0

I'm working in a notebook for fastai that previously worked correctly but now is generating an error.

I think the error is due to this newly added labeling function, RegexLabeller:

https://docs.fast.ai/data.transforms.html#regexlabeller

The cells in my notebook prior to the error are:

from fastai.vision.all import *
import timm
path = untar_data(URLs.PETS)/'images'

My third cell is

dls = ImageDataLoaders.from_name_func('.',
    get_image_files(path), valid_pct=0.2, seed=42,
    label_func=RegexLabeller(pat = r'^([^/]+)_\\d+'),
    item_tfms=Resize(224))

... which throws a long assertion error starting with

AssertionError                            Traceback (most recent call last)
Input In [5], in <cell line: 1>()
----> 1 dls = ImageDataLoaders.from_name_func('.',
      2     get_image_files(path), valid_pct=0.2, seed=42,
      3     label_func=RegexLabeller(pat = r'^([^/]+)_\\d+'),
      4     item_tfms=Resize(224))

File /usr/local/lib/python3.9/dist-packages/fastai/vision/data.py:149, in ImageDataLoaders.from_name_func(cls, path, fnames, label_func, **kwargs)
    147     raise ValueError("label_func couldn't be lambda function on Windows")
    148 f = using_attr(label_func, 'name')
--> 149 return cls.from_path_func(path, fnames, f, **kwargs)

... and ending with

AssertionError: Failed to find "re.compile('^([^/]+)_\\\\d+')" in "pug_130.jpg"

It was the RegexLabeller that changed in what was previously working code:

label_func=RegexLabeller(pat = r'^([^/]+)_\\d+'),

Any feedback appreciated!

  • I resolved this question. Something was conflicting with the Paperspace Gradient notebook. It worked fine in a Google Colab Jupyter notebook. – brooksjordan Feb 20 '23 at 22:52

0 Answers0