imagenet_256_cc.yml runtime error

[bahjat-kawar]

imagenet_256_cc.yml runtime error

[lshaw8317]

#18 is related. I also had the same error. Adding global seed_worker to Diffusion.sample_sequence in diffusion.py fails to resolve issue:

Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "C:\Users\shaw\Anaconda3\lib\multiprocessing\spawn.py", line 116, in spawn_main
    exitcode = _main(fd, parent_sentinel)
  File "C:\Users\shaw\Anaconda3\lib\multiprocessing\spawn.py", line 126, in _main
    self = reduction.pickle.load(from_parent)
AttributeError: Can't get attribute 'seed_worker' on <module 'runners.diffusion' from 'C:\\Users\\shaw\\Documents\\Year 2\\Diffusion Models\\ddrm\\runners\\diffusion.py'>

The reason (in my case) is that when running on Windows the multiprocessing module uses spawn and so one must (according to docs):

Wrap most of you main script’s code within if name == 'main': block, to make sure it doesn’t run again (most likely generating error) when each worker process is launched. You can place your dataset and DataLoader instance creation logic here, as it doesn’t need to be re-executed in workers.

Make sure that any custom collate_fn, worker_init_fn or dataset code is declared as top level definitions, outside of the main check. This ensures that they are available in worker processes. (this is needed since functions are pickled as references only, not bytecode.)

It is difficult to implement this advice since the seed_worker function needs access to the input args coming from the config file.
Simplest "solution" was to just set the worker_init_fn argument to None as below (within Diffusion.sample_sequence):

val_loader = data.DataLoader(
            test_dataset,
            batch_size=config.sampling.batch_size,
            shuffle=True,
            num_workers=config.data.num_workers,
            worker_init_fn=None,
            generator=g,
        )