Freigeben über

Attention(Int32, Int32, Int32, Int32, List<Int32>) Konstruktor

  • Referenz

Definition

Namespace:
Microsoft.ML.TorchSharp.AutoFormerV2
Assembly:
Microsoft.ML.TorchSharp.dll
Paket:
Microsoft.ML.TorchSharp v0.22.1
Paket:
Microsoft.ML.TorchSharp v0.22.0
Paket:
Microsoft.ML.TorchSharp v0.23.0-preview.1.25125.4
Quelle:
Attention.cs
Quelle:
Attention.cs
Quelle:
Attention.cs

Initialisiert eine neue Instanz der Attention-Klasse.

public Attention(int inChannels, int keyChannels, int numHeads = 8, int attnRatio = 4, System.Collections.Generic.List<int> windowResolution = default);
new Microsoft.ML.TorchSharp.AutoFormerV2.Attention : int * int * int * int * System.Collections.Generic.List<int> -> Microsoft.ML.TorchSharp.AutoFormerV2.Attention
Public Sub New (inChannels As Integer, keyChannels As Integer, Optional numHeads As Integer = 8, Optional attnRatio As Integer = 4, Optional windowResolution As List(Of Integer) = Nothing)

Parameter

inChannels
Int32

Die Eingabekanäle.

keyChannels
Int32

Die wichtigsten Kanäle.

numHeads
Int32

Die Anzahl der Blöcke.

attnRatio
Int32

Das Verhältnis der Aufmerksamkeit.

windowResolution
List<Int32>

Die Auflösung des Fensters.

Gilt für: