Error While Running distributed/FSDP/T5_training.py

[nariaki3551]

Context

While running the distributed/FSDP/T5_training.py example, I encountered an error when loading the wikihow dataset. I would like to know if this is a bug or if there is a way to resolve it.

• PyTorch version: 2.7.0a0+git49bdc41
• Operating System and version: Ubuntu 20.04

Your Environment

• Installed using source? [yes/no]: yes
• Are you planning to deploy it using docker container? [yes/no]: no
• Is it a CPU or GPU environment?: GPU (CUDA)
• Which example are you using: distributed/FSDP/T5_training.py
• Link to code or data to repro [if any]: N/A
• version: commit 1bef748

Expected Behavior

The wikihow dataset should be successfully loaded using the following command:

dataset = load_dataset('wikihow', 'all', data_dir='data/')

Current Behavior

The script fails with a ConnectionError, indicating that the dataset could not be downloaded from the specified URL.

Possible Solution

The issue might be related to the URL used in the dataset script:
https://raw.githubusercontent.com/mahnazkoupaee/WikiHow-Dataset/master/all_train.txt.
If the file is unavailable, an updated URL or alternative data source could resolve the issue.

Steps to Reproduce

1. Install the requirements:

   sh download_dataset.sh
   pip install -r requirements.txt

2. Run the following command:

   OMP_NUM_THREADS=1 torchrun --nproc_per_node 1 T5_training.py

Failure Logs [if any]

Downloading and preparing dataset wikihow/all (download: Unknown size, generated: Unknown size, post-processed: Unknown size, total: Unknown size) to /home/tateiwa/.cache/huggingface/datasets/wikihow/all/1.2.0/cfb412ca2191fac028cae9a5a9a03ba21b08ff2b4bf46f8a0473d7303a3e3683...
[rank0]: Traceback (most recent call last):
[rank0]:   File "/home/tateiwa/pytorch_test/examples/distributed/FSDP/T5_training.py", line 222, in <module>
[rank0]:     fsdp_main(args)
[rank0]:   File "/home/tateiwa/pytorch_test/examples/distributed/FSDP/T5_training.py", line 90, in fsdp_main
[rank0]:     dataset = load_dataset('wikihow', 'all', data_dir='data/')
...
[rank0]: ConnectionError: Couldn't reach https://raw.githubusercontent.com/mahnazkoupaee/WikiHow-Dataset/master/all_train.txt

full log is here.

ownloading and preparing dataset wikihow/all (download: Unknown size, generated: Unknown size, post-processed: Unknown sizetotal: Unknown size) to /home/tateiwa/.cache/huggingface/datasets/wikihow/all/1.2.0/cfb412ca2191fac028cae9a5a9a03ba21b08ff2b4bf46f8a0473d7303a3e3683...
[rank0]: Traceback (most recent call last):
[rank0]:   File "/home/tateiwa/pytorch_test/examples/distributed/FSDP/T5_training.py", line 222, in <module>
[rank0]:     fsdp_main(args)
[rank0]:   File "/home/tateiwa/pytorch_test/examples/distributed/FSDP/T5_training.py", line 90, in fsdp_main
[rank0]:     dataset = load_dataset('wikihow', 'all', data_dir='data/')
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/load.py", line 548, in load_dataset
[rank0]:     builder_instance.download_and_prepare(
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/builder.py", line 462, in download_and_prepare
[rank0]:     self._download_and_prepare(
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/builder.py", line 518, in _download_and_prepare
[rank0]:     split_generators = self._split_generators(dl_manager, **split_generators_kwargs)
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/datasets/wikihow/cfb412ca2191fac028cae9a5a9a03ba21b08ff2b4bf46f8a0473d7303a3e3683/wikihow.py", line 126, in _split_generators
[rank0]:     dl_path = dl_manager.download_and_extract(_URLS)
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/utils/download_manager.py", line 220, in download_and_extract
[rank0]:     return self.extract(self.download(url_or_urls))
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/utils/download_manager.py", line 155, in download
[rank0]:     downloaded_path_or_paths = map_nested(
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/utils/py_utils.py", line 163, in map_nested
[rank0]:     return {
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/utils/py_utils.py", line 164, in <dictcomp>
[rank0]:     k: map_nested(
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/utils/py_utils.py", line 191, in map_nested
[rank0]:     return function(data_struct)
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/utils/download_manager.py", line 156, in <lambda>
[rank0]:     lambda url: cached_path(url, download_config=self._download_config,), url_or_urls,
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/utils/file_utils.py", line 191, in cached_path
[rank0]:     output_path = get_from_cache(
[rank0]:   File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/nlp/utils/file_utils.py", line 356, in get_from_cache
[rank0]:     raise ConnectionError("Couldn't reach {}".format(url))
[rank0]: ConnectionError: Couldn't reach https://raw.githubusercontent.com/mahnazkoupaee/WikiHow-Dataset/master/all_train.txt
E0118 06:12:01.182000 1191202 torch/distributed/elastic/multiprocessing/api.py:870] failed (exitcode: 1) local_rank: 0 (pid: 1191213) of binary: /home/tateiwa/pytorch_test/venv/bin/python
Traceback (most recent call last):
  File "/home/tateiwa/pytorch_test/venv/bin/torchrun", line 33, in <module>
    sys.exit(load_entry_point('torch==2.7.0a0+git49bdc41', 'console_scripts', 'torchrun')())
  File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
    return f(*args, **kwargs)
  File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/torch/distributed/run.py", line 918, in main
    run(args)
  File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/torch/distributed/run.py", line 909, in run
    elastic_launch(
  File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
    return launch_agent(self._config, self._entrypoint, list(args))
  File "/home/tateiwa/pytorch_test/venv/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
    raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
============================================================
T5_training.py FAILED
------------------------------------------------------------
Failures:
  <NO_OTHER_FAILURES>
------------------------------------------------------------
Root Cause (first observed failure):
[0]:
  time      : 2025-01-18_06:12:01
  host      : snail03
  rank      : 0 (local_rank: 0)
  exitcode  : 1 (pid: 1191213)
  error_file: <N/A>
  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
============================================================

Could you provide guidance on how to resolve this issue? Alternatively, if this is a bug, are there any workarounds or fixes available?

Thank you for your help!