Concatenate#

The Concatenate operation merges two or more tensors into one, along the specified axis. The user may also specify an in-place merge. The operation provides the capabilities of the cudnn backend’s concatenate operation.

C++ API#

std::shared_ptr<Tensor_attributes>
concatenate(std::vector<std::shared_ptr<Tensor_attributes>>, Concatenate_attributes);

Concatenate attributes is a lightweight structure with inputs, outputs, and setters:

std::vector<std::shared_ptr<Tensor_attributes>> inputs;

std::unordered_map<output_names, std::shared_ptr<Tensor_attributes>> outputs;

Concatenate_attributes&
set_axis(int64_t const value)

Concatenate_attributes&
set_in_place_index(int64_t const value)