Question a bout model tokenizer

#2
by Bachhoang - opened

When using the GGUF model and checking the vocabulary metadata, I noticed a slight difference compared to the base model's vocabulary:

The GGUF model does not have a padding token, as shown below:

image.png

However, the base model includes a padding token in its vocabulary:

image.png

Could someone explain why this difference exists and how I can handle like it?

Sign up or log in to comment