The mini AI series ; I found that i needed smaller models for other models such as VisionEncoderDecoder : and other models which contain llms as a par