Extend the start up time from 1200 to 1500 (for aws silicon instances) + add HF_HUB_ENABLE_HF_TRANSFER for faster downloads of the model