Experiments

Overview

Create and manage experiments for training machine learning models.

Experiments

An experiment is a container for training a specific machine learning task. It groups together multiple runs and their resulting models.

A run is a single training execution within an experiment, containing the dataset, training configuration, metrics, and model weights.

Experiments Overview

Workflow

  1. Create Experiment - Select modality and task type
  2. Create Run - Add dataset and configure training
  3. Freeze Model - Save trained models for production use

Run Summary

Aggregated run statistics across all Flow AIs

Create Experiment

Set up a new experiment with modality, task, and dataset.

On this page

On this page

ExperimentsWorkflow