I am trying to retrain a pretrained model (
pytorch/fairseq/xlmr.base) loaded from the hub. The examples provided on the documentation focus on fine-tuning and evaluation. What if I want to retrain a pretrained model on additional data? Is there an interface that allows this?
I should specify that the architecture mentioned above is not available in fairseq’s registered architectures and AFAIK, it is not possible to use the utility
fairseq-train with a pretrained model from the hub.
Thanks a lot