Uptated the dataset with the latest releases, also made some adjustments to the pre train settings, hope it works better than the previous one.