lqyiii
lqyiii
i also encountered this problem, using gpu v100. a little trick was used to avoid it, that is using gpu to train model except to calculate model loss, before calculating...
@ccnankai @10183308 have you solved this problem? i also meet this error...
i added this command (--expt-relaxed-constexpr) after which line has nvcc. the error disappear. eg, nvcc -std=c++11 -c -o roi_pooling_op.cu.o roi_pooling_op_gpu.cu.cc \ -I $TF_INC -I $NSYNC_INC -D GOOGLE_CUDA=1 -L $CUDA_HOME/lib64 -x...