JoaoZaokk/Wan2.2-TI2V-5B-W4A8-ConvRot
Same weights, converted vs Wan-AI/Wan2.2-TI2V-5B. Same weights stored in another dtype (F32 → I8/U8), plus 900 new tensors.
Tensors changed: Attention 0/240, Norms 0/180, MLP 0/60, other 0/35, Embeddings 0/10
Same weights, converted vs Wan-AI/Wan2.2-TI2V-5B. Same weights stored in another dtype (F32 → I8/U8), plus 900 new tensors.
Tensors changed: Attention 0/240, Norms 0/180, MLP 0/60, other 0/35, Embeddings 0/10