Keep Training but no output

[Lyt859165290]

I used the suggested command 'python main.py \ --dataset cifar10 \ --labels data-local/labels/cifar10/1000_balanced_labels/00.txt \ --arch cifar_shakeshake26 \ --consistency 100.0 \ --consistency-rampup 5 \ --labeled-batch-size 62 \ --epochs 180 \ --lr-rampdown-epochs 210

I'm using ubuntu 18.04 python3.6, pytorch 0.3.0. numpy 1.14.2, and cuda8.0.and 2 gtx1080Ti When I run the main.py, it can start training but there is no output information(epochs ,accuracy) during the training process (in a few hours), the only outputs are like this:

INFO:main:=> creating model 'cifar_shakeshake26' INFO:main:=> creating EMA model 'cifar_shakeshake26' INFO:main: List of model parameters:

module.conv1.weight 16 3 3 3 = 432 module.layer1.0.conv_a1.weight 96 16 3 3 = 13,824 module.layer1.0.bn_a1.weight 96 = 96 module.layer1.0.bn_a1.bias 96 = 96 ..... module.fc2.weight 10 * 384 = 3,840 module.fc2.bias 10 = 10

all parameters sum of above = 26,197,316

I have checked the results folder and there is no checkpoint file in it.

[CheukNgai]

Hi, I got the same hang up issue when training with pytorch 0.3.1 and torchvision 0.2.0 under 4 gtx1080Ti. Could anybody tell me how to fix it? @tarvaina