SplitBatchLengthLayer.md

CSplitBatchLengthLayer Class

• CSplitBatchLengthLayer Class
  • Settings
    • Outputs size
  • Trainable parameters
  • Inputs
  • Outputs

This class implements a layer that splits a single input blob into several output blobs along the BatchLength coordinate.

Settings

Outputs size

void SetOutputCounts(const CArray<int>& outputCounts);

Sets the values of the BatchLength dimension for the output blobs. See below for more details on the size and number of the outputs.

void SetOutputCounts2(int count0);
void SetOutputCounts3(int count0, int count1);
void SetOutputCounts4(int count0, int count1, int count2);

These auxiliary methods set the values of the BatchLength dimension for the outputs blobs, in the cases where you want 2, 3, or 4 outputs. Equivalent to calling the SetOutputCounts(const CArray<int>&) with the array of 1, 2, or 3 elements. See below for more details on the size and number of the outputs.

Trainable parameters

This layer has no trainable parameters.

Inputs

There is a single input which accepts a blob of any size. The only limitation is that its BatchLength dimension must be greater or equal to the sum of the elements of the GetOutputCounts() array.

Outputs

The layer has at least GetOutputCounts().Size() outputs. Each of the outputs contains a blob of the dimensions:

• BatchLength is equal to the corresponding element of the GetOutputCount() array; for example, the first output BatchLength is equal to GetOutputCount()[0] and so on.
• BatchWidth, ListSize, Height, Width, Depth, Channels are equal to the corresponding dimensions of the input.

However, if the BatchLength dimension of the input is greater than the sum of the GetOutputCounts() elements, the layer will have one more output of the dimensions:

• BatchLength is equal to the difference between the input BatchLength and the sum of the GetOutputCount() elements.
• BatchWidth, ListSize, Height, Width, Depth, Channels are equal to the corresponding dimensions of the input.