mradermacher/TaRO-8B-GGUF

Same weights, converted (fewer layers) vs zhengmh/TaRO-8B. mradermacher/TaRO-8B-GGUF matches part of zhengmh/TaRO-8B (399 of its 750 tensors; the rest of zhengmh/TaRO-8B has no counterpart). On the shared part: Same weights stored in another dtype (BF16 → F16/F32).

Tensors changed: Embeddings 0/1, MLP 0/108, Norms 0/145, Attention 0/144, lm_head 0/1

On the Hub · Downloads on Model Pulse