Tokenizer config uses the wrong type which causes breaks with txt-gen-inf

#3
by samulrich1 - opened

Tokenizer type is PreTrainedTokenizerFast but should logically be Qwen2Tokenizer unless there is a reason it is not?

Also missing vocab.json file unless this truly is a completely different tokenizer than the original Qwen 1.5

Qwen org

The next version will fix the issue of inconsistent tokenizer.

huybery changed discussion status to closed

@huybery As in a new model or there will be an update to this model's repo?

Sign up or log in to comment