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)