ModelOpt Quantize UNet

ModelOptQuantizeUNet

Quantize UNet model with NVIDIA ModelOpt (INT8/FP8/INT4)

Pack: comfy-modelopt

custom_nodes.comfy-modelopt

Inputs (5)

Name	Type	Required
model	MODEL	required
precision	COMBO	required
calibration_steps	INT	required
calibration_data	LATENT	optional
skip_layers	STRING	optional

Outputs (1)

Name	Type
quantized_model	MODEL