potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-29 07:13:02 +04:00
Code Packages Releases Wiki Activity

Merge pull request #17939 from YashasSamaga:cuda4dnn-fix-eltwise-fusion

Browse Source 
* fix eltwise fusion segfault, more eltwise fusions, fix power fusion

* add assertion
This commit is contained in:
Yashas Samaga B L
2020-08-01 17:33:07 +05:30
committed by GitHub
parent e421233a1d
commit f53f491cd2
3 changed files with 25 additions and 37 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/dnn/src/layers/convolution_layer.cpp
+6 -8
View File
@@ -250,7 +250,8 @@ public:
#ifdef HAVE_CUDA
cuda4dnn::ConvolutionConfiguration::FusionMode cudaFusionMode;
cuda4dnn::ConvolutionConfiguration::ActivationType cudaActType;
float cuda_relu_slope, cuda_crelu_floor, cuda_crelu_ceil, cuda_power_exp;
float cuda_relu_slope, cuda_crelu_floor, cuda_crelu_ceil;
float cuda_power_exp, cuda_power_scale, cuda_power_shift;
#endif
ConvolutionLayerImpl(const LayerParams &params) : BaseConvolutionLayerImpl(params)
@@ -457,13 +458,8 @@ public:
Ptr<PowerLayer> activ_power = activ.dynamicCast<PowerLayer>();
if (!activ_power.empty())
{
if (activ_power->scale != 1.f || activ_power->shift != 0.f)
{
const int outCh = blobs[0].size[0];
fuseWeights(Mat(1, outCh, CV_32F, Scalar(activ_power->scale)),
Mat(1, outCh, CV_32F, Scalar(activ_power->shift)));
}
cuda_power_scale = activ_power->scale;
cuda_power_shift = activ_power->shift;
cuda_power_exp = activ_power->power;
cudaActType = cuda4dnn::ConvolutionConfiguration::ActivationType::POWER;
}
@@ -1591,6 +1587,8 @@ public:
config.crelu_floor = cuda_crelu_floor;
config.crelu_ceil = cuda_crelu_ceil;
config.power_exp = cuda_power_exp;
config.power_scale = cuda_power_scale;
config.power_shift = cuda_power_shift;
Mat filtersMat = fusedWeights ? weightsMat : blobs[0];
Mat biasMat = (hasBias() || fusedBias) ? Mat(output_feature_maps, 1, CV_32F, biasvec.data()) : Mat();