Struct MlflowTrackingClient

pub struct MlflowTrackingClient { /* private fields */ }

Provides access to a MLflow tracking server via REST API.

Support Mlflow API version 2.0.

Implementations

impl MlflowTrackingClient

pub fn new(base_url: impl AsRef<str>) -> Self

pub fn basic_auth( self, user_name: impl AsRef<str>, password: impl AsRef<str>, ) -> Self

Sets user name and password for basic authentication of the tracking server.

pub fn set_experiment( self, name: impl AsRef<str>, ) -> Result<Self, GetExperimentIdError>

Sets an experiment with the given name to this struct.

pub fn get_experiment(&self, name: impl AsRef<str>) -> Option<Experiment>

Gets [Experiment] by name from the tracking server.

If the experiment with given name does not exist in the trackingserver, it will be created.

TODO: Better error handling

pub fn create_recorder<E, R>( &self, run_name: impl AsRef<str>, ) -> Result<MlflowTrackingRecorder<E, R>>

where E: Env, R: ReplayBufferBase,

Create MlflowTrackingRecorder corresponding to a run.

If name is empty (""), a run name is generated by the tracking server.

If a Run with name exists in the tracking server, the run is used to create the recorder. If two or more runs with name exists, this method panics.

You need to set an experiment using MlflowTrackingClient::set_experiment() before calling this method.

This method uses MLFLOW_DEFAULT_ARTIFACT_ROOT environment variable as the directory where artifacts, like model parameters, will be saved. It is recommended to set this environment variable mlruns directory to which the tracking server persists experiment and run data.

pub fn get_runs_by_name(&self, name: impl AsRef<str>) -> Result<Vec<Run>>

Get runs by names.

This method queries the tracking server and returns Runs.

Trait Implementations

impl Debug for MlflowTrackingClient

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.