Omartificial-Intelligence-Space/Arabic-base-all-nli-stsb-quora
Partial fine-tune vs google-bert/bert-base-multilingual-cased. Partial change: 48% of the weights changed (attention, embeddings, other), in 12 of 12 layers.
Tensors changed: Embeddings 2/3, Attention 96/96, other 48/50