mradermacher/Hi-Token-GGUF

Same weights, converted (fewer layers) vs xyzzzh/Hi-Token. mradermacher/Hi-Token-GGUF matches part of xyzzzh/Hi-Token (435 of its 825 tensors; the rest of xyzzzh/Hi-Token has no counterpart). On the shared part: Same weights stored in another dtype (BF16 → F16/F32).

Tensors changed: lm_head 0/1, Embeddings 0/1, MLP 0/108, Attention 0/252, Norms 0/73

On the Hub · Downloads on Model Pulse