Cache

Full name Microsoft.VisualBasic.MachineLearning.Transformer.DecoderLayer.Cache Assembly Microsoft.VisualBasic.DeepLearning Members 38

Forward intermediates of one decode step, required by the backward pass.

01 Syntax

Microsoft.VisualBasic.MachineLearning.Transformer.DecoderLayer.Cache

02 Fields

Name	Overloads	Summary
Input	2	The input of this layer for the step.
MaskedAttention	2	Output of the masked self attention sub layer.
MaskedAttentionDropped	2	Masked self attention output after dropout.
Normalized1	2	Result of the first AddNorm.
Norm1Mean	2	Mean used by the first layer normalization.
Norm1InvStd	2	Inverse standard deviation used by the first layer normalization.
CrossAttention	2	Output of the cross attention sub layer.
CrossAttentionDropped	2	Cross attention output after dropout.
Normalized2	2	Result of the second AddNorm.
Norm2Mean	2	Mean used by the second layer normalization.
Norm2InvStd	2	Inverse standard deviation used by the second layer normalization.
FeedForwardOutput	2	Output of the feed forward sub layer.
FeedForwardDropped	2	Feed forward output after dropout.
Norm3Mean	2	Mean used by the third layer normalization.
Norm3InvStd	2	Inverse standard deviation used by the third layer normalization.
DropoutApplied	2	Indicates whether dropout was applied during this step.
MaskedCache	2	Forward cache snapshot of the masked self attention sub layer.
CrossCache	2	Forward cache snapshot of the cross attention sub layer.
FfCache	2	Forward cache snapshot of the feed forward sub layer.

03 Members

field Input #

Input

The input of this layer for the step.

field MaskedAttention #

MaskedAttention

Output of the masked self attention sub layer.

field MaskedAttentionDropped #

MaskedAttentionDropped

Masked self attention output after dropout.

field Normalized1 #

Normalized1

Result of the first AddNorm.

field Norm1Mean #

Norm1Mean

Mean used by the first layer normalization.

field Norm1InvStd #

Norm1InvStd

Inverse standard deviation used by the first layer normalization.

field CrossAttention #

CrossAttention

Output of the cross attention sub layer.

field CrossAttentionDropped #

CrossAttentionDropped

Cross attention output after dropout.

field Normalized2 #

Normalized2

Result of the second AddNorm.

field Norm2Mean #

Norm2Mean

Mean used by the second layer normalization.

field Norm2InvStd #

Norm2InvStd

Inverse standard deviation used by the second layer normalization.

field FeedForwardOutput #

FeedForwardOutput

Output of the feed forward sub layer.

field FeedForwardDropped #

FeedForwardDropped

Feed forward output after dropout.

field Norm3Mean #

Norm3Mean

Mean used by the third layer normalization.

field Norm3InvStd #

Norm3InvStd

Inverse standard deviation used by the third layer normalization.

field DropoutApplied #

DropoutApplied

Indicates whether dropout was applied during this step.

field MaskedCache #

MaskedCache

Forward cache snapshot of the masked self attention sub layer.

field CrossCache #

CrossCache

Forward cache snapshot of the cross attention sub layer.

field FfCache #

FfCache

Forward cache snapshot of the feed forward sub layer.

field Input overload 2 #

Input

field MaskedAttention overload 2 #

MaskedAttention

field MaskedAttentionDropped overload 2 #

MaskedAttentionDropped

field Normalized1 overload 2 #

Normalized1

field Norm1Mean overload 2 #

Norm1Mean

field Norm1InvStd overload 2 #

Norm1InvStd

field CrossAttention overload 2 #

CrossAttention

field CrossAttentionDropped overload 2 #

CrossAttentionDropped

field Normalized2 overload 2 #

Normalized2

field Norm2Mean overload 2 #

Norm2Mean

field Norm2InvStd overload 2 #

Norm2InvStd

field FeedForwardOutput overload 2 #

FeedForwardOutput

field FeedForwardDropped overload 2 #

FeedForwardDropped

field Norm3Mean overload 2 #

Norm3Mean

field Norm3InvStd overload 2 #

Norm3InvStd

field DropoutApplied overload 2 #

DropoutApplied

field MaskedCache overload 2 #

MaskedCache

field CrossCache overload 2 #

CrossCache

field FfCache overload 2 #

FfCache