Thanks for tuning in to Google I/O. View all sessions on demandWatch on demand

TensorListConcatV2

public final class TensorListConcatV2

Concats all tensors in the list along the 0th dimension.

Requires that all tensors have the same shape except the first dimension.

input_handle: The input list. element_shape: The shape of the uninitialized elements in the list. If the first dimension is not -1, it is assumed that all list elements have the same leading dim. leading_dims: The list of leading dims of uninitialized list elements. Used if the leading dim of input_handle.element_shape or the element_shape input arg is not already set. tensor: The concated result. lengths: Output tensor containing sizes of the 0th dimension of tensors in the list, used for computing the gradient.

Public Methods

static <U, T extends Number> TensorListConcatV2<U>
create(Scope scope, Operand<?> inputHandle, Operand<T> elementShape, Operand<Long> leadingDims, Class<U> elementDtype)
Factory method to create a class wrapping a new TensorListConcatV2 operation.
Output<Long>
Output<U>
tensor()

Inherited Methods

Public Methods

public static TensorListConcatV2<U> create (Scope scope, Operand<?> inputHandle, Operand<T> elementShape, Operand<Long> leadingDims, Class<U> elementDtype)

Factory method to create a class wrapping a new TensorListConcatV2 operation.

Parameters
scope current scope
Returns
  • a new instance of TensorListConcatV2

public Output<Long> lengths ()

public Output<U> tensor ()