File size: 237 Bytes
df6a6b5
 
 
 
 
 
 
9b97240
1
2
3
4
5
6
7
8
9
default_stage:
  default_modifiers:
    QuantizationModifier:
      targets: [Linear]
      ignore: [lm_head, 're:.*mlp.gate[.].*']
      scheme: FP8_DYNAMIC
      bypass_divisibility_checks: false
      requires_calibration_data: false