This is the final version, I used SVD training script to avoid model overfitting the style of project dive future tone. Now this lora performs well in different checkpoint.
This is the final version, I used SVD training script to avoid model overfitting the style of project dive future tone. Now this lora performs well in different checkpoint.