Clarification on dicksonhk/RolmOCR-mlx-4Bit training origin

#1
by dqdw - opened

Dear [Developer/Team],

I really appreciate your work on dicksonhk/RolmOCR-mlx-4Bit. I've been using it recently and it's been very helpful.

Before building on top of it, I would like to understand its connection with reducto/RolmOCR:

Direct Fine-tuning: Is dicksonhk/RolmOCR-mlx-4Bit a direct fine-tuned version of reducto/RolmOCR, or were there intermediate models/checkpoints involved?

Inheritance: Are the weights directly inherited from reducto/RolmOCR?

This clarification would help me avoid compatibility issues.

I really appreciate your help.

Sign up or log in to comment