mirror/darknet
1
0
Fork 0
mirror of https://github.com/pjreddie/darknet.git synced 2023-08-10 21:13:14 +03:00
Code Issues Projects Releases Wiki Activity

Improve training performance - batch-norm using cuDNN.

Browse Source
This commit is contained in:
AlexeyAB
2018-03-20 02:16:51 +03:00
parent 2f52cfeb07
commit 537d135feb
12 changed files with 193 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
src/connected_layer.c
View File

@ -97,6 +97,12 @@ connected_layer make_connected_layer(int batch, int inputs, int outputs, ACTIVAT
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;
@ -280,12 +286,13 @@ void forward_connected_layer_gpu(connected_layer l, network_state state)
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);
}