SC117/QwenPaw-Flash-9B-heretic
Partial fine-tune vs Qwen/Qwen3.5-9B. Partial change: 82% of the weights changed (attention, lm_head, mlp, norms), in 32 of 32 layers.
Tensors changed: lm_head 1/1, Embeddings 0/5, MLP 96/211, Attention 187/336, other 0/1, Norms 102/221
5 derivatives checked
- mradermacher/QwenPaw-Flash-9B-heretic-i1-GGUF: Quantization
- mradermacher/QwenPaw-Flash-9B-heretic-GGUF: Same weights, converted
- SC117/QwenPaw-Flash-9B-heretic-mlx-4Bit: Quantization
- SC117/QwenPaw-Flash-9B-heretic-mlx-8Bit: Quantization
- SC117/QwenPaw-Flash-9B-heretic-mlx-6Bit: Quantization