Using Python 3.8.3, fastai v2.1.10, and image_tabular v0.0.6 on Windows 10:

The following imports return an error:

import fastai
import image_tabular
from fastai.vision import *
from fastai.tabular import *
from image_tabular.core import *

The error is:

Traceback (most recent call last):

  File "<ipython-input-6-3fb4e005a29c>", line 7, in <module>
    from image_tabular.core import *

  File "C:\Users\username\Anaconda3\envs\fastai_env\lib\site-packages\image_tabular\core.py", line 19, in <module>
    def _normalize_batch_image_tab(b:Tuple[Tensor,Tensor],

NameError: name 'Tuple' is not defined

After review, it seems like this may be a forward reference issue. I tried adding from __future__ import annotations above import fastai because my python version is >3.7, but I still get the same error.

Similarly, from image_tabular.dataset import * yields NameError: name 'Dataset' is not defined,
from image_tabular.model import * returns NameError: name 'Module' is not defined, and
from image_tabular.metric import * returns NameError: name 'Tensor' is not defined.

Any advice on how I may fix these issues and use this neat looking tool?

Thanks!

I was trying to use the image_tabular package in the Kaggle kernel but while importing the libraries I received this error :

i was tried using multiclass and got some error

find learning rate to train the last layer group first

fucntion when error :
learn.freeze()
learn.lr_find()
learn.recorder.plot()

error message : can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.

i search in stackoverflow too but still got error , i read some link and its said , cause i use multiclass , and your example in binary

what should i do ?