CamemBERT

We explore the impact of the training data size on a French version of RoBERTa. (Equal contribution by the first three authors).

We explore the impact of the training data size and heterogeneity on French language modeling. (Equal contribution by the first three authors).