Regarding choosing BATCH_SIZE

[mouryarahul]

Hi all,
Maybe I'm missing something, but in almost all the examples, the batch_size = 1 or batch_size = num_gpus. I have a single GPU, can I still use batch_size > 1 ?
When I'm trying to use batch_size = 2, I'm getting the following error message:

Traceback (most recent call last):
File "/media/rahul/DATA/WorkSpace/Multimodal-Data-Processing/Projects/MRI-OPTNET/Codes/fastMRI_demo.py", line 220, in
run_cli()
File "/media/rahul/DATA/WorkSpace/Multimodal-Data-Processing/Projects/MRI-OPTNET/Codes/fastMRI_demo.py", line 217, in run_cli
cli_main(args)
File "/media/rahul/DATA/WorkSpace/Multimodal-Data-Processing/Projects/MRI-OPTNET/Codes/fastMRI_demo.py", line 170, in cli_main
for batch in train_dataloader:
File "/home/rahul/anaconda3/envs/pytorch/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 652, in next
data = self._next_data()
File "/home/rahul/anaconda3/envs/pytorch/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 1347, in _next_data
return self._process_data(data)
File "/home/rahul/anaconda3/envs/pytorch/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 1373, in _process_data
data.reraise()
File "/home/rahul/anaconda3/envs/pytorch/lib/python3.10/site-packages/torch/_utils.py", line 461, in reraise
raise exception
RuntimeError: Caught RuntimeError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/home/rahul/anaconda3/envs/pytorch/lib/python3.10/site-packages/torch/utils/data/_utils/worker.py", line 302, in _worker_loop
data = fetcher.fetch(index)
File "/home/rahul/anaconda3/envs/pytorch/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py", line 52, in fetch
return self.collate_fn(data)
File "/home/rahul/anaconda3/envs/pytorch/lib/python3.10/site-packages/torch/utils/data/_utils/collate.py", line 165, in default_collate
return elem_type(*(default_collate(samples) for samples in zip(batch)))
File "/home/rahul/anaconda3/envs/pytorch/lib/python3.10/site-packages/torch/utils/data/_utils/collate.py", line 165, in
return elem_type((default_collate(samples) for samples in zip(*batch)))
File "/home/rahul/anaconda3/envs/pytorch/lib/python3.10/site-packages/torch/utils/data/utils/collate.py", line 140, in default_collate
out = elem.new(storage).resize(len(batch), *list(elem.size()))
RuntimeError: Trying to resize storage that is not resizable

[mmuckley]

Hello @mouryarahul, the code for the variational network is configured to only work with a batch size of 1 (and we should probably raise an exception if this isn't the case). The reason is that all the different k-spaces in the dataset have different shapes and if you didn't use a batch size of 1 then you'd have to figure out how to collate them. There are also several parts of the code that are basically hardcoded for a batch size of 1.

At least for the U-Net with knee data, I think you can use a batch sizes greater than 1 because every sample in the knee dataset has a crop size of 320 x 320.

[mouryarahul]

@mmuckley Thanks for the clarification.