How to fine tune a model using two different datasets with different number of class.

[lxrswdd]

So currently I trained a 7-class (7 emotions) repvgg model using dataset A.
I wish to further fine-tune the model to fit a 2-class (gaze engaged and gaze not engaged) dataset.
The classes are different for the two datasets but present similarities in the features.

I tried to resume training from the checkpoint saved for 7-class model. I changed the raf-db-basic_bs32_subfolder.py file to fit the second dataset. I ran the following codes:

python tools/train.py /home/guest/Xiangrui/mmpretrain/mmpretrain/configs/repvgg/repvgg-A1_8xb32_in1k_lessaug.py --cfg-options /home/guest/Xiangrui/mmpretrain/mmpretrain/work_dirs/repvgg-A1_8xb32_in1k_lessaug/last_checkpoint

This will result in ValueError: not enough values to unpack (expected 2, got 1)

May I know whats the correct way to do so?