Shiba-2-shiba/HY-OmniWeaving_HunyuanVideo_1.5_FP8_Patched

Full fine-tune vs tencent/HY-OmniWeaving. Fine-tune: 98% of the weights changed, across 0 of 2 layers.

Tensors changed: MLP 216/448, other 110/240, Norms 0/226, Embeddings 0/11, Attention 432/872

On the Hub · Downloads on Model Pulse