Could not run the code

[wangq95]

Hi, @speedinghzl , I try to train on Cityscapes dataset using pytorch 0.4.0, but I got an error as follows:

Traceback (most recent call last): File "train.py", line 253, in main() File "train.py", line 217, in main preds = model(images, args.recurrence) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 491, in call result = self.forward(*input, **kwargs) File "/usr/local/lib/python3.6/dist-packages/torch/nn/parallel/data_parallel.py", line 112, in forward return self.module(*inputs[0], **kwargs[0]) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 491, in call result = self.forward(*input, **kwargs) File "/userhome/segmentation/CCNet/networks/ccnet.py", line 196, in forward x = self.relu1(self.bn1(self.conv1(x))) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 491, in call result = self.forward(*input, **kwargs) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/conv.py", line 301, in forward self.padding, self.dilation, self.groups) RuntimeError: CUDNN_STATUS_MAPPING_ERROR

Could you give me some advise? Thanks

[sreeragh-ar]

One possibility can be that you have less number of visible/available GPUs than the configured number of GPUs to use. i.e

In the run_local.sh file (in branch pytorch-1.1)

--nproc_per_node=4 specifies to use 4 GPUs for training but if you have GPU_IDS=0,1 only 2 GPUs are made visible

This may cause an error.