3

I'm writing a library using mlflow REST APIs. I'm looking for mlflow REST api for logging different mlflow models.

In the doc, https://www.mlflow.org/docs/latest/rest-api.html#log-model it says the api will be removed in future and doesn't have description about model_json request body.

If I see github, https://github.com/mlflow/mlflow/blob/master/docs/source/rest-api.rst mlflow REST API for Log model is missing.

desertnaut
  • 57,590
  • 26
  • 140
  • 166
veeresh patil
  • 1,168
  • 1
  • 11
  • 18
  • Any reason why you wouldn't use the well-supported and documented python api? – Rinat Veliakhmedov Mar 31 '22 at 09:35
  • Python api doesn't support authentication and authorisation. The mlflow tracking UI also doesn't have user based access. Hence i want to create a python library which uses my backend rest service to handle access mgmt which in-turn uses mlflow REST apis. – veeresh patil Mar 31 '22 at 15:29
  • It supports personal access tokens. – Rinat Veliakhmedov Mar 31 '22 at 15:41
  • Could you please elaborate more on personal access token usage or provide some links, that would be really helpful. Do you mean mlflow tracking server url supports access tokens? And you can access only user related( based on personal token) information like runs, experiments in mlflow server? or Do you mean personal access for backend artifact store? I need access management for metadata also like runs, metric/param logs, not just artifacts store – veeresh patil Mar 31 '22 at 16:37
  • Sorry, I got confused. We use mlflow with Databricks and I confused Databricks tokens with mlflow. My bad. – Rinat Veliakhmedov Apr 01 '22 at 20:48
  • No problem, Thanks for clarifying. – veeresh patil Apr 04 '22 at 06:59

1 Answers1

-1

You can develop mlflow multi tenancy features for user level access to view models as per tenant.

Rakesh Kumar
  • 1
  • 1