# Slot: batch_mode 


_How the per-subject fits are executed. ``fan_out`` (default) runs one per-subject job, each binding its own targets at run time. ``on_device`` evaluates the whole cohort together as one vectorised batch on a single device — the subject axis mapped in parallel while the shared network is reused across subjects. Backend-independent intent: the scientific spec is identical either way; only the execution strategy differs (a backend realises the batch with its own vectorised-map primitive)._



<div data-search-exclude markdown="1">



URI: [tvbo_study:slot/batch_mode](https://w3id.org/tvbo/study/slot/batch_mode)
<!-- no inheritance hierarchy -->





## Applicable Classes

| Name | Description | Modifies Slot |
| --- | --- | --- |
| [Dataset](../classes/Dataset.qmd) | A collection of subjects for a multi-subject study |  no  |
| [DBSDataset](../classes/DBSDataset.qmd) | Collection of data related to a specific DBS study |  no  |






## Properties

### Type and Range

| Property | Value |
| --- | --- |
| Range | [SubjectBatchMode](../enums/SubjectBatchMode.qmd) |
| Domain Of | [Dataset](../classes/Dataset.qmd) |

### Cardinality and Requirements

| Property | Value |
| --- | --- |
### Slot Characteristics

| Property | Value |
| --- | --- |
| If Absent | `string(fan_out)` |
| Owner | [Dataset](../classes/Dataset.qmd) |












## Identifier and Mapping Information





### Schema Source


* from schema: https://w3id.org/tvbo/study




## Mappings

| Mapping Type | Mapped Value |
| ---  | ---  |
| self | tvbo_study:batch_mode |
| native | tvbo_study:batch_mode |




## LinkML Source

<details>
```yaml
name: batch_mode
description: 'How the per-subject fits are executed. ``fan_out`` (default) runs one
  per-subject job, each binding its own targets at run time. ``on_device`` evaluates
  the whole cohort together as one vectorised batch on a single device — the subject
  axis mapped in parallel while the shared network is reused across subjects. Backend-independent
  intent: the scientific spec is identical either way; only the execution strategy
  differs (a backend realises the batch with its own vectorised-map primitive).'
from_schema: https://w3id.org/tvbo/study
rank: 1000
ifabsent: string(fan_out)
owner: Dataset
domain_of:
- Dataset
range: SubjectBatchMode

```
</details></div>