inclusionAI/Ring-lite-distill-preview

Full fine-tune vs inclusionAI/Ling-lite. Fine-tune: 100% of the weights changed, across 28 of 28 layers.

Tensors changed: Embeddings 1/1, Norms 57/57, Attention 56/56, MLP 2884/2884, lm_head 1/1

2 derivatives checked