neta-art/neta-lumina-gguf

Same weights, converted (fewer layers) vs neta-art/Neta-Lumina. neta-art/neta-lumina-gguf matches part of neta-art/Neta-Lumina (400 of its 934 tensors; the rest of neta-art/Neta-Lumina has no counterpart). On the shared part: Same weights stored in another dtype (BF16/F16 → BF16/F32).

Tensors changed: Embeddings 0/9, Norms 0/181, Attention 0/60, MLP 0/90, other 0/60

On the Hub · Downloads on Model Pulse