pymlex/roberta-spanish-cefr
Full fine-tune vs bertin-project/bertin-roberta-base-spanish. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: other 50/50, Embeddings 3/3, Attention 96/96
Full fine-tune vs bertin-project/bertin-roberta-base-spanish. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: other 50/50, Embeddings 3/3, Attention 96/96