Skip to main content
Version: Next

Request model training

Updated 2024.09.25

Request for training​

Streams → Actions → Train Model

Requesting model training involves sending the Dataset and Train Pipeline execution information to AI Conductor through the Stream and receiving the AI Model in return. When requesting model training, the following information is transmitted from Edge Conductor to AI Conductor:

  • Training Dataset: The Dataset selected for model training, which the AI Model will learn from. Users can select one or more datasets.
  • AI Solution: The Solution selected when creating a stream includes information about the model training environment settings. The AI Conductor references the Solution information to execute the Train Pipeline.
  • Train Pipeline: The Train Pipeline of the AI Solution currently in use.
  • Parameters: User parameter information inputted into the Train Pipeline during Stream creation.

AI Conductor executes the Train Pipeline, returns the trained AI Model to Edge Conductor, and logs the training history in the Stream's Model History.

To request model training:​

  1. Navigate to the Edge Conductor and login.
  2. In the left navigation pane, choose Stream.
  3. Choose the stream for model training. The stream status should be ready to train.
  4. Press the train button (brain icon) in the stream status pane or click Actions in the upper right corner and select train model.
  5. Choose the dataset to train model. (or User can train model using the sample dataset provided by the solution by performing check - Start with Sample Dataset )
  6. In the Train Resource section, select the system resources to be used for training.
  7. Press Train button.

Stop training​

Users can cancel the training of a model that is currently being trained.

In the Stream List table of the Streams menu, users can check the training status in the Status column. For streams that are currently being trained, the stop button will be activated. Users can click the stop button to halt the ongoing training. The successfully halted training history can be checked in the Model Status on the Inf. Model History screen.


Status of Model Train​

Users can check the training status of the model through the Status in the Stream table and the Model Status on the Inf. Model History screen.

  • Training: Model is currently being trained in AI Conductor, including the process of uploading training data to S3.
  • Training Failure: Training has failed. Users can check the training logs through the Log View function in the Train Artifacts of the model item in the Inf. Model History.
  • Ready to Deploy: Training has been successful, and the solution and model are ready to be deployed to the Edge.
  • Train Stopped: Training has been stopped while it was in progress.