ToastyPigeon/Gemma-3-Starshine-12B
Full fine-tune (fewer layers) vs ToastyPigeon/Gemma-3-Confetti-12B. ToastyPigeon/Gemma-3-Confetti-12B matches part of ToastyPigeon/Gemma-3-Starshine-12B (626 of its 1065 tensors; the rest of ToastyPigeon/Gemma-3-Starshine-12B 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 289/289, MLP 144/144, Attention 192/192
5 derivatives checked
- bartowski/ToastyPigeon_Gemma-3-Starshine-12B-GGUF: Same weights, converted (fewer layers)
- mradermacher/Gemma-3-Starshine-12B-i1-GGUF: Quantization (fewer layers)
- mradermacher/Gemma-3-Starshine-12B-GGUF: Quantization (fewer layers)
- ToastyPigeon/gemma-3-starshine-merge-B: Full fine-tune
- ToastyPigeon/gemma-3-starshine-merge-A: LoRA-style