Couldn't reach server GPT-2

[MSMOON]

I have tried to use gpt2 using ubuntu and vagrant. This is the code:
import torch from lm_scorer.models.auto import AutoLMScorer as LMScorer scorer = LMScorer.from_pretrained("gpt2")
I get this error:

AH01215: OSError: Couldn't reach server at 'https://s3.amazonaws.com/models.huggingface.co/bert/gpt2-config.json' to download pretrained model configuration file.

It has worked before but I had to reset my virtual environment and now it no longer worked. I think it has something to do with apache configurations.

Also, It works in the terminal but not in python script.

[MSMOON]

I also created this stack overflow question here: https://stackoverflow.com/questions/61944526/oserror-couldnt-reach-server-gpt2-config-json

[BramVanroy]

It is odd that it works from CLI but not from within your script. That does not make a lot of sense. Can you try this?

scorer = LMScorer.from_pretrained("gpt2", force_download=True)

[MSMOON]

Thanks, Bram. I tried that code but I get the same error. I believe it has something to do with 1) apache2 or ubuntu config and what I've allowed it to connect to or 2) some download I am missing because it has worked previously.

I tried downloading the gpt2 config, model.bin, and vocab file but I would either get the same error as above or get this error:

ValueError: Unrecognized model name.Can be one of: gpt2, gpt2-medium, gpt2-large, gpt2-xl, distilgpt2:

[BramVanroy]

If you are sure that you have manually downloaded all files to the correct folder, you can disable the online look-up. This is useful if you have, as you say, network restrictions.

scorer = LMScorer.from_pretrained("gpt2", local_files_only=True)

[MSMOON]

Still nothing yet.

1. I downloaded everything from https://huggingface.co/gpt2#list-files
2. Added files to the same directory the script is in
3. changed names (e.g. gpt2-config.json to config.json)

Is there anything I am missing?

[BramVanroy]

This is likely to be a problem with the LMScorer rather than with this transformers library. Looking t the source code, it does not pass they keyword arguments down to model init. I suggest that you make an issue over at the library that you used.

https://github.com/simonepri/lm-scorer/blob/master/lm_scorer/models/gpt2.py

[MSMOON]

Still nothing. I believe it is my apache2 configurations for access but I haven't figure out how yet.

[BramVanroy]

Closing this. See continuation here: simonepri/lm-scorer#8 (comment)