Enigma-for-AI/bert-base-sudabert
Full fine-tune vs asafaya/bert-base-arabic. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 4/4, Norms 48/48, Attention 96/96, other 50/50
Full fine-tune vs asafaya/bert-base-arabic. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 4/4, Norms 48/48, Attention 96/96, other 50/50