mgoin
/

Mixtral-8x7B-Instruct-v0.1-FP8

Model card Files Files and versions Community

File size: 156 Bytes

cd23aab

DEFAULT_stage:
  DEFAULT_modifiers:
    QuantizationModifier:
      ignore: [lm_head, 're:.*block_sparse_moe.gate']
      targets: Linear
      scheme: FP8