bug ! NO consistent EOS token #14
Description
Why is VLLM's inference script use eos token as , TransFormers's inference script use tokenizer.eos(im_end), and why is it necessary to output tokens one by one.....
{{ message }}
Why is VLLM's inference script use eos token as , TransFormers's inference script use tokenizer.eos(im_end), and why is it necessary to output tokens one by one.....