Continuously retrain ULMFiT

Hi all,

I’m working on a classification task using ULMFiT. I’m iteratively adding to my training set, and everytime I train a new ULMFiT model. But since everytime 80% of the dataset is the same to the one used previously, I’m wondering if there is a way to load in my trained model and just continue training on it. That would save me a lot of time. I tried loading in my trained model the way you initially have to load the pretrained model from ULMFiT, but that doesn’t work.

If there are any resources about this question, feel free to share them with me.

Thanks in advance,