zelk12/MT-Gen1-gemma-3-12B

Full fine-tune vs IlyaGusev/saiga_gemma3_12b. Fine-tune: 97% of the weights changed, across 48 of 48 layers.

Tensors changed: Embeddings 1/4, Norms 270/400, MLP 144/252, Attention 192/408, other 1/1

2 derivatives checked

• mradermacher/MT-Gen1-gemma-3-12B-i1-GGUF: Quantization (fewer layers)
• mradermacher/MT-Gen1-gemma-3-12B-GGUF: Quantization (fewer layers)

On the Hub · Downloads on Model Pulse