surogate/amami-357m-ro
Full fine-tune vs nvidia/magpie_tts_multilingual_357m. surogate/amami-357m-ro is a GGUF (F32/I32) of different weights. Fine-tune: 100% of the weights changed, across 12 of 12 layers.
Tensors changed: Embeddings 20/22, Norms 66/66, Attention 76/76, other 98/98