OS-Copilot/OS-Atlas-Base-4B
Full fine-tune vs OpenGVLab/InternVL2-4B. Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: Norms 161/161, MLP 166/166, Attention 160/160, Embeddings 4/4, other 48/48
Full fine-tune vs OpenGVLab/InternVL2-4B. Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: Norms 161/161, MLP 166/166, Attention 160/160, Embeddings 4/4, other 48/48