gbrixi/minerva-mlm-8k
Full fine-tune vs gbrixi/minerva-mlm. Fine-tune: 100% of the weights changed, across 33 of 33 layers.
Tensors changed: other 12/19, lm_head 2/2, Attention 66/66, Norms 62/66, MLP 99/99, Embeddings 1/1
Full fine-tune vs gbrixi/minerva-mlm. Fine-tune: 100% of the weights changed, across 33 of 33 layers.
Tensors changed: other 12/19, lm_head 2/2, Attention 66/66, Norms 62/66, MLP 99/99, Embeddings 1/1