From dbdd31ee211fe8b1ac7e93ceadf7b34b8d304f34 Mon Sep 17 00:00:00 2001
From: Roland Singer <roland.singer@desertbit.com>
Date: Wed, 22 Aug 2018 11:56:41 +0000
Subject: [PATCH] updated README to include information about learning rate adjustment for multiple GPUs
---
src/connected_layer.c | 50 ++++++++++++++++++++++++++++++++++++++++++--------
1 files changed, 42 insertions(+), 8 deletions(-)
diff --git a/src/connected_layer.c b/src/connected_layer.c
index 623e6c8..e6dc759 100644
--- a/src/connected_layer.c
+++ b/src/connected_layer.c
@@ -36,6 +36,10 @@
l.weights = calloc(outputs*inputs, sizeof(float));
l.biases = calloc(outputs, sizeof(float));
+ l.forward = forward_connected_layer;
+ l.backward = backward_connected_layer;
+ l.update = update_connected_layer;
+
//float scale = 1./sqrt(inputs);
float scale = sqrt(2./inputs);
for(i = 0; i < outputs*inputs; ++i){
@@ -66,6 +70,10 @@
}
#ifdef GPU
+ l.forward_gpu = forward_connected_layer_gpu;
+ l.backward_gpu = backward_connected_layer_gpu;
+ l.update_gpu = update_connected_layer_gpu;
+
l.weights_gpu = cuda_make_array(l.weights, outputs*inputs);
l.biases_gpu = cuda_make_array(l.biases, outputs);
@@ -89,10 +97,16 @@
l.x_gpu = cuda_make_array(l.output, l.batch*outputs);
l.x_norm_gpu = cuda_make_array(l.output, l.batch*outputs);
+#ifdef CUDNN
+ cudnnCreateTensorDescriptor(&l.normTensorDesc);
+ cudnnCreateTensorDescriptor(&l.dstTensorDesc);
+ cudnnSetTensor4dDescriptor(l.dstTensorDesc, CUDNN_TENSOR_NCHW, CUDNN_DATA_FLOAT, l.batch, l.out_c, l.out_h, l.out_w);
+ cudnnSetTensor4dDescriptor(l.normTensorDesc, CUDNN_TENSOR_NCHW, CUDNN_DATA_FLOAT, 1, l.out_c, 1, 1);
+#endif
}
#endif
l.activation = activation;
- fprintf(stderr, "Connected Layer: %d inputs, %d outputs\n", inputs, outputs);
+ fprintf(stderr, "connected %4d -> %4d\n", inputs, outputs);
return l;
}
@@ -187,7 +201,7 @@
{
int i, j;
for(i = 0; i < l.outputs; ++i){
- float scale = l.scales[i]/sqrt(l.rolling_variance[i] + .00001);
+ float scale = l.scales[i]/sqrt(l.rolling_variance[i] + .000001);
for(j = 0; j < l.inputs; ++j){
l.weights[i*l.inputs + j] *= scale;
}
@@ -198,6 +212,25 @@
}
}
+
+void statistics_connected_layer(layer l)
+{
+ if(l.batch_normalize){
+ printf("Scales ");
+ print_statistics(l.scales, l.outputs);
+ /*
+ printf("Rolling Mean ");
+ print_statistics(l.rolling_mean, l.outputs);
+ printf("Rolling Variance ");
+ print_statistics(l.rolling_variance, l.outputs);
+ */
+ }
+ printf("Biases ");
+ print_statistics(l.biases, l.outputs);
+ printf("Weights ");
+ print_statistics(l.weights, l.outputs);
+}
+
#ifdef GPU
void pull_connected_layer(connected_layer l)
@@ -253,12 +286,13 @@
float * b = l.weights_gpu;
float * c = l.output_gpu;
gemm_ongpu(0,1,m,n,k,1,a,k,b,k,1,c,n);
- if(l.batch_normalize){
- forward_batchnorm_layer_gpu(l, state);
- }
- for(i = 0; i < l.batch; ++i){
- axpy_ongpu(l.outputs, 1, l.biases_gpu, 1, l.output_gpu + i*l.outputs, 1);
- }
+ if (l.batch_normalize) {
+ forward_batchnorm_layer_gpu(l, state);
+ }
+ else {
+ add_bias_gpu(l.output_gpu, l.biases_gpu, l.batch, l.outputs, 1);
+ }
+ //for(i = 0; i < l.batch; ++i) axpy_ongpu(l.outputs, 1, l.biases_gpu, 1, l.output_gpu + i*l.outputs, 1);
activate_array_ongpu(l.output_gpu, l.outputs*l.batch, l.activation);
}
--
Gitblit v1.10.0