Hangzhi Guo

Hi @chnyutao thanks for raising this issue. Yes, this is a bug. Feel free to submit a PR. Note that we are using nbdev for development/document/test. See: https://nbdev.fast.ai/tutorials/tutorial.html#make-your-first-edit

Hi @pluiez All dataloader backends do not assume the statefulness. For now, you should manually store the states (e.g., batch index, arguments, seeds, etc), and load the states by re-init...

@GarrettMerz Thank you very much for your constructive feedback and suggestions. Below, we clarify your concerns about our library: > -The "installation instructions" are incomplete. I'd like some more clarity...

Hi @duhd1993, Is there anything else you need our help to clarify about this repo?

@duhd1993 Thank you very much for your constructive feedback and suggestions. Below, we clarify your concerns about our library: > The installation instruction is not complete. When I tried to...

Hi @duhd1993, thank you for your help in improving this submission. I am wondering if you have other questions or need any additional information about this repo. cc @Fei-Tao

Hi @aspannaus - Thanks for your suggestions. This library tries to bring the PyTorch dataloader API that works on the four supported datasets and three backends. It seems like your...

Thanks for your feedback. One workaround for now is to call `manual_seed` each time before initializing a dataloader. E.g., ```python data_key, model_key, loader_key, train_key, sample_key = jr.split(key, 5) # first...

Another idea is to introduce the `Generator` API, which is actually a part of PyTorch DataLoader's class parameter. We could define the `jdl.DataLoader` as something like this: ```python class DataLoader:...

This is addressed in #41 Thank you for your suggestion. Feel free to play around with this API and let me know if you encounter any issues.