Finetuned T5 on the french part of Lang-8 to automatically correct sentences.

Since the Lang-8 dataset contains really short sentences, the model does not generalize well with sentences larger than 10 words. I'll upload soon the cleaned dataset that I've used for training.