CodeGoat24/UnifiedReward-Flex-qwen35-9b
Partial fine-tune vs CodeGoat24/UnifiedReward-Think-qwen35-9b. Partial change: 84% of the weights changed (attention, lm_head, mlp, norms), in 32 of 32 layers.
Tensors changed: lm_head 1/1, Embeddings 0/4, Norms 77/215, Attention 190/332, MLP 96/208
2 derivatives checked
- mradermacher/UnifiedReward-Flex-qwen35-9b-i1-GGUF: Full fine-tune
- mradermacher/UnifiedReward-Flex-qwen35-9b-GGUF: Partial fine-tune