How to convert LORA to use with a 2.1 ckpt model?
Is it possible to convert a LORA trained in 1.5 to use with a 2.1 model? Can anyone tell me how to do that?
It is possible to convert a LORA trained in 1.5 to use with a 2.1 model, but the process can be complex and may require adjusting the hyperparameters to achieve optimal performance. One approach to do this is to fine-tune the 1.5 model on additional data using the 2.1 architecture, and then retrain the model on the original dataset with the updated architecture. Another approach is to use a tool like ONNX to convert the 1.5 model to a format that is compatible with the 2.1 model, but this may also require adjustments to the hyperparameters.