Skip to content

Conversation

@jasonrusselwang
Copy link

When using import-model to import from an exported open source project to Databricks, I was receiving the following error:

mlflow.exceptions.MlflowException: Model version creation failed for model name: help-mlops_reference_project-normal_tickets version: 2 with status: FAILED_REGISTRATION and message: Failed registration. The given source path `dbfs:/databricks/mlflow-tracking/2485056402562980/6ddc7ac453ea44abb2f740628fe3b423/artifacts/45/007930631053402f957be01de073fc3e/artifacts/model` does not exist.

This is due to the source field in model.json being an S3 bucket that contains "artifacts" (i.e. s3://test-analytics-us-east-1-mlflow-artifacts) which results in the pattern "artifacts" occurring twice, resulting in:
dbfs:/databricks/mlflow-tracking/2485056402562980/6ddc7ac453ea44abb2f740628fe3b423/artifacts/45/007930631053402f957be01de073fc3e/artifacts/model
instead of:
dbfs:/databricks/mlflow-tracking/2485056402562980/6ddc7ac453ea44abb2f740628fe3b423/artifacts/model.

I've added a check for source S3 URLs and parses the path so that the bucket name will no longer interfere with the creation of the model_path variable.

@jasonrusselwang
Copy link
Author

@mac-macoy
@safwanislam

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant