diff options
| author | Patrick Follmann <follmann@mvtec.com> | 2016-12-29 14:46:16 +0100 |
|---|---|---|
| committer | Jeff Donahue <jeff.donahue@gmail.com> | 2017-04-13 13:29:28 -0700 |
| commit | e93fcd267582888f960ca48d6e0c2e719d4ea09b (patch) | |
| tree | 30426e3968c5c6d5b235b3ea004e9e19b853a996 | |
| parent | e98023af4a570e3105486b661e4c4d1855c0dd79 (diff) | |
| download | caffe-e93fcd267582888f960ca48d6e0c2e719d4ea09b.tar.gz caffe-e93fcd267582888f960ca48d6e0c2e719d4ea09b.tar.bz2 caffe-e93fcd267582888f960ca48d6e0c2e719d4ea09b.zip | |
GPU BatchNormLayer: replace powx with mul and sqrt
| -rw-r--r-- | src/caffe/layers/batch_norm_layer.cu | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/src/caffe/layers/batch_norm_layer.cu b/src/caffe/layers/batch_norm_layer.cu index c21713c8..a35e778e 100644 --- a/src/caffe/layers/batch_norm_layer.cu +++ b/src/caffe/layers/batch_norm_layer.cu @@ -48,14 +48,14 @@ void BatchNormLayer<Dtype>::Forward_gpu(const vector<Blob<Dtype>*>& bottom, if (!use_global_stats_) { // compute variance using var(X) = E((X-EX)^2) - caffe_gpu_powx(top[0]->count(), top_data, Dtype(2), + caffe_gpu_mul(top[0]->count(), top[0]->gpu_data(), top[0]->gpu_data(), temp_.mutable_gpu_data()); // (X-EX)^2 caffe_gpu_gemv<Dtype>(CblasNoTrans, channels_ * num, spatial_dim, 1. / (num * spatial_dim), temp_.gpu_data(), spatial_sum_multiplier_.gpu_data(), 0., num_by_chans_.mutable_gpu_data()); - caffe_gpu_gemv<Dtype>(CblasTrans, num, channels_, 1., - num_by_chans_.gpu_data(), batch_sum_multiplier_.gpu_data(), 0., + caffe_gpu_gemv<Dtype>(CblasTrans, num, channels_, Dtype(1.), + num_by_chans_.gpu_data(), batch_sum_multiplier_.gpu_data(), Dtype(0.), variance_.mutable_gpu_data()); // E((X_EX)^2) // compute and save moving average @@ -72,7 +72,7 @@ void BatchNormLayer<Dtype>::Forward_gpu(const vector<Blob<Dtype>*>& bottom, // normalize variance caffe_gpu_add_scalar(variance_.count(), eps_, variance_.mutable_gpu_data()); - caffe_gpu_powx(variance_.count(), variance_.gpu_data(), Dtype(0.5), + caffe_gpu_sqrt(variance_.count(), variance_.gpu_data(), variance_.mutable_gpu_data()); // replicate variance to input size |