Ngixdev/Qwen3.5-4B-OBLITERATED
Partial fine-tune (fewer layers) vs Qwen/Qwen3.5-4B. Ngixdev/Qwen3.5-4B-OBLITERATED matches part of Qwen/Qwen3.5-4B (426 of its 738 tensors; the rest of Qwen/Qwen3.5-4B has no counterpart). On the shared part: Partial change: 89% of the weights changed (attention, embeddings, mlp), in 30 of 32 layers.
Tensors changed: Embeddings 1/1, MLP 90/96, Attention 76/224, Norms 0/105
2 derivatives checked
- mradermacher/Qwen3.5-4B-OBLITERATED-i1-GGUF: Quantization
- mradermacher/Qwen3.5-4B-OBLITERATED-GGUF: Same weights, converted