kiyotaka-xiaoli

icon company Tencent

Results 2 comments of kiyotaka-xiaoli

使用mnn_offline_quant.py时，执行到quant_func方法的net.forward时候报错了Tensorflow's interp's shape should be length two, Segmentation fault (core dumped)

看起来是到c++的代码了，我该如何往下排查

使用mnn_offline_quant.py时，执行到quant_func方法的net.forward时候报错了Tensorflow's interp's shape should be length two, Segmentation fault (core dumped)

使用gdb打印出来的堆栈日志 `0x00007fffe7ef27b1 in MNN::Express::ConvBNReluFusedModule::fakeQuantFeatureWithMinMax(MNN::Express::VARP, MNN::Express::VARP, MNN::Express::VARP, MNN::Express::VARP, std::vector) () from /home/ubuntu/miniconda3/envs/mnn/lib/python3.7/site-packages/_mnncengine.cpython-37m-x86_64-linux-gnu.so (gdb) backtrace #0 0x00007fffe7ef27b1 in MNN::Express::ConvBNReluFusedModule::fakeQuantFeatureWithMinMax(MNN::Express::VARP, MNN::Express::VARP, MNN::Express::VARP, MNN::Express::VARP, std::vector) () from /home/ubuntu/miniconda3/envs/mnn/lib/python3.7/site-packages/_mnncengine.cpython-37m-x86_64-linux-gnu.so #1 0x00007fffe7f0b329 in MNN::Express::ConvBNReluFusedModule::onForward(std::vector const&) ()...