twinkle-ai/gemma-3-4B-T1-it
Full fine-tune (fewer layers) vs google/gemma-3-4b-pt. twinkle-ai/gemma-3-4B-T1-it matches part of google/gemma-3-4b-pt (444 of its 883 tensors; the rest of google/gemma-3-4b-pt has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 34 of 34 layers.
Tensors changed: Embeddings 1/1, Norms 205/205, MLP 102/102, Attention 136/136
1 derivatives checked
- twinkle-ai/gemma-3-4B-T1-it-GGUF: Same weights, converted