bgg1996/Melinoe-Qwen3-5-27B-VL
Full fine-tune vs Qwen/Qwen3.5-27B. Fine-tune: 94% of the weights changed, across 64 of 64 layers.
Tensors changed: lm_head 1/1, Embeddings 0/4, MLP 192/304, Attention 352/556, Norms 104/319
3 derivatives checked
- mradermacher/Melinoe-Qwen3-5-27B-VL-i1-GGUF: Quantization
- mradermacher/Melinoe-Qwen3-5-27B-VL-GGUF: Quantization
- bgg1996/Melinoe-Qwen3-5-27B-VL-Q4_K_M-GGUF: Not comparable yet