recipe.yaml
| 1 | quant_stage: |
| 2 | quant_modifiers: |
| 3 | QuantizationModifier: |
| 4 | ignore: [lm_head] |
| 5 | config_groups: |
| 6 | group_0: |
| 7 | weights: {num_bits: 8, type: float, strategy: channel, dynamic: false, symmetric: true} |
| 8 | input_activations: {num_bits: 8, type: float, strategy: token, dynamic: true, symmetric: true} |
| 9 | targets: [Linear] |
| 10 | |