Hatsunama/Qwen3-4B-OBLITERATED-repaired

Full fine-tune vs OBLITERATUS/Qwen3-4B-OBLITERATED. 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

3 derivatives checked

• mradermacher/Qwen3-4B-OBLITERATED-repaired-i1-GGUF: Quantization
• mradermacher/Qwen3-4B-OBLITERATED-repaired-GGUF: Same weights, converted
• Hatsunama/Qwen3-4B-OBLITERATED-toolcall-lora: Adapter

On the Hub · Downloads on Model Pulse