opus/dnn/nnet.c

/* Copyright (c) 2018 Mozilla
                 2008-2011 Octasic Inc.
                 2012-2017 Jean-Marc Valin */
/*
   Redistribution and use in source and binary forms, with or without
   modification, are permitted provided that the following conditions
   are met:

   - Redistributions of source code must retain the above copyright
   notice, this list of conditions and the following disclaimer.

   - Redistributions in binary form must reproduce the above copyright
   notice, this list of conditions and the following disclaimer in the
   documentation and/or other materials provided with the distribution.

   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
   ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
   A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR
   CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
   EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
   PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
   PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
   LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
   NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
   SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#include <stdlib.h>
#include <math.h>
#include "opus_types.h"
#include "arch.h"
#include "tansig_table.h"
#include "nnet.h"
#include "nnet_data.h"
#include "dred_rdovae_constants.h"
#include "plc_data.h"
#include "os_support.h"

#ifdef NO_OPTIMIZATIONS
#if defined(_MSC_VER)
#pragma message ("Compiling without any vectorization. This code will be very slow")
#else
#warning Compiling without any vectorization. This code will be very slow
#endif
#endif


#define SOFTMAX_HACK

#define MAX_ACTIVATIONS (4096)

static OPUS_INLINE void vec_swish(float *y, const float *x, int N)
{
   int i;
   float tmp[MAX_ACTIVATIONS];
   celt_assert(N <= MAX_ACTIVATIONS);
   vec_sigmoid(tmp, x, N);
   for (i=0;i<N;i++)
      y[i] = x[i]*tmp[i];
}

static OPUS_INLINE float relu(float x)
{
   return x < 0 ? 0 : x;
}

void compute_linear(const LinearLayer *linear, float *out, const float *in)
{
   int i, M, N;
   const float *bias;
   celt_assert(in != out);
   bias = linear->bias;
   M = linear->nb_inputs;
   N = linear->nb_outputs;
   if (linear->float_weights != NULL) {
     if (linear->weights_idx != NULL) sparse_sgemv8x4(out, linear->float_weights, linear->weights_idx, N, in);
     else sgemv(out, linear->float_weights, N, M, N, in);
   } else if (linear->weights != NULL) {
     if (linear->weights_idx != NULL) sparse_cgemv8x4(out, linear->weights, linear->weights_idx, linear->scale, N, M, in);
     else cgemv8x4(out, linear->weights, linear->scale, N, M, in);
     /* Only use SU biases on for integer matrices on SU archs. */
#ifdef USE_SU_BIAS
     bias = linear->subias;
#endif
   }
   else OPUS_CLEAR(out, N);
   if (bias != NULL) {
      for (i=0;i<N;i++) out[i] += bias[i];
   }
   if (linear->diag) {
      /* Diag is only used for GRU recurrent weights. */
      celt_assert(3*M == N);
      for (i=0;i<M;i++) {
         out[i] += linear->diag[i]*in[i];
         out[i+M] += linear->diag[i+M]*in[i];
         out[i+2*M] += linear->diag[i+2*M]*in[i];
      }
   }
}

void compute_generic_dense(const LinearLayer *layer, float *output, const float *input, int activation)
{
   compute_linear(layer, output, input);
   compute_activation(output, output, layer->nb_outputs, activation);
}

#define MAX_RNN_NEURONS_ALL IMAX(IMAX(MAX_RNN_NEURONS, PLC_MAX_RNN_NEURONS), DRED_MAX_RNN_NEURONS)


void compute_generic_gru(const LinearLayer *input_weights, const LinearLayer *recurrent_weights, float *state, const float *in)
{
  int i;
  int N;
  float zrh[3*MAX_RNN_NEURONS_ALL];
  float recur[3*MAX_RNN_NEURONS_ALL];
  float *z;
  float *r;
  float *h;
  celt_assert(3*recurrent_weights->nb_inputs == recurrent_weights->nb_outputs);
  celt_assert(input_weights->nb_outputs == recurrent_weights->nb_outputs);
  N = recurrent_weights->nb_inputs;
  z = zrh;
  r = &zrh[N];
  h = &zrh[2*N];
  celt_assert(recurrent_weights->nb_outputs <= 3*MAX_RNN_NEURONS_ALL);
  celt_assert(in != state);
  compute_linear(input_weights, zrh, in);
  compute_linear(recurrent_weights, recur, state);
  for (i=0;i<2*N;i++)
     zrh[i] += recur[i];
  compute_activation(zrh, zrh, 2*N, ACTIVATION_SIGMOID);
  for (i=0;i<N;i++)
     h[i] += recur[2*N+i]*r[i];
  compute_activation(h, h, N, ACTIVATION_TANH);
  for (i=0;i<N;i++)
     h[i] = z[i]*state[i] + (1-z[i])*h[i];
  for (i=0;i<N;i++)
     state[i] = h[i];
}

void compute_gated_activation(const LinearLayer *layer, float *output, const float *input, int activation)
{
   int i;
   float act1[MAX_INPUTS];
   float act2[MAX_INPUTS];
   celt_assert(layer->nb_inputs == layer->nb_outputs);
   compute_activation(act1, input, layer->nb_outputs, activation);
   compute_linear(layer, act2, input);
   compute_activation(act2, act2, layer->nb_outputs, ACTIVATION_SIGMOID);
   for (i=0;i<layer->nb_outputs;i++) output[i] = act1[i]*act2[i];
}

void compute_activation(float *output, const float *input, int N, int activation)
{
   int i;
   if (activation == ACTIVATION_SIGMOID) {
      vec_sigmoid(output, input, N);
   } else if (activation == ACTIVATION_TANH) {
      vec_tanh(output, input, N);
   } else if (activation == ACTIVATION_SWISH) {
      vec_swish(output, input, N);
   } else if (activation == ACTIVATION_RELU) {
      for (i=0;i<N;i++)
         output[i] = relu(input[i]);
   } else if (activation == ACTIVATION_SOFTMAX) {
#ifdef SOFTMAX_HACK
      OPUS_COPY(output, input, N);
      /*for (i=0;i<N;i++)
         output[i] = input[i];*/
#else
      float sum = 0;
      softmax(output, input, N);
      for (i=0;i<N;i++) {
         sum += output[i];
      }
      sum = 1.f/(sum+1e-30);
      for (i=0;i<N;i++)
         output[i] = sum*output[i];
#endif
   } else {
      celt_assert(activation == ACTIVATION_LINEAR);
      if (input != output) {
         for (i=0;i<N;i++)
            output[i] = input[i];
      }
   }
}

void _lpcnet_compute_dense(const DenseLayer *layer, float *output, const float *input)
{
   LinearLayer matrix;
   celt_assert(input != output);
   matrix.bias = layer->bias;
   matrix.subias = NULL;
   matrix.float_weights = layer->input_weights;
   matrix.weights = NULL;
   matrix.weights_idx = NULL;
   matrix.diag = NULL;
   matrix.nb_inputs = layer->nb_inputs;
   matrix.nb_outputs = layer->nb_neurons;
   matrix.scale = NULL;
   compute_linear(&matrix, output, input);
   compute_activation(output, output, layer->nb_neurons, layer->activation);
}

int sample_mdense(const MDenseLayer *layer, const float *input, const float *sampling_logit_table, kiss99_ctx *rng)
{
   int b, j, N, M, C, stride;
   int val=0;
   float thresholds[8];
   M = layer->nb_inputs;
   N = layer->nb_neurons;
   C = layer->nb_channels;
   celt_assert(N*C <= MAX_MDENSE_TMP);
   stride = M*C;

   celt_assert(N <= DUAL_FC_OUT_SIZE);

   /* Computing all the random thresholds in advance. These thresholds are directly
      based on the logit to avoid computing the sigmoid.*/
   for (b=0;b<8;b+=4) {
       uint32_t r = kiss99_rand(rng);
       thresholds[b] = sampling_logit_table[r&0xFF];
       thresholds[b+1] = sampling_logit_table[(r>>8)&0xFF];
       thresholds[b+2] = sampling_logit_table[(r>>16)&0xFF];
       thresholds[b+3] = sampling_logit_table[(r>>24)&0xFF];
   }

   for (b=0;b<8;b++)
   {
      int bit;
      int i;
      float sum1, sum2;

      i = (1<<b) | val;

      sum1 = layer->bias[i];
      sum2 = layer->bias[i + N];
      for (j=0;j<M;j++) {
         sum1 += layer->input_weights[i*stride + j]*input[j];
         sum2 += layer->input_weights[i*stride + j + M]*input[j];
      }
      sum1 = layer->factor[i]*tanh_approx(sum1);
      sum2 = layer->factor[N + i]*tanh_approx(sum2);
      sum1 += sum2;
      /*sum1 = 1.f/(1 + exp(-sum1));*/
#if 1 /* Sample the decision based on the logit. */
      bit = thresholds[b] < sum1;
#else
      sum1 = sigmoid_approx(sum1);
      bit = .025+.95*((rand()+.5f)/(RAND_MAX+1.f)) < sum1;
#endif
      val = (val << 1) | bit;
   }
   return val;

}

#ifdef USE_SU_BIAS
#define bias_type subias
#else
#define bias_type bias
#endif
#define MAX_IDX_SIZE 8192

void compute_gruB(const GRULayer *gru, const float* gru_b_condition, float *state, const float *input)
{
  LinearLayer in_matrix, rec_matrix;
  int i, M, N;
  float bias[3*MAX_RNN_NEURONS_ALL];
  float scale[3*MAX_RNN_NEURONS_ALL];
  M = gru->nb_inputs;
  N = gru->nb_neurons;

  in_matrix.bias = bias;
  in_matrix.diag = NULL;
  in_matrix.nb_inputs = M;
  in_matrix.nb_outputs = 3*N;
  in_matrix.subias = bias;
#ifdef DISABLE_DOT_PROD
  for (i=0;i<3*N;i++) bias[i] = gru->bias[i] + gru_b_condition[i];
  in_matrix.scale = NULL;
  in_matrix.float_weights = gru->input_weights;
  in_matrix.weights = NULL;
#else
  for (i=0;i<3*N;i++) bias[i] = gru->bias_type[i] + gru_b_condition[i];
  for (i=0;i<3*N;i++) scale[i] = SCALE_1;
  in_matrix.scale = scale;
  in_matrix.weights = gru->input_weights;
  in_matrix.float_weights = NULL;
#endif
  in_matrix.weights_idx = gru->input_weights_idx;

  rec_matrix.bias = &gru->bias[3*N];
  rec_matrix.diag = NULL;
  rec_matrix.nb_inputs = N;
  rec_matrix.nb_outputs = 3*N;
  rec_matrix.scale = scale;
  rec_matrix.subias = &gru->subias[3*N];
#ifdef DISABLE_DOT_PROD
  rec_matrix.scale = NULL;
  rec_matrix.float_weights = gru->recurrent_weights;
  rec_matrix.weights = NULL;
#else
  rec_matrix.scale = scale;
  rec_matrix.weights = gru->recurrent_weights;
  rec_matrix.float_weights = NULL;
#endif
  rec_matrix.weights_idx = NULL;
  compute_generic_gru(&in_matrix, &rec_matrix, state, input);
}

/* The input of this GRU is after the input matrix multiply. */
void compute_sparse_gru(const SparseGRULayer *gru, float *state, const float *input)
{
  LinearLayer in_matrix, rec_matrix;
  int i, N;
  float scale[3*MAX_RNN_NEURONS_ALL];
  N = gru->nb_neurons;

  in_matrix.bias = input;
  in_matrix.diag = NULL;
  in_matrix.nb_inputs = N;
  in_matrix.nb_outputs = 3*N;
  in_matrix.subias = input;
  in_matrix.scale = NULL;
  in_matrix.float_weights = NULL;
  in_matrix.weights = NULL;
  in_matrix.weights_idx = NULL;

  rec_matrix.bias = &gru->bias[3*N];
  rec_matrix.diag = gru->diag_weights;
  rec_matrix.nb_inputs = N;
  rec_matrix.nb_outputs = 3*N;
  rec_matrix.subias = &gru->subias[3*N];
#ifdef DISABLE_DOT_PROD
  rec_matrix.scale = NULL;
  rec_matrix.float_weights = gru->recurrent_weights;
  rec_matrix.weights = NULL;
#else
  for (i=0;i<3*N;i++) scale[i] = SCALE_1;
  rec_matrix.scale = scale;
  rec_matrix.weights = gru->recurrent_weights;
  rec_matrix.float_weights = NULL;
#endif
  rec_matrix.weights_idx = gru->idx;
  compute_generic_gru(&in_matrix, &rec_matrix, state, input);
}

#define MAX_CONV_INPUTS_ALL IMAX(MAX_CONV_INPUTS, DRED_MAX_CONV_INPUTS)

void compute_generic_conv1d(const LinearLayer *layer, float *output, float *mem, const float *input, int input_size, int activation)
{
   float tmp[MAX_CONV_INPUTS_ALL];
   celt_assert(input != output);
   celt_assert(layer->nb_inputs <= MAX_CONV_INPUTS_ALL);
   OPUS_COPY(tmp, mem, layer->nb_inputs-input_size);
   OPUS_COPY(&tmp[layer->nb_inputs-input_size], input, input_size);
   compute_linear(layer, output, tmp);
   compute_activation(output, output, layer->nb_outputs, activation);
   OPUS_COPY(mem, &tmp[input_size], layer->nb_inputs-input_size);
}

void compute_conv1d(const Conv1DLayer *layer, float *output, float *mem, const float *input)
{
   LinearLayer matrix;
   int N, M;
   M = layer->nb_inputs*layer->kernel_size;
   N = layer->nb_neurons;
   matrix.bias = layer->bias;
   matrix.subias = NULL;
   matrix.float_weights = layer->input_weights;
   matrix.weights = NULL;
   matrix.weights_idx = NULL;
   matrix.diag = NULL;
   matrix.nb_inputs = M;
   matrix.nb_outputs = N;
   matrix.scale = NULL;
   compute_generic_conv1d(&matrix, output, mem, input, layer->nb_inputs, layer->activation);
}

/* Computes non-padded convolution for input [ ksize1 x in_channels x (len2+ksize2) ],
   kernel [ out_channels x in_channels x ksize1 x ksize2 ],
   storing the output as [ out_channels x len2 ].
   We assume that the output dimension along the ksize1 axis is 1,
   i.e. processing one frame at a time. */
void conv2d_float(float *out, const float *weights, int in_channels, int out_channels, int ktime, int kheight, const float *in, int len2)
{
   int i;
   int in_stride;
   in_stride = len2+kheight-1;
   OPUS_CLEAR(out, out_channels*len2);
   for (i=0;i<out_channels;i++) {
      int m;
      for (m=0;m<in_channels;m++) {
         int t;
         for (t=0;t<ktime;t++) {
            int h;
            for (h=0;h<kheight;h++) {
               int j;
               for (j=0;j<len2;j++) {
                  out[i*len2 + j] += weights[i*in_channels*ktime*kheight + m*ktime*kheight + t*kheight + h] *
                                     in[t*in_channels*in_stride + m*in_stride + j + h];
               }
            }
         }
      }
   }
}

#define MAX_CONV2D_INPUTS 2048

void compute_conv2d(const Conv2dLayer *conv, float *out, float *mem, const float *in, int len2, int activation)
{
   int i;
   const float *bias;
   float in_buf[MAX_CONV2D_INPUTS];
   int time_stride;
   celt_assert(in != out);
   time_stride = conv->in_channels*(len2+conv->kheight);
   celt_assert(conv->ktime*time_stride <= MAX_CONV2D_INPUTS);
   OPUS_COPY(in_buf, mem, (conv->ktime-1)*time_stride);
   OPUS_COPY(&in_buf[(conv->ktime-1)*time_stride], in, time_stride);
   OPUS_COPY(mem, &in_buf[time_stride], (conv->ktime-1)*time_stride);
   bias = conv->bias;
   conv2d_float(out, conv->float_weights, conv->in_channels, conv->out_channels, conv->ktime, conv->kheight, in_buf, len2);
   if (bias != NULL) {
      for (i=0;i<conv->out_channels*len2;i++) out[i] += bias[i];
   }
   compute_activation(out, out, conv->out_channels*len2, activation);
}


void compute_embedding(const EmbeddingLayer *layer, float *output, int input)
{
   int i;
   celt_assert(input >= 0);
   celt_assert(input < layer->nb_inputs);
   /*if (layer->dim == 64) printf("%d\n", input);*/
   for (i=0;i<layer->dim;i++)
   {
      output[i] = layer->embedding_weights[input*layer->dim + i];
   }
}

void compute_gru_a_input(float *output, const float *input, int N, const EmbeddingLayer *layer1, int val1, const EmbeddingLayer *layer2, int val2, const EmbeddingLayer *layer3, int val3) {
   int i;
   for (i=0;i<3*N;i++) {
      output[i] = input[i] + layer1->embedding_weights[val1*layer1->dim + i]
                           + layer2->embedding_weights[val2*layer2->dim + i]
                           + layer3->embedding_weights[val3*layer3->dim + i];
   }
}

void accum_embedding(const EmbeddingLayer *layer, float *output, int input)
{
   int i;
   celt_assert(input >= 0);
   celt_assert(input < layer->nb_inputs);
   /*if (layer->dim == 64) printf("%d\n", input);*/
   for (i=0;i<layer->dim;i++)
   {
      output[i] += layer->embedding_weights[input*layer->dim + i];
   }
}