Jiunsong/supergemma4-e4b-abliterated
Partial fine-tune (fewer layers) vs google/gemma-4-E4B-it. Jiunsong/supergemma4-e4b-abliterated matches part of google/gemma-4-E4B-it (719 of its 2130 tensors; the rest of google/gemma-4-E4B-it has no counterpart). On the shared part: Partial change: 17% of the weights changed (attention, mlp, other), in 35 of 42 layers.
Tensors changed: Embeddings 0/2, Norms 0/296, other 4/127, MLP 39/126, Attention 43/168
9 derivatives checked
- NidAll/supergemma4-e4b-abliterated-Q4_K_M-GGUF: Quantization
- mradermacher/supergemma4-e4b-abliterated-GGUF: Same weights, converted
- mradermacher/supergemma4-e4b-abliterated-i1-GGUF: Quantization
- Jiunsong/supergemma4-e4b-abliterated-mlx: Quantization
- Abiray/supergemma4-e4b-abliterated-GGUF: Quantization
- kanarena/supergemma4-e4b-abliterated-multimodal-gguf-4bit: Quantization
- DreamFoundries/supergemma4-e4b-abliterated-6bit: Quantization
- DreamFoundries/supergemma4-e4b-abliterated-4bit: Quantization
- DreamFoundries/supergemma4-e4b-abliterated-8bit: Quantization
1 other repos hold byte-for-byte identical weight files.