Make model compatible with Transformers v5
#34
by hmellor HF Staff - opened
bytes_to_unicode should be imported from convert_slow_tokenizer, not from GPT2.
I'd recommend making the same change in your other models which do this.
courage17340 changed pull request status to merged