Build Quantize Error for ResNet50
pras...@gmail.com
Hi,
I am getting the error shown below, while running the following command:
./build/tools/ristretto quantize
--model=$HOME/ResNet-50-deploy.prototxt
--weights=$HOME/ResNet-50-model.caffemodel
--model_quantized=$HOME/minifloat_resnet.prototxt
--iterations=2000 --trimming_mode=mini_floating_point --error_margin=1 --gpu=0
======ERROR===================================================
*** Aborted at 1528805069 (unix time) try "date -d @1528805069" if you are using GNU date ***
PC: @ 0x2b85116f9825 caffe::Blob<>::mutable_gpu_data()
*** SIGSEGV (@0x0) received by PID 30284 (TID 0x2b8526059a80) from PID 0; stack trace: ***
@ 0x3afda0f7e0 (unknown)
@ 0x2b85116f9825 caffe::Blob<>::mutable_gpu_data()
@ 0x2b8511849c59 caffe::ConvolutionRistrettoLayer<>::Forward_gpu()
@ 0x2b8511825c95 caffe::Net<>::ForwardFromTo()
@ 0x2b8511826007 caffe::Net<>::Forward()
@ 0x2b85118261da caffe::Net<>::Forward()
@ 0x2b85117068d2 Quantization::RunForwardBatches()
@ 0x2b851170a0bc Quantization::Quantize2MiniFloatingPoint()
@ 0x2b851170abc8 Quantization::QuantizeNet()
@ 0x403371 quantize()
@ 0x40281c main
@ 0x3afce1ed1d (unknown)
@ 0x403069 (unknown)
Segmentation fault
The framework has been built for CPU only. Why is the function caffe::ConvolutionRistrettoLayer<>::Forward_gpu() called when --gpu=0?
It would be grateful if anyone can resolve the above issue.
Thanks in Advance,
Prasad.
