alphaZimuth/Hy-MT2-1.8B-SpecDraft-30B
Same weights, converted vs tencent/Hy-MT2-1.8B. Same weights stored in another dtype (BF16 → F16/F32).
Tensors changed: Norms 0/129, MLP 0/96, Attention 0/128
Same weights, converted vs tencent/Hy-MT2-1.8B. Same weights stored in another dtype (BF16 → F16/F32).
Tensors changed: Norms 0/129, MLP 0/96, Attention 0/128