@EthanMarx 's interval fine-tune models are achieving validation AUROCs way below the original model. He's currently running from-scratch retrainings, but early results indicate these will arrive at the same values. Filing this issue to share results when they're in and discuss the implications to finalize a plot for the paper.
@EthanMarx 's interval fine-tune models are achieving validation AUROCs way below the original model. He's currently running from-scratch retrainings, but early results indicate these will arrive at the same values. Filing this issue to share results when they're in and discuss the implications to finalize a plot for the paper.