---
title: dropout
framework: ML Compute
role: symbol
role_heading: Instance Property
platforms: [iOS 14.0+, iPadOS 14.0+, Mac Catalyst 14.0+, macOS 11.0+, tvOS 14.0+]
path: mlcompute/mlcmultiheadattentiondescriptor/dropout
---

# dropout

The dropout rate you apply to the output projection weights.

## Declaration

```swift
var dropout: Float { get }
```

## Discussion

Discussion If dropout is non-zero, the layer adds a dropout operation to all the output projection weights except the last, with the dropout rate equal to dropout.

## See Also

### Inspecting Multi-Head Attention Descriptors

- [modelDimension](mlcompute/mlcmultiheadattentiondescriptor/modeldimension.md)
- [keyDimension](mlcompute/mlcmultiheadattentiondescriptor/keydimension.md)
- [valueDimension](mlcompute/mlcmultiheadattentiondescriptor/valuedimension.md)
- [headCount](mlcompute/mlcmultiheadattentiondescriptor/headcount.md)
- [hasBiases](mlcompute/mlcmultiheadattentiondescriptor/hasbiases.md)
- [hasAttentionBiases](mlcompute/mlcmultiheadattentiondescriptor/hasattentionbiases.md)
- [addsZeroAttention](mlcompute/mlcmultiheadattentiondescriptor/addszeroattention.md)