Skip to main content
This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub
The ModelComputeDtype node allows you to change the computational data type used by a model during inference. It creates a copy of the input model and applies the specified data type setting, which can help optimize memory usage and performance depending on your hardware capabilities. This is particularly useful for debugging and testing different precision settings.

Inputs

ParameterData TypeRequiredRangeDescription
modelMODELYes-The input model to modify with a new compute data type
dtypeSTRINGYes”default"
"fp32"
"fp16"
"bf16”
The computational data type to apply to the model

Outputs

Output NameData TypeDescription
modelMODELThe modified model with the new compute data type applied