aisingapore/Gemma-SEA-LION-v4-27B
Partial fine-tune vs google/gemma-3-27b-it. Partial change: 35% of the weights changed (attention, mlp, norms), in 59 of 62 layers.
Tensors changed: Embeddings 0/4, Norms 29/484, MLP 70/294, Attention 88/464, other 0/1
3 derivatives checked
- aisingapore/Gemma-SEA-LION-v4-27B-IT: Full fine-tune
- mradermacher/Gemma-SEA-LION-v4-27B-i1-GGUF: Quantization (fewer layers)
- mradermacher/Gemma-SEA-LION-v4-27B-GGUF: Quantization (fewer layers)