cjvt/GaMS3-12B-Instruct
Full fine-tune (fewer layers) vs google/gemma-3-12b-pt. cjvt/GaMS3-12B-Instruct matches part of google/gemma-3-12b-pt (626 of its 1065 tensors; the rest of google/gemma-3-12b-pt has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 1/1, Norms 211/289, MLP 144/144, Attention 192/192
2 derivatives checked
- mradermacher/GaMS3-12B-Instruct-i1-GGUF: Quantization
- mradermacher/GaMS3-12B-Instruct-GGUF: Quantization