utter-project/TowerVision-2B

Full fine-tune (fewer layers) vs Unbabel/Tower-Plus-2B. Unbabel/Tower-Plus-2B matches part of utter-project/TowerVision-2B (287 of its 714 tensors; the rest of utter-project/TowerVision-2B has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 26 of 26 layers.

Tensors changed: Norms 105/105, MLP 78/78, Attention 104/104

1 derivatives checked

On the Hub · Downloads on Model Pulse