prithivMLmods/Alpha-UMa-Qwen-14B

Full fine-tune vs prithivMLmods/Galactic-Qwen-14B-Exp2. Fine-tune: 100% of the weights changed, across 48 of 48 layers.

Tensors changed: Norms 97/97, MLP 144/144, Attention 336/336

2 derivatives checked