Kernel keeps in running state after running metric = evaluate.load("seqeval")

[pratikchhapolika]

I am trying to load metric = evaluate.load("seqeval")
But it keeps running in GPU jupyter notebook.

transformers==4.21.1
Python 3.7.13
evaluate==0.2.2

[lvwerra]

I just ran the command with the same Python and library versions successfully (here a colab notebook). Are there any outputs in the Jupyter console giving a hint where it is hanging?

[pratikchhapolika]

I just ran the command with the same Python and library versions successfully (here a colab notebook). Are there any outputs in the Jupyter console giving a hint where it is hanging?

@lvwerra

I am in restricted zone. Where there is no internet. My guess is metric = evaluate.load("seqeval") is downloading something externally.

So I tried to download folder: huggingface/datasets/Metrics/seqeva/ put it in my local folder. and then run:
metric = evaluate.load("seqeval")

But no luck. It throws me

TypeError: 'NoneType' object is not callable

[lvwerra]

Indeed, metrics are loaded from hf.co. E.g. seqeval is loaded from here: https://huggingface.co/spaces/evaluate-metric/seqeval/tree/main

If you don't have any internet connection you can download the folder (in this case also just the seqeval.py script) into a seqeval/ folder. Then you should be able to load it with evaluate.load("./seqeval/"). The issue with your approach is that you tried to load an datasets metric instead of the evaluate metric.

Does that work?

[pratikchhapolika]

Yayyyy!!! worked. Thank you so much @lvwerra