metadata
license: apache-2.0
4-bit OmniQuant quantized version of Mixtral-8x7B-Instruct-v0.1. Note that the embedding and the MoE gate weights are not quantized in this version.
license: apache-2.0
4-bit OmniQuant quantized version of Mixtral-8x7B-Instruct-v0.1. Note that the embedding and the MoE gate weights are not quantized in this version.