nejumi
/

Qwen2.5-14B-Instruct-1M-W8A8-calib-ja-1k

8-bit precision

compressed-tensors

Model card Files Files and versions Community

Qwen2.5-14B-Instruct-1M-W8A8-calib-ja-1k

File size: 170 Bytes

e828c5a

DEFAULT_stage:
  DEFAULT_modifiers:
    SmoothQuantModifier: {smoothing_strength: 0.8}
    GPTQModifier:
      scheme: W8A8
      targets: Linear
      ignore: [lm_head]