snuh/hari-q2.5-thinking

Full fine-tune vs Qwen/Qwen2.5-72B. Fine-tune: 98% of the weights changed, across 80 of 80 layers.

Tensors changed: Embeddings 0/1, MLP 240/240, Attention 560/560, Norms 145/161, lm_head 1/1

2 derivatives checked