Darknet: (Question) Fine Tuning to Transfer Learning Transition

[silvernine209]

I was wondering if I can use my fine tuning weights yolo-obj_fine_tuning_50300.weights (still ongoing and training) and switch it to transfer learning by removing stopbackward=1 from my .cfg file.
I understand it's the best to start scratch again for transfer learning, but I already put in days to train my weights up to 50300 iteration.

Below is iteration vs mAP for my 1.6 million train images and 500 classes fine tuning learning.

[AlexeyAB]

Yes, you can try to remove stopbackward=1 and continue training.

---

You must train from the begining only if you changed number of: classes, anchors, filters, layers, or type of layers.

[silvernine209]

Thank you for the clarification!