Imp.Tracking.MLflow (Imp v0.5.0)

MLflow tracking backend implemented over the official HTTP API.

Summary

Types

t()

Functions

log_artifact(state, path, body, content_type \\ "application/octet-stream")

Uploads a raw artifact through an mlflow-artifacts proxy URI.

log_config(state, config)

Logs configuration values as MLflow parameters in one log-batch request.

log_metrics(state, metrics, opts \\ [])

Logs numeric metrics in one log-batch request.

log_summary(state, summary)

Logs summary values as imp.summary.* run tags in one log-batch request.

log_table(state, path, data)

Appends rows to an MLflow split-orient JSON table and records its table tag.

Types

t()

@type t() :: %Imp.Tracking.MLflow{
  artifact_uri: String.t() | nil,
  experiment_id: String.t(),
  finish_adopted_run: boolean(),
  headers: Imp.Tracking.Transport.headers(),
  owned_run: boolean(),
  run_id: String.t(),
  tracking_uri: String.t(),
  transport: module(),
  transport_opts: keyword()
}

Functions

log_artifact(state, path, body, content_type \\ "application/octet-stream")

@spec log_artifact(t(), String.t(), iodata(), String.t()) :: :ok | {:error, term()}

Uploads a raw artifact through an mlflow-artifacts proxy URI.

log_config(state, config)

@spec log_config(t(), map() | keyword()) :: :ok | {:error, term()}

Logs configuration values as MLflow parameters in one log-batch request.

log_metrics(state, metrics, opts \\ [])

@spec log_metrics(t(), map() | keyword(), keyword()) :: :ok | {:error, term()}

Logs numeric metrics in one log-batch request.

log_summary(state, summary)

@spec log_summary(t(), map() | keyword()) :: :ok | {:error, term()}

Logs summary values as imp.summary.* run tags in one log-batch request.

log_table(state, path, data)

@spec log_table(t(), String.t(), [map()] | map()) :: :ok | {:error, term()}

Appends rows to an MLflow split-orient JSON table and records its table tag.