File size: 200 Bytes
8c660d1 |
1 2 3 4 5 6 7 8 |
default_stage:
default_modifiers:
SmoothQuantModifier: {smoothing_strength: 0.5}
QuantizationModifier:
targets: [Linear]
ignore: [lm_head, model.embed_tokens]
scheme: W8A8
|