legraphista/dolphin-2.9.1-mixtral-1x22b-IMat-GGUF

Same weights, converted vs dphn/dolphin-2.9.1-mixtral-1x22b. Same weights stored in another dtype (BF16 → F16/F32), plus 168 new tensors.

Tensors changed: Embeddings 0/1, Norms 0/113, Attention 0/224, lm_head 0/1

On the Hub · Downloads on Model Pulse