Model Compare
Compare two models registered in Model Registry side-by-side to review training configurations, metrics, and loss curves before assigning Aliases.
Prerequisites
- Account with access to AI Training
- At least 2 models in the same category registered in Model Registry
- At least one project exists
Step 1. Navigate to Model Compare
- Click Models → Model compare in the left sidebar.
- The Compare Models page opens with the following guidance at the top.
Select a model category and two models to compare their training configuration, metrics, and loss curves.
Step 2. Select Models to Compare
In Section 1, select a category and two models to compare. You can compare exactly 2 models at a time.
- Select the category containing the models from the Model category dropdown. If no category is registered, it shows
No Data. - Select a model in Model (Option 1) and Model (Option 2). These are enabled only after a category is selected.
- Click Run Comparison.
Step 3. Review Comparison Result
After running the comparison, the results appear below Section 1.
- Model details: A table comparing the two models' training configuration side by side (Experiment, Base model, Status, Batch size, Learning rate, Max length, Method, Training mode, etc.).
- Metrics comparison: Charts comparing the two models' evaluation metrics and loss curves. Loss shows test/validation/training loss and Training Runtime shows training time as bar charts.
Before you select a category and models, a Select a model category placeholder is shown.
Step 4. Assign Alias Based on Comparison
Once you select the better model, assign an Alias in Model Registry. See Assign Alias in Browse Model Registry.
| Alias | Description |
|---|---|
| Best | Best-performing version |
| Active | Currently used version |
| Custom | User-defined alias |