Changing batch_size and setting num_workers > 0 or 1 throws the following: 'TypeError: __init__() missing 1 required positional argument: 'dtype'

1

Hi all, still learning this module (and about 6-8 months of python coding experience). I was wondering if some one can point me to the correct direction in what this may be telling me? (error in the image below)
Trains but stops and complains

I am following a book along (Deep Learning with Pytorch by Eli Stevens and Thomas Viehmann) about implementing some deep learning models but I am not entirely sure what is going on in this error (been stuck for 1 day or 2 now). If I set the num workers to 0 and some batch_size, the model trains but if I start setting the num_workers to 2 (in the figure above, the num_workers was set to 2 by default) or higher along with increasing the batch_sizes this error usually will be thrown…if i set num_workers higher like 12 I get an “OSError: [WinError 1455] The paging file is too small for this operation to complete”

Other things I notice:

  • from the windows task manager GPU utilization for cuda: it barely goes over 10% (i see spikes)
  • the memory usage does go up

I have the following:
OS: Windows 10
CPU: AMD Ryzen 9 3900X 12-Core Processor
GPU: GTX 1660ti
32gb ram

So from these, do you think it could be a hardware issue or what? I’ve been trying to figure this out, no idea what else to do so this is my first post…so hopefully someone sees this and can point me to some direction. Thank you very much in advanced

1 Like