Shuffling automatically set to True in Config API - not compatible with iterable datasets

🐛 Bug Report

In the config API it appears that it is assumed that the train dataloader should be shuffled, as per the following line: https://github.com/catalyst-team/catalyst/blob/a7bc302a762d7d9f462ded6d9cd6ae70f8b656aa/catalyst/utils/data.py#L201

This behaviour is particularly undesirable when using iterable datasets, as they are incompatible with shuffle=True. It would probably be better to let the user specify the desired value of shuffle in the config.yml file. At the moment, if the user passes: loaders_params: {"shuffle": False} it is overwritten in the mentioned line, which leads to the pytorch error:

ValueError: DataLoader with IterableDataset: expected unspecified shuffle option, but got shuffle=True

How To Reproduce

Steps to reproduce the behavior:

1. Create an iterable dataset that is created in customRunner.get_datasets()
2. Use the config API to create a loader for the dataset with e.g. the following params:

 loaders: &loaders     
   batch_size: None  
   num_workers: 0  
   drop_last: False 
   per_gpu_scaling: False  
   loaders_params: {"shuffle": False} 

1. See the following pytorch error:

ValueError: DataLoader with IterableDataset: expected unspecified shuffle option, but got shuffle=True

Expected behavior

When passing loaders_params: {"shuffle": False} one would expect shuffling to be turned off for both loaders.

Environment

Catalyst version: 21.03.2
PyTorch version: 1.7.1
TensorFlow version: N/A
TensorBoard version: 2.4.1
OS: Ubuntu 16.04.6 LTS
Python version: 3.7
Nvidia driver version: 460.32.03

Checklist

• [ x] bug description
• [ x] steps to reproduce
• [ x] expected behavior
• [ x] environment
• code sample / screenshots

[Scitator]

Oh, I see, as a possible workaround:

loaders: &loaders     
  batch_size: None  
  num_workers: 0  
  drop_last: False 
  per_gpu_scaling: False  
  loaders_params:
    train: {"shuffle": False}
    valid:  {"shuffle": False}
    other_loader_key: {"shuffle": False}

Nevetheless, it would be great if you could inject a hotfix here.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.