to load the model.

Yet another way is to run the training code again in Azure Databricks notebook, giving you a fresh start.