mradermacher/Metis-8B-RL-GGUF

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

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

On the Hub · Downloads on Model Pulse