HazyResearch/domino

imagenette is not being installed ? Error Below

Alla-Abdella opened this issue · 3 comments

import os

dp = mk.datasets.get("imagenette")

we'll only be using the validation data

dp = dp.lz[dp["split"] == "valid"]


Downloading: 100%
1.56G/1.56G [00:42<00:00, 40.5MB/s]

AttributeError Traceback (most recent call last)
in ()
1 import os
2
----> 3 dp = mk.datasets.get("imagenette")
4
5 # we'll only be using the validation data

4 frames
/usr/local/lib/python3.7/dist-packages/meerkat/datasets/utils.py in extract(path, dst, extractor)
22 return extractor.extract(path, dst)
23 for extractor in Extractor.extractors:
---> 24 if extractor.is_extractable(path):
25 return extractor.extract(path, dst)
26 raise ValueError("Extraction method not found for {}".format(path))

AttributeError: 'str' object has no attribute 'is_extractable'

This is a bug introduced in meerkat==0.2.5. I have submitted a patch to HazyResearch/meerkat#260

If you force install of meerkat<0.2.5 the snippet will work

Not a developer, but I found a workaround for this issue if you run the notebook locally:

  1. Download imagenette from https://github.com/fastai/imagenette
  2. Put the downloaded imaginette folder in ~/.meerkat/datasets/imagenette/full/ (make sure the folder name is imaginette2)
  3. You should now be able to run dp = mk.datasets.get("imagenette")

This should be fixed in the most recent update -- let me know if the problem persists