root
/
MNN
mirror of https://github.com/alibaba/MNN.git
1
0
Fork 0
Code Issues Actions 1 Packages Projects Releases Wiki Activity

quant weight valid range issue

This commit is contained in:
tianbu.xsw 2021-09-02 15:35:16 +08:00 committed by xiaying
parent 5dfe97e4c8
commit 612199d0ee
1 changed files with 3 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
source/backend/opencl/execution/buffer/ConvBufExecution.cpp
View File

@ -306,6 +306,7 @@ ConvBufExecution::ConvBufExecution(const std::vector<Tensor *> &inputs, const st
std::string kernelName = "conv_2d_c4h1w4";
mInputChannel = inputs[0]->channel();
std::shared_ptr<ConvolutionCommon::Int8Common> quanCommon;
if (inputs.size() != 1) {
// Multi - Input
mConv1x1Opt = false;
@ -315,7 +316,6 @@ ConvBufExecution::ConvBufExecution(const std::vector<Tensor *> &inputs, const st
mRasterExe.reset(new RasterBufExecution({virtualFilter.get()}, mOpenCLBackend));
} else {
int weightSize = 0;
std::shared_ptr<ConvolutionCommon::Int8Common> quanCommon;
ConvolutionCommon::getConvParameters(&quanCommon, conv2dParams, &mFilterDataPtr, &weightSize);
//select opt conv method
mConv1x1Opt = (mKernelHeight == mKernelWidth && mKernelHeight == 1 && mPaddings[0] == 0 &&
@ -470,10 +470,11 @@ ErrorCode ConvBufExecution::onResize(const std::vector<Tensor *> &inputs, const
mLocalWorkSize = {retTune.first[0], retTune.first[1]};
}
}
std::shared_ptr<ConvolutionCommon::Int8Common> quanCommon;
int min_index = min_cost.second;
if(min_index >= c8_index_start) {//if best kernel is "conv_2d_1x1_c8h1w4", set weight packCout to 8
int weightSize = 0;
std::shared_ptr<ConvolutionCommon::Int8Common> quanCommon;
ConvolutionCommon::getConvParameters(&quanCommon, mConv2dParams, &mFilterDataPtr, &weightSize);
setConv1x1WeightBuffer(8, 4, mFilterDataPtr);
}