Skip to content

R Tracking/Fluent API#548

Merged
javierluraschi merged 89 commits intomlflow:masterfrom
kevinykuo:feature/r-tracking
Sep 27, 2018
Merged

R Tracking/Fluent API#548
javierluraschi merged 89 commits intomlflow:masterfrom
kevinykuo:feature/r-tracking

Conversation

@kevinykuo
Copy link
Contributor

This change splits the tracking feature into a "fluent" component and a "client" component to be consistent with other languages and fix #432.

The explicit Tracking Service/client API, which is a thin layer on top of the internal REST wrappers, is exposed to the user in tracking-client.R. Similar to their Python counterparts, these methods require an mlflow_client object (created in R via mlflow_client()) to be passed and do not depend on global state. The higher-level fluent API, which implicitly manages active runs and experiments, lives in tracking-fluent.R. These are the functions we expect most users to be using.

Note that the "connection" constructs, e.g. mlflow_connect() and the "active connection" constructs have been deprecated since mlflow_client now provides the necessary mechanism to track services.

This PR is WIP: There are a couple methods yet to be implemented, specifically list_run_infos and download_artifacts. The documentation also needs to be updated.

@tomasatdatabricks
Copy link
Contributor

Awesome, thanks Kevin!

@kevinykuo
Copy link
Contributor Author

@tomasatdatabricks The fluent and tracking client functions have been separated as described above.

@javierluraschi javierluraschi changed the title [WIP] R Tracking/Fluent API R Tracking/Fluent API Sep 27, 2018
@javierluraschi javierluraschi self-requested a review September 27, 2018 04:53
@javierluraschi
Copy link
Contributor

LGTM

@javierluraschi javierluraschi merged commit 885bf0c into mlflow:master Sep 27, 2018
@javierluraschi javierluraschi added the language/r R APIs and clients label Sep 27, 2018
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

language/r R APIs and clients

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Consider mapping Service API 1:1 in R

3 participants