Skip to content

train_adjust

Adjust hyperparameters of a running training run.

Overview

train_adjust writes updated hyperparameter values to a running training run's configuration. The training worker picks up the changes at the start of the next epoch.

Input Schema

Parameter Type Required Default Description
run_id string Yes — ID of the running training run.
lr float No null New learning rate.
batch integer No null New batch size.
epochs integer No null New total epoch count.

Output Schema

Field Type Description
run_id string Run ID.
adjusted object Map of adjusted parameter names to their new values.
status string Current run status.

Examples

CLI

curl -X POST http://127.0.0.1:7823/tools/train_adjust \
  -H "Authorization: Bearer $(cat ~/.fovux/auth.token)" \
  -H "Content-Type: application/json" \
  -d '{"run_id": "abc123", "lr": 0.001, "epochs": 200}'

Python

from fovux.tools.train_adjust import train_adjust
result = train_adjust("abc123", lr=0.001, epochs=200)

Notes & Limits

  • Only running or paused runs can be adjusted.
  • Changes take effect at the next epoch boundary, not immediately.
  • Not all hyperparameters may be adjustable depending on the training backend.

Failure Modes

  • FovuxRunNotFoundError if the run ID is not found.
  • FovuxValidationError if the run is not in a running state.