24 #ifndef __ARM_COMPUTE_NEFULLYCONNECTEDLAYER_H__
25 #define __ARM_COMPUTE_NEFULLYCONNECTEDLAYER_H__
69 bool _transpose_weights;
70 bool _is_batched_fc_layer;
109 Tensor _interleave4x4_output;
110 Tensor _reshape_weights_output;
111 bool _are_weights_reshaped;
112 bool _is_batched_fc_layer;
113 bool _linearize_input;
114 bool _accumulate_biases;
Base class for all functions.
Basic function to reshape the weights of Fully Connected layer with NEON.
void configure(const ITensor *input, ITensor *output, bool transpose_weights, bool is_batched_fc_layer)
Set the input and output tensors.
Interface for the im2col reshape kernel.
NEFullyConnectedLayer(std::shared_ptr< IMemoryManager > memory_manager=nullptr)
Constructor.
void configure(const ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output, bool transpose_weights=true, bool are_weights_reshaped=false)
Set the input and output tensors.
Interface for NEON tensor.
NEON kernel to interleave the elements of a matrix.
NEON kernel which transposes the elements of a matrix.
NEON kernel to add a bias to each row of the input tensor.
NEON kernel which transposes the elements of a matrix in chunks of 1xW, where W is equal to (16 / ele...
Basic implementation of the tensor interface.
void run() override
Run the kernels contained in the function.
Basic function to compute a Fully Connected layer on NEON.
void run() override
Run the kernels contained in the function.
NEFullyConnectedLayerReshapeWeights(std::shared_ptr< IMemoryManager > memory_manager=nullptr)
Constructor.
NEON kernel to multiply two input matrices "A" and "B".