Multiple training one after another

[MITsVision]

I have limited knowledge in this domain and here trying to learn so that I can improve my self, forgive me if I am asking a junky question [But i really need advice].

I have trained a model in the past for 6 labels (this was used : tiny-yolo-voc-6c.cfg) and that is working perfectly. now I want to train a model for 18 labels(6 are the same as of the previous model) for that I have created the cfg file: tiny-yolo-voc-18c.cfg.

i want to train this model on the top of the previous model and for that, i am using the command flow --model cfg/tiny-yolo-voc-18c.cfg --load -1 --train --annotation train/annotation --dataset train/input_image where --load -1 ,i am loading the last check point from previous model.

now my problem is that

1. if i am going to use tiny-yolo-voc-18c.cfg then the checkpoint will be missing because all the checkpoints are in the format tiny-yolo-voc-6c-xx.xx so simply i can not use this approach.
2. if i am going the edit the previous tiny-yolo-voc-6c.cfg then it will work perfectly. but here i have lack of knowledge, is it the right approach or not

please guide me which approach to choose or is there any better way to proceed.

Note: only thing i want is to train current model on the top of the previously trained model.

Thanks in advance.

[yasser-taha]

Hey @savanmorya why don't you just load the "tiny-yolo-voc-6c.cfg" as follows: flow --model cfg/tiny-yolo-voc-18c.cfg --load tiny-yolo-voc-6c.cfg --train --annotation train/annotation --dataset train/input_image