| default_stage: | |
| default_modifiers: | |
| QuantizationModifier: | |
| config_groups: | |
| attention_quant: | |
| targets: ['re:.*self_attn$'] | |
| weights: null | |
| input_activations: | |
| num_bits: 8 | |
| type: float | |
| symmetric: true | |
| group_size: null | |
| strategy: !!python/object/apply:compressed_tensors.quantization.quant_args.QuantizationStrategy [ | |
| tensor] | |
| block_structure: null | |
| dynamic: false | |
| actorder: null | |
| observer: minmax | |
| observer_kwargs: {} | |
| output_activations: null | |
| targets: [Linear] | |
| ignore: [lm_head] | |