CNR-ILC/gs-GreBerta
Partial fine-tune vs bowphs/GreBerta. Partial change: 68% of the weights changed (attention, lm_head, norms, other), in 12 of 12 layers.
Tensors changed: lm_head 5/5, Embeddings 0/5, Norms 48/48, Attention 96/96, other 48/48
Partial fine-tune vs bowphs/GreBerta. Partial change: 68% of the weights changed (attention, lm_head, norms, other), in 12 of 12 layers.
Tensors changed: lm_head 5/5, Embeddings 0/5, Norms 48/48, Attention 96/96, other 48/48