File size: 294 Bytes
122e4f1
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
default_stage:
  default_modifiers:
    QuantizationModifier:
      targets: [Linear]
      ignore: [lm_head, 're:.*visual.*', 're:.*linear_attn.in_proj_a$', 're:.*linear_attn.in_proj_b$']
      scheme: FP8_DYNAMIC
      bypass_divisibility_checks: false
      requires_calibration_data: false