ReadError: file could not be opened successfully

[annabechang]

Hi,
I tried to load the ImageNetV2Dataset on both linux and colab, both returned the same error:

! pip install git+https://github.com/modestyachts/ImageNetV2_pytorch

from imagenetv2_pytorch import ImageNetV2Dataset
from torch.utils.data import DataLoader

Collecting git+https://github.com/modestyachts/ImageNetV2_pytorch
Cloning https://github.com/modestyachts/ImageNetV2_pytorch to /tmp/pip-req-build-mjyllfja
Running command git clone --filter=blob:none --quiet https://github.com/modestyachts/ImageNetV2_pytorch /tmp/pip-req-build-mjyllfja
Resolved https://github.com/modestyachts/ImageNetV2_pytorch to commit 9017469
Preparing metadata (setup.py) ... done

images = ImageNetV2Dataset(transform=preprocess)

Extracting....
---------------------------------------------------------------------------
ReadError                                 Traceback (most recent call last)
Cell In[19], line 1
----> 1 images = ImageNetV2Dataset(transform=preprocess)

File ~/.conda/envs/py38/lib/python3.8/site-packages/imagenetv2_pytorch/ImageNetV2_dataset.py:82, in ImageNetV2Dataset.__init__(self, variant, transform, location)
     80         assert False, f"Downloading from {URLS[variant]} failed"
     81 print("Extracting....")
---> 82 tarfile.open(self.tar_root).extractall(f"{location}")
     83 shutil.move(f"{location}/{FNAMES[variant]}", self.dataset_root)
     84 self.fnames = list(self.dataset_root.glob("**/*.jpeg"))

File ~/.conda/envs/py38/lib/python3.8/tarfile.py:1608, in TarFile.open(cls, name, mode, fileobj, bufsize, **kwargs)
   1606                 fileobj.seek(saved_pos)
   1607             continue
-> 1608     raise ReadError("file could not be opened successfully")
   1610 elif ":" in mode:
   1611     filemode, comptype = mode.split(":", 1)

ReadError: file could not be opened successfully

[shionhonda]

You can just try re-installing the package. The problem is solved in the current version!