To using LoHa u need to install extension for WebUI.
KohakuBlueleaf/LyCORIS: Lora beYond Conventional methods, Other Rank adaptation Implementations for Stable diffusion. (github.com)

v1: This is a LoHa model for style learning test. It sometimes may not generate the pic u want. The weight is recommanded to be more than 1.0 .

v3: This version works better on eyes.

any: This version is trained based on Anything-v3 model. It mantains the original style better on more checkpoints but sometimes shows worse generalization ability in test.