Logging Machine Learning model to a centralized workspace from a Databricks workspace

SKandiga — Thu, 06 Jul 2023 05:48:47 GMT

hi all,

I have been working on the concept on of centralized workspace where I have been trying to create a centralized workspace repo.

I am running my model on Workspace A, while I am logging my model on to a remote Databricks workspace B,

I have connected the two repo's by using set_tracking_uri(),

destination_workspace_url = "xxxxxxxxxxxxxxxxxxxxxx"

destination_access_token = "xxxxxxxxxxxxxxxxxxxxxxxx"

import mlflow

import os

from mlflow.tracking import MlflowClient

client = MlflowClient()

os.environ['MLFLOW_TRACKING_TOKEN'] = destination_access_token

mlflow.set_tracking_uri(destination_workspace_url)

tracking_uri = mlflow.set_tracking_uri(destination_workspace_url)

experiment = mlflow.set_experiment("/Users/xxxxxxxxxxxxx/mlFlow_centralized_rep")

print(experiment)

Now the connection is successful, also the hyper parameters and the output score is getting logged but when I am trying to log the model using the command ,

mlflow.pyfunc.log_model("random_forest_model_v10", python_model=wrappedModel, conda_env=conda_env, signature=signature)

this doesn't seem to work I am encountered with the following error,

REST EXCEPTION: Resource id "xxxxxxxxx" does not exist.

I have tried configuring the compute cluster with the databricks configure --token

But this does not seem to work, it's not logging the artifacts that's the main issue here. Any suggestions on this please ?

topic Logging Machine Learning model to a centralized workspace from a Databricks workspace in Machine Learning

Logging Machine Learning model to a centralized workspace from a Databricks workspace