Skip to main content

Model Training

Model training creates a training task from an uploaded and converted data artifact, then configures the model and compute resources before submission.

Training Configuration​

Training configuration confirms the training input, model type, and compute resource:

  • Select a training dataset and converted artifact to make sure the input comes from organized data.
  • Choose the model type according to the task and configure training parameters.
  • When selecting a training specification, view available resources, estimated cost, and current point balance.
  • Submit the training task after confirming the configuration.

Current Training​

Current training tracks tasks that are running:

  • View task status, training progress, and training steps.
  • Use the loss curve to observe how the model changes during training.
  • Refresh or synchronize the training status to confirm whether the task is still running.

History​

History keeps submitted training tasks and training results:

  • View the status, configuration, and completion time of historical training tasks.
  • Download the trained model files to a local directory after training completes.
  • Pause, resume, open the folder, or cancel model download tasks.