prithivMLmods/Sombrero-Opus-14B-Sm4

Full fine-tune vs Qwen/Qwen2.5-14B-Instruct-1M. Fine-tune: 100% of the weights changed, across 48 of 48 layers.

Tensors changed: Embeddings 1/1, Norms 97/97, MLP 144/144, Attention 336/336, lm_head 1/1

3 derivatives checked