IWSLT datasets are not properly unpacked from tgz file

[MarkusSpanring]

🐛 Bug

IWSLT datasets are not properly unpacked from the downloaded tgz file when using torchtext.datasets.IWSLT*.
When I unpack the nested tgz files by hand everything works as expected.

To Reproduce

E.g. for IWSLT2017 download the .tgz file from here and put it under ./data/datasets/IWSLT2017

import torchtext

a = iter(torchtext.datasets.IWSLT2017(root="./data", split="train"))
next(a)

---------------------------------------------------------------------------
Exception                                 Traceback (most recent call last)
Input In [9], in <cell line: 4>()
      1 import torchtext
      3 a = iter(torchtext.datasets.IWSLT2016(root="./data", split="train"))
----> 4 next(a)

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/_typing.py:514, in hook_iterator.<locals>.wrap_generator(*args, **kwargs)
    512         response = gen.send(None)
    513 else:
--> 514     response = gen.send(None)
    516 while True:
    517     request = yield response

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/iter/grouping.py:41, in ShardingFilterIterDataPipe.__iter__(self)
     40 def __iter__(self):
---> 41     for i, item in enumerate(self.source_datapipe):
     42         if i % self.num_of_instances == self.instance_id:
     43             yield item

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/_typing.py:514, in hook_iterator.<locals>.wrap_generator(*args, **kwargs)
    512         response = gen.send(None)
    513 else:
--> 514     response = gen.send(None)
    516 while True:
    517     request = yield response

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/iter/combinatorics.py:122, in ShufflerIterDataPipe.__iter__(self)
    120 def __iter__(self) -> Iterator[T_co]:
    121     if not self._enabled:
--> 122         for x in self.datapipe:
    123             yield x
    124     else:

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/_typing.py:514, in hook_iterator.<locals>.wrap_generator(*args, **kwargs)
    512         response = gen.send(None)
    513 else:
--> 514     response = gen.send(None)
    516 while True:
    517     request = yield response

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/iter/combining.py:513, in ZipperIterDataPipe.__iter__(self)
    512 def __iter__(self) -> Iterator[Tuple[T_co]]:
--> 513     for data in zip(*self.datapipes):
    514         yield data

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/_typing.py:514, in hook_iterator.<locals>.wrap_generator(*args, **kwargs)
    512         response = gen.send(None)
    513 else:
--> 514     response = gen.send(None)
    516 while True:
    517     request = yield response

File /user/conda/envs/storch/lib/python3.9/site-packages/torchdata/datapipes/iter/util/plain_text_reader.py:121, in LineReaderIterDataPipe.__iter__(self)
    120 def __iter__(self) -> Iterator[Union[Str_Or_Bytes, Tuple[str, Str_Or_Bytes]]]:
--> 121     for path, file in self.source_datapipe:
    122         stream = self._helper.skip_lines(file)
    123         stream = self._helper.strip_newline(stream)

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/_typing.py:514, in hook_iterator.<locals>.wrap_generator(*args, **kwargs)
    512         response = gen.send(None)
    513 else:
--> 514     response = gen.send(None)
    516 while True:
    517     request = yield response

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/iter/fileopener.py:68, in FileOpenerIterDataPipe.__iter__(self)
     67 def __iter__(self):
---> 68     yield from get_file_binaries_from_pathnames(self.datapipe, self.mode, self.encoding)

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/utils/common.py:86, in get_file_binaries_from_pathnames(pathnames, mode, encoding)
     83 if mode in ('b', 't'):
     84     mode = 'r' + mode
---> 86 for pathname in pathnames:
     87     if not isinstance(pathname, str):
     88         raise TypeError("Expected string type for pathname, but got {}"
     89                         .format(type(pathname)))

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/_typing.py:514, in hook_iterator.<locals>.wrap_generator(*args, **kwargs)
    512         response = gen.send(None)
    513 else:
--> 514     response = gen.send(None)
    516 while True:
    517     request = yield response

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/iter/combining.py:51, in ConcaterIterDataPipe.__iter__(self)
     49 def __iter__(self) -> Iterator:
     50     for dp in self.datapipes:
---> 51         for data in dp:
     52             yield data

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/_typing.py:514, in hook_iterator.<locals>.wrap_generator(*args, **kwargs)
    512         response = gen.send(None)
    513 else:
--> 514     response = gen.send(None)
    516 while True:
    517     request = yield response

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/iter/filelister.py:58, in FileListerIterDataPipe.__iter__(self)
     57 def __iter__(self) -> Iterator[str] :
---> 58     for path in self.datapipe:
     59         yield from get_file_pathnames_from_root(path, self.masks, self.recursive, self.abspath, self.non_deterministic)

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/_typing.py:514, in hook_iterator.<locals>.wrap_generator(*args, **kwargs)
    512         response = gen.send(None)
    513 else:
--> 514     response = gen.send(None)
    516 while True:
    517     request = yield response

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/iter/callable.py:116, in MapperIterDataPipe.__iter__(self)
    114 def __iter__(self) -> Iterator[T_co]:
    115     for data in self.datapipe:
--> 116         yield self._apply_fn(data)

File /user/conda/envs/storch/lib/python3.9/site-packages/torch/utils/data/datapipes/iter/callable.py:81, in MapperIterDataPipe._apply_fn(self, data)
     79 def _apply_fn(self, data):
     80     if self.input_col is None and self.output_col is None:
---> 81         return self.fn(data)
     83     if self.input_col is None:
     84         res = self.fn(data)

File /user/conda/envs/storch/lib/python3.9/site-packages/torchdata/datapipes/iter/util/cacheholder.py:300, in _wait_promise_fn(timeout, filename)
    298     time.sleep(0.01)
    299     if time.time() - start > timeout:
--> 300         raise Exception(
    301             f"OnDiskCache Exception: {filename} expected to be written by different process, "
    302             + f"but file is not ready in {timeout} seconds."
    303         )
    304 return filename

Exception: OnDiskCache Exception: ./data/datasets/IWSLT2016/2016-01/texts/de/en/de-en/train.de-en.de expected to be written by different process, but file is not ready in 300 seconds.
This exception is thrown by __iter__ of MapperIterDataPipe(datapipe=UnBatcherIterDataPipe, fn=functools.partial(<function _wait_promise_fn at 0x7ff9b89dfee0>, 300), input_col=None, output_col=None)

Expected behavior

Dataset is properly unpacked from the downloaded tgz file when using torchtext.datasets.IWSLT*

Environment

Collecting environment information...
PyTorch version: 1.12.1
Is debug build: False
CUDA used to build PyTorch: 11.6
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.1 LTS (x86_64)
GCC version: (Ubuntu 11.2.0-19ubuntu1) 11.2.0
Clang version: Could not collect
CMake version: Could not collect
Libc version: glibc-2.35

Python version: 3.9.12 (main, Jun  1 2022, 11:38:51)  [GCC 7.5.0] (64-bit runtime)
Python platform: Linux-5.15.0-46-generic-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: Could not collect
GPU models and configuration: 
GPU 0: NVIDIA A100-SXM4-80GB
GPU 1: NVIDIA A100-SXM4-80GB
GPU 2: NVIDIA A100-SXM4-80GB
GPU 3: NVIDIA A100-SXM4-80GB

Nvidia driver version: 470.141.03
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

Versions of relevant libraries:
[pip3] mypy==0.971
[pip3] mypy-extensions==0.4.3
[pip3] numpy==1.22.4
[pip3] pytest-mypy==0.9.1
[pip3] pytorch-lightning==1.6.5
[pip3] torch==1.12.1
[pip3] torch-fidelity==0.3.0
[pip3] torchdata==0.4.1
[pip3] torchmetrics==0.9.3
[pip3] torchtext==0.13.1
[pip3] torchvision==0.13.1
[conda] blas                      1.0                         mkl    conda-forge
[conda] cudatoolkit               11.6.0              hecad31d_10    conda-forge
[conda] mkl                       2022.0.1           h06a4308_117  
[conda] numpy                     1.22.4                   pypi_0    pypi
[conda] pytorch                   1.12.1          py3.9_cuda11.6_cudnn8.3.2_0    pytorch
[conda] pytorch-lightning         1.6.5                    pypi_0    pypi
[conda] pytorch-mutex             1.0                        cuda    pytorch
[conda] torch-fidelity            0.3.0                    pypi_0    pypi
[conda] torchdata                 0.4.1                    pypi_0    pypi
[conda] torchmetrics              0.9.3                    pypi_0    pypi
[conda] torchtext                 0.13.1                   pypi_0    pypi
[conda] torchvision               0.13.1                   pypi_0    pypi

[Nayef211]

cc @ejguan and @VitalyFedyunin as this seems to be an issue with the CacheHolder datapipe from torchdata

[ejguan]

Will take a look

[VitalyFedyunin]

Likely related to not yet landed race condition fix

[joecummings]

Likely related to not yet landed race condition fix

@VitalyFedyunin Can you link the Issue/PR?

[VitalyFedyunin]

Yep, In my TODO for 1st half of the week.

[MarkusSpanring]

@VitalyFedyunin is there any update on this?

[VitalyFedyunin]

Working on it right now. Want to make sure it is part of the next official branch cut. So it requires intensive testing.

As mitigation cleaning cache folder and restart should help.

[VitalyFedyunin]

pytorch/data#652 should fix it.

[ejguan]

This is fixed by #1942 and pytorch/data#810