LSTM Layer Functions

Functions
riscv_nmsis_nn_status	riscv_lstm_unidirectional_s16_s8 (nmsis_nn_lstm_context *scratch_buffers, const int8_t *input_data, const nmsis_nn_lstm_dims *lstm_dims, const int8_t *in_to_in_weights, const int8_t *in_to_forget_weights, const int8_t *in_to_cell_weights, const int8_t *in_to_out_weights, const int8_t *recurrent_to_in_weights, const int8_t *recurrent_to_forget_weights, const int8_t *recurrent_to_cell_weights, const int8_t *recurrent_to_out_weights, const int16_t *cell_to_in_weights, const int16_t *cell_to_forget_weights, const int16_t *cell_to_out_weights, const int8_t *projection_weights, const nmsis_nn_lstm_params *lstm, int8_t *output_state, int16_t *cell_state, int8_t *output_data)
	LSTM unidirectional function with 8 bit input and output and 16 bit gate output Peephole connections, projection, clipping, combined input/forget gate and layer normalization are not supported. More...

Detailed Description

Function Documentation

riscv_lstm_unidirectional_s16_s8()

riscv_nmsis_nn_status riscv_lstm_unidirectional_s16_s8	(	nmsis_nn_lstm_context *	scratch_buffers,
		const int8_t *	input_data,
		const nmsis_nn_lstm_dims *	lstm_dims,
		const int8_t *	input_to_input_weights,
		const int8_t *	input_to_forget_weights,
		const int8_t *	input_to_cell_weights,
		const int8_t *	input_to_output_weights,
		const int8_t *	recurrent_to_input_weights,
		const int8_t *	recurrent_to_forget_weights,
		const int8_t *	recurrent_to_cell_weights,
		const int8_t *	recurrent_to_output_weights,
		const int16_t *	cell_to_input_weights,
		const int16_t *	cell_to_forget_weights,
		const int16_t *	cell_to_output_weights,
		const int8_t *	projection_weights,
		const nmsis_nn_lstm_params *	lstm,
		int8_t *	output_state,
		int16_t *	cell_state,
		int8_t *	output_data
	)

LSTM unidirectional function with 8 bit input and output and 16 bit gate output Peephole connections, projection, clipping, combined input/forget gate and layer normalization are not supported.

Parameters

[in]	scratch_buffers	Struct containing scratch buffers Expected size for each scratch buffer is lstm_dims->num_batches * lstm_dims->num_outputs.
[in]	input_data	Pointer to input data
[in]	lstm_dims	LSTM input parameters related to dimensions
[in]	input_to_input_weights	Input to input weights
[in]	input_to_forget_weights	Input to forget weights
[in]	input_to_cell_weights	Input to cell weights
[in]	input_to_output_weights	Input to output weights
[in]	recurrent_to_input_weights	Recurrent to input weights
[in]	recurrent_to_forget_weights	Recurrent to forget weights
[in]	recurrent_to_cell_weights	Recurrent to cell weights
[in]	recurrent_to_output_weights	Recurrent to output weights
[in]	cell_to_input_weights	Cell to input weights. Not used.
[in]	cell_to_forget_weights	Cell to forget weights. Not used.
[in]	cell_to_output_weights	Cell to output weights. Not used.
[in]	projection_weights	Projection weights. Not used.
[in]	lstm	LSTM parameters. See struct declaration
[in]	output_state	Pointer to (recurrent) output state
[in]	cell_state	Pointer to cell state
[in]	output_data	Pointer to output state

Note

Following assumptions are done based on LSTM functionality as supported by Keras version 2.9.0 at the time of development. As stated here, https://github.com/tensorflow/community/blob/master/rfcs/20180920-unify-rnn-interface.md Keras's LSTMCell is equivalent to TensorFlow's BasicLSTMCell, which does not support peephole, clipping or projection. Layer normalization and combined input/forget gate are not supported either.

1 Input to input weight can not be nullptr. Otherwise nullptr for combined input/forgat gate. 2 Cell weights are not used and should be nullptr. Otherwise needed for peephole connections. 3 Projection weight is not used and should be nullpr. Otherwise needed for projection.

Returns

The function returns RISCV_NMSIS_NN_SUCCESS

1. Supported framework: TensorFlow Lite micro