V2: More refined dataset with better weights for more stable output, should do well if you want thicker . and were utilized for the training set, so please prompt accordingly.