Make model compatible with Transformers v5

`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.

tokenization_moonshot.py

@@ -17,7 +17,7 @@ from tiktoken.load import load_tiktoken_bpe
 from tokenizers import AddedToken
 from transformers.tokenization_utils import PreTrainedTokenizer
 from transformers.utils import to_py_obj
-from transformers.models.gpt2.tokenization_gpt2 import bytes_to_unicode
+from transformers.convert_slow_tokenizer import bytes_to_unicode
 
 
 logger = getLogger(__name__)