Fine-tune PP-OCRv4 DET Server Model

[lsch0lz]

Hi! I want to fine-tune the server detection model. Am I right assuming, that this is simply only the teacher model, without any students? So training with python3 tools/train.py -c configs/det/ch_PP-OCRv4/ch_PP-OCRv4_det_teacher.yml should fine-tune the from you provided server model? Do I have to use the pre-trained weights LINK anyhow? I see in the config, that there is a different model being used as pretrained_model.

Thanks!