lapa-llm/lapa-12b-pt
Full fine-tune vs google/gemma-3-12b-pt. Fine-tune: 100% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 3/4, Norms 242/400, MLP 198/252, Attention 300/408, other 1/1
2 derivatives checked
- lapa-llm/lapa-v0.1.2-instruct: Full fine-tune
- lapa-llm/lapa-v0.1.3-instruct: Full fine-tune