If I wanted to use my own dataset (l large medical corpus) as the pretrained model and then finetune with my task specific data what would be the best way to go about this? I also want to create a smaller network (1 layer and 200 hidden activation units).
Has anyone done this with fast.ai v1?
Thanks