Versioned name: DynamicReshape-1
Category: Shape manipulation
Short description: DynamicReshape operation changes dimensions of the input tensor according to the specified order. Input tensor volume is equal to output tensor volume, where volume is the product of dimensions. Output tensor may have a different memory layout from input tensor. DynamicReshape is not guaranteed to return a view or a copy when input tensor and output tensor can be inplaced, user should not depend on this behavior. In DynamicReshape, shape is given as an input at runtime. It’s useful when the target shape is unknown during the operator creation. Use DynamicReshape if shape is not constant or is not available until runtime. Otherwise, use StaticReshape.
Description: special_zero controls how zero values in
shapeare interpreted. If special_zero is
0is interpreted as-is which means that output shape will contain a zero dimension at the specified location. Input and output tensors are empty in this case. If special_zero is
true, then all zeros in
shapeimplies the copying of corresponding dimensions from
data.shapeinto the output shape.
Range of values:
Default value: None
data– multidimensional input tensor of type T. Required.
shape– specifies the output shape. The values in this tensor could be -1, 0 and any positive integer number.
-1means that this dimension is calculated to keep the overall elements count the same as in the input tensor.
0is interpreted by attr special_zero. No more than one
-1can be used in
1: Output tensor with the same content as input
databut with shape defined by input