Retrained with a more refined dataset with different captioning.
Overall the fidelity has improved and the model is less overfitting.
Recommended Weight: 0.6~1.0
Trigger Words: