Reduce SmoothQuant Repr

[kylesayrs]

Purpose

• Reduce verbosity when debugging SmoothQuantModifier

Changes

resolved_mappings_: Optional[List] = Field(default=None, repr=False)
scales_: Optional[Dict] = Field(default=None, repr=False)

Testing

Before

SmoothQuantModifier(index=0, group='DEFAULT', start=None, end=None, update=None, initialized_=False, finalized_=False, st
arted_=False, ended_=False, smoothing_strength=0.8, ignore=[], num_calibration_steps=None, calibration_function=None, res
olved_mappings_=[SmoothQuantMapping(smooth_name='model.layers.0.input_layernorm', smooth_layer=LlamaRMSNorm((4096,), eps=
1e-05), balance_layers=[Linear(in_features=4096, out_features=4096, bias=False), Linear(in_features=4096, out_features=10
24, bias=False), Linear(in_features=4096, out_features=1024, bias=False)]),  ...

After

SmoothQuantModifier(index=0, group='DEFAULT', start=None, end=None, update=None, initialized_=False, finalized_=False, started_=False, ended_=False, smoothing_strength=0.8, mappings=[LayerMap(balance_layers=['re:.*q_proj', 're:.*k_proj', 're:.*v_proj'], smooth_layers='re:.*input_layernorm'), LayerMap(balance_layers=['re:.*gate_proj', 're:.*up_proj'], smooth_layers='re:.*post_attention_layernorm')], ignore=[], num_calibration_steps=None, calibration_function=None, scales_=None)

[github-actions]

👋 Hi! Thank you for contributing to llm-compressor. Please add the ready label when the PR is ready for review.

Note: This is required to complete the testing suite, please only add the label once the PR is code complete and local testing has been performed.

[brian-dellabetta]

cool!