mudler/ced-gguf
Same weights, converted vs mispeech/ced-base. Same weights stored in another dtype (F32/I64 → F16/F32), plus 2 new tensors.
Tensors changed: Attention 0/48, MLP 0/48, Norms 0/50, Embeddings 0/4, other 0/8
Same weights, converted vs mispeech/ced-base. Same weights stored in another dtype (F32/I64 → F16/F32), plus 2 new tensors.
Tensors changed: Attention 0/48, MLP 0/48, Norms 0/50, Embeddings 0/4, other 0/8