Getting error to use Multiple GPU in pytorch

vision
1

Hi, I am running my deep learning model using pytorch. When I am using single GUP, my model is running fine but facing problem using multiple GPU. I searched this problem in google and got idea to downgrade pytorch version 0.3.0. I don’t know where this is the way to rid of my problem or not. However I was trying to downgrade pytorch using
pip install torch=0.3.0
But getting error to install. Is there someone to help me to downgrade pytorch using pip.

I got below error during running multiple GPU:
Exception NameError: “global name ‘FileNotFoundError’ is not defined” in <bound method _DataLoaderIter.del of <torch.utils.data.dataloader._DataLoaderIter object at 0x7fb3bbaba8d0>> ignored
Traceback (most recent call last):
File “train.py”, line 61, in
Variable(data[‘image’]), Variable(data[‘feat’]), infer=save_fake)
File “/netscratch/reza/venv/lib/python2.7/site-packages/torch/nn/modules/module.py”, line 491, in call
result = self.forward(*input, **kwargs)
File “/netscratch/reza/venv/lib/python2.7/site-packages/torch/nn/parallel/data_parallel.py”, line 114, in forward
outputs = self.parallel_apply(replicas, inputs, kwargs)
File “/netscratch/reza/venv/lib/python2.7/site-packages/torch/nn/parallel/data_parallel.py”, line 124, in parallel_apply
return parallel_apply(replicas, inputs, kwargs, self.device_ids[:len(replicas)])
File “/netscratch/reza/venv/lib/python2.7/site-packages/torch/nn/parallel/parallel_apply.py”, line 65, in parallel_apply
raise output
TypeError: forward() takes at least 5 arguments (2 given)

2

I don’t think downgrading is an optimal solution. I would rather want to fix the issue using the latest release.
Which PyTorch version are you currently using? If < 0.4.1, could you update to it?

Also, could you share a bit of executable code so that we can have a look at this problem?

3

Hi ptrblck, thanks for your quick reply. Earlier I was using torch 0.4.0. After getting your reply I updated it into torch 0.4.1 but facing the same problem.

Actually I am running pix2pixHD pytorch model for my image dataset. It run smoothly when I use single GPU but getting error if I want to use multiple GPU. My image size is big and having large dataset so I have to use multiple GPU to save time. Please help me to rid of this problem.

4

Could you post a small executable code snippet reproducing this error?
This would make it easier to debug this issue.

5

Here is the screen shot of the error.

Screenshot%20from%202018-08-07%2017-33-36
Screenshot%20from%202018-08-07%2017-33-361366×768 234 KB

6

The above problem has solved. To see the error it seems the error come from torch but actually there was a limitation in pix2pixHD model and also for torch version 0.4.0. Finally the model ran with multiple GPU after update torch version (0.4.1) and avoid model limitation.

8

I have get error to use Multiple GPU in pytorch when i train pix2pixHD，how to solve the limitation in pix2pixHD model？