LiquidAI/LFM2.5-Audio-1.5B-JP

Full fine-tune vs LiquidAI/LFM2.5-Audio-1.5B. Fine-tune: 100% of the weights changed, across 17 of 17 layers.

Tensors changed: other 158/160, Norms 323/323, Embeddings 29/29, MLP 202/202, Attention 217/217

2 derivatives checked