News for hackers. Stuff that matters.
Re:Our Humble Attempt at “How Much Data Is Needed to Fine-Tune by Scene_Cast2 on Sunday September 24, 2023 @09:34PM (#37638770)
I'm surprised that this few samples made such a big difference. Perhaps the learning rate is jacked up? Otherwise, the last couple of batches during regular training would set the overall model tone.
Re:Our Humble Attempt at “How Much Data Is Needed to Fine-Tune by 3abiton on Monday September 25, 2023 @06:38AM (#37641679)
One thing they did not test was the complexity of the task. I doubt with "hundred" samples it would be possible to get it to be that accurately usable. But experiments are needed to back up that claim.