Skip to main content
This node sets the compute dtype used when a VAE encodes or decodes, without changing the rest of the model graph. It returns a copy of the input VAE with the requested precision, so a workflow can run one branch at a higher precision while others stay unchanged. “default” keeps whatever compute dtype the input VAE already uses.

Inputs

Outputs

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): 239e5829d76b4bd43ec300fc2b7c8c408b5bd4a93918da417ecf43817aee3e12