get_default_qat_qconfig_mapping()Build the default QAT QConfigMapping.
Convenience constructor returning a fresh mapping whose global tier is
get_default_qat_qconfig and whose per-type / per-name tiers are empty. Every
eligible module therefore trains with lucid.quantization.FakeQuantize on
its activations and weights unless you add an override. This is the mapping
lucid.quantization.prepare_qat uses when called without an explicit one.
Returns
QConfigMappingA mapping whose global QConfig is the default QAT config and whose
type / name tiers are empty (ready to receive overrides).
Notes
- Add exceptions with
QConfigMapping.set_object_type/QConfigMapping.set_module_name; mapping a module toNonetrains it in float (no fake-quant inserted). - For static post-training quantization use
get_default_qconfig_mappinginstead — it installs bare observers rather than fake-quant modules.
Examples
>>> import lucid.nn as nn
>>> import lucid.quantization as Q
>>> mapping = Q.get_default_qat_qconfig_mapping()
>>> type(mapping.get_qconfig(nn.Linear, "fc").activation()).__name__
'FakeQuantize'See Also
get_default_qat_qconfig—The recipe installed as the global tier.get_default_qconfig_mapping—The static-PTQ counterpart.QConfigMapping—The three-tier resolution object this returns.