save_state() and load_state() do not work correctly with multi-gpu with shuffle=True in dataloader

[isayoften]

System Info

- `Accelerate` version: 1.0.0
- Platform: Linux-5.15.154+-x86_64-with-glibc2.35
- `accelerate` bash location: /opt/conda/bin/accelerate
- Python version: 3.10.14
- Numpy version: 1.26.4
- PyTorch version (GPU?): 2.4.0 (True)
- PyTorch XPU available: False
- PyTorch NPU available: False
- PyTorch MLU available: False
- PyTorch MUSA available: False
- System RAM: 31.36 GB
- GPU type: Tesla T4
- `Accelerate` default config:
    Not found

Information

• [ ] The official example scripts
• [X] My own modified scripts

Tasks

• [ ] One of the scripts in the examples/ folder of Accelerate or an officially supported no_trainer script in the examples folder of the transformers repo (such as run_no_trainer_glue.py)
• [X] My own task or dataset (give details below)

Reproduction

https://www.kaggle.com/code/amanattheedge/demonstration

Expected behavior

Maybe I'm doing something wrong, but save_state() and load_state() should memorize the RNG states so that the previous shuffling of data within the new epoch can be restored.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.