File size: 344 Bytes
c25bd83
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
default_stage:
  default_modifiers:
    QuantizationModifier:
      targets: ['re:.*language_model\.layers\.\d+\.mlp\.(gate|up|down)_proj$']
      ignore: [lm_head, 're:.*visual.*', 're:.*linear_attn\.in_proj_a$', 're:.*linear_attn\.in_proj_b$']
      scheme: NVFP4
      bypass_divisibility_checks: false
      requires_calibration_data: true