OBLITERATUS/Qwen3-4B-OBLITERATED
Full fine-tune vs Qwen/Qwen3-4B. Fine-tune: 95% of the weights changed, across 34 of 36 layers.
Tensors changed: Embeddings 1/1, Norms 0/145, MLP 102/108, Attention 136/144
4 derivatives checked
- mradermacher/Qwen3-4B-OBLITERATED-GGUF: Full fine-tune
- mradermacher/Qwen3-4B-OBLITERATED-i1-GGUF: Full fine-tune
- Hatsunama/Qwen3-4B-OBLITERATED-Q4_K_S: Full fine-tune
- Hatsunama/Qwen3-4B-OBLITERATED-repaired: Full fine-tune