> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/sdgm/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/sdgm/_mcp/server.

# early_stopping

#### `early_stopping: <list[list[EarlyStopper]]>` (Optional)

## Description

A list of potential early stopping strategies for AutoML to explore.

An early stopping strategy can be empty (in which case no early stopping is applied), or it can be configured through two parameters: `min_delta` and `patience`:

* `min_delta` (default: `0.0`): The minimum absolute change in the monitored tune metric to qualify as an improvement. Must be `>= 0.0`.
* `patience` (default: `3`): The number of epochs with no improvement after which training will be stopped. Must be `> 0`.

By default, training is stopped after `3` consecutive epochs with no improvements of the tune metric on the validation set.

### Supported Task Types

* All

### Example

```yaml
early_stopping:
- min_delta: 0.0
  patience: 3
- min_delta: 0.1
  patience: 0
```

### Default Values

```yaml
early_stopping:
- min_delta: 0.0
  patience: 3
```