{% from 'partials/hero_cta.html' import hero_cta %}
{% call hero_cta( theme='amber', icon_class='fa-save', icon_shape='rounded', title='Training Checkpoints', description='This page displays checkpoints saved during training runs. Checkpoints are snapshots of your model at specific training steps, allowing you to resume training or use intermediate versions.', features=[ {'icon': 'fa-history', 'color': 'text-warning', 'label': 'Version History'}, {'icon': 'fa-cloud-upload-alt', 'color': 'text-info', 'label': 'HuggingFace Upload'}, {'icon': 'fa-broom', 'color': 'text-success', 'label': 'Retention Cleanup'}, ], show_condition='showHeroCTA()', dismiss_method='dismissHeroCTA()', cta_primary={'label': 'Dismiss', 'icon': 'fa-check', 'action': 'dismissHeroCTA()'}, tip_text='Configure checkpoint frequency and retention limits in the Checkpointing section on the Training tab.' ) %}
When Checkpoints Appear

Checkpoints are saved at intervals you configure (e.g., every 500 steps). They appear here once training has run long enough to trigger a save.

Configuring Saves

Set Checkpointing Steps and Checkpoints to Keep on the Training tab to control how often saves occur and how many are retained.

Publishing

Upload checkpoints directly to HuggingFace Hub. Configure your credentials on the Publishing tab to enable one-click uploads.

{% endcall %}