utils.callbacks.qat

utils.callbacks.qat

QAT Callback for HF Causal Trainer

Classes

Name Description
QATCallback Callback to toggle fake quantization for the model.

QATCallback

utils.callbacks.qat.QATCallback(cfg)

Callback to toggle fake quantization for the model.

Functions

Name Description
toggle_fake_quant Toggle fake quantization for any fake quantized linear or embedding layers in the model.

toggle_fake_quant

utils.callbacks.qat.toggle_fake_quant(mod, enable)

Toggle fake quantization for any fake quantized linear or embedding layers in the model.

Parameters

Name Type Description Default
mod nn.Module The module to toggle fake quantization for. required
enable bool Whether to enable or disable fake quantization. required