Experiments
Create Run
Configure dataset and start a training run within an experiment.
Create Run
A run is a single training execution within an experiment.
1. Add Dataset
Import your training data. The dataset undergoes validation before use:
| Status | Action |
|---|---|
| Pass | Proceeds to run configuration |
| Fail | Shows error reason, must fix data |
| Warning | Shows warning, can proceed |
Only validated datasets can be used for training.
2. Run Configuration
| Field | Description |
|---|---|
| Run Name | Identifier for this training run |
| Base Model | Select the model to fine-tune |
| Advanced Parameters | Training hyperparameters with optimized defaults |
Advanced parameters have best-practice defaults. Adjust only if you have specific requirements.
Run Contents
Each completed run contains:
- Configuration - Selected parameters and base model
- Metadata - Timestamps, duration, status
- Metrics - Training and validation metrics (loss, accuracy, F1, etc.)
- Model Weights - Trained model artifacts
Review metrics to evaluate model performance before freezing.