Nitral-Archive/Irix-12B-Qwen3-Tokenizer-Homunculus-Slerp

Full fine-tune vs arcee-ai/Homunculus. Fine-tune: 97% of the weights changed, across 40 of 40 layers.

Tensors changed: Norms 61/81, MLP 117/120, Attention 156/160

1 derivatives checked

On the Hub · Downloads on Model Pulse