Interp.plot_top_losses TypeError when fastai installed through setuptools

fastai
1

Hello

I wanted to try to upload fastai as a pypi package. So I installed fastai through setup.py with the version that was working for me:

from setuptools import setup

setup(
    name='challenge',
    version='0.1.0',
    description='Python package challenge',
    url='https://github.com/bla/challenge',
    packages=['challenge'],
    install_requires=['torch',
                      'torchvision',
                      'fastai==2.1.5',
                      'fastcore==1.3.2'           
                      ],
    )

I am running the pet breed lesson, when I run:

interp.plot_top_losses(9, figsize=(15,10))

I got the error :

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-19-be969192d4f7> in <module>
----> 1 interp.plot_top_losses(9, figsize=(15,10))

~/.virtualenvs/challenge/lib/python3.6/site-packages/fastai/interpret.py in plot_top_losses(self, k, largest, **kwargs)
     38         if isinstance(self.inputs[0], Tensor): inps = tuple(o[idx] for o in self.inputs)
     39         else: inps = self.dl.create_batch(self.dl.before_batch([tuple(o[i] for o in self.inputs) for i in idx]))
---> 40         b = inps + tuple(o[idx] for o in (self.targs if is_listy(self.targs) else (self.targs,)))
     41         x,y,its = self.dl._pre_show_batch(b, max_n=k)
     42         b_out = inps + tuple(o[idx] for o in (self.decoded if is_listy(self.decoded) else (self.decoded,)))

~/.virtualenvs/challenge/lib/python3.6/site-packages/fastai/interpret.py in <genexpr>(.0)
     38         if isinstance(self.inputs[0], Tensor): inps = tuple(o[idx] for o in self.inputs)
     39         else: inps = self.dl.create_batch(self.dl.before_batch([tuple(o[i] for o in self.inputs) for i in idx]))
---> 40         b = inps + tuple(o[idx] for o in (self.targs if is_listy(self.targs) else (self.targs,)))
     41         x,y,its = self.dl._pre_show_batch(b, max_n=k)
     42         b_out = inps + tuple(o[idx] for o in (self.decoded if is_listy(self.decoded) else (self.decoded,)))

~/.virtualenvs/challenge/lib/python3.6/site-packages/fastai/torch_core.py in __torch_function__(self, func, types, args, kwargs)
    315 
    316     def __torch_function__(self, func, types, args=(), kwargs=None):
--> 317         with torch._C.DisableTorchFunction(): ret = _convert(func(*args, **(kwargs or {})), self.__class__)
    318         if isinstance(ret, TensorBase): ret.set_meta(self, as_copy=True)
    319         return ret

TypeError: only integer tensors of a single element can be converted to an index

which is odd since this worked perfectly well with my other fastai install using exactly the same data and code. Anyone has this problem before?