OpenGVLab/InternVL2_5-26B-MPO
Full fine-tune vs OpenGVLab/InternVL2_5-26B. Fine-tune: 98% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 3/5, Norms 143/277, Attention 231/231, other 89/91, MLP 329/330
Full fine-tune vs OpenGVLab/InternVL2_5-26B. Fine-tune: 98% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 3/5, Norms 143/277, Attention 231/231, other 89/91, MLP 329/330