Setup questions

Use this topic for any questions related to Chapter 0 of the course.

Hi @sgugger, thank you for the course. Is it advisable to use a dockerized container as a coding environment on a windows machine? Or inside a WSL2 layer? Or does something or the other may create issues later? (Asking because I have faced a lot of issues with NCCL failure while trying to run XGboost et al on a RAPIDS installation on WSL2).

I haven’t personally tries on WSL2 but it should work pretty well as long as your GPU is recognized.


Hi, @sgugger!
In order to go through the first code snippet like

from transformers import pipeline
classifier = pipeline("sentiment-analysis")
classifier("I've been waiting for a HuggingFace course my whole life.")

I needed to do the next things after activating the virtual environment:

pip install ipykernel
python -m ipykernel install --user --name=.env

After that I was able to choose needed virtual environment in jupyter notebook and run the code without errors.
Otherwise I’ve had only Python 3 kernel in jupyter notebook and got messages like “ModuleNotFoundError: No module named ‘transformers’” even though I went through all the setup steps and which python and pip list showed expected results (for the first I saw path to the venv python and for second I saw transformers library in the list).

Is it just me or setup chapter doesn’t cover all the steps?

P.S. I’m using Windows 10 and Python 3.8.10

The setup does not cover locally installing jupyter notebook, either you are experienced enough to do it yourself, or we recommend using colab. That’s why those instructions are not included :slight_smile:


When I try running the command “from transformers import pipeline
section 3 of chapter 1 of the course in colab, I am getting the following error:

OSError                                   Traceback (most recent call last)
<ipython-input-2-69a9fd07ccac> in <module>()
----> 1 from transformers import pipeline

20 frames
/usr/lib/python3.7/ctypes/ in __init__(self, name, mode, handle, use_errno, use_last_error)
    363         if handle is None:
--> 364             self._handle = _dlopen(self._name, mode)
    365         else:
    366             self._handle = handle

OSError: /usr/local/lib/python3.7/dist-packages/torchaudio/ undefined symbol: _ZN2at6detail10noopDeleteEPv

Any idea how to fix it?

It comes from torchaudio, for some reason. I’ve removed that from the deps installed so if you retry to open the notebook on Colab, it should work now.

1 Like

It is working now. Thank you.

1 Like

Thank you for creating this course! :blush:

I am going through the setup doc where it asks to install development version by running the following command.

!pip install transformers[sentencepiece]

Could you explain why we are passing sentencepiece as an extra instead of passing “dev” like below

!pip install transformers[dev]

The dev install was broken on Colab for a few days, because there were incompatible versions of torch and torchaudio, so we switch to something lighter.

1 Like