3 ms·
Can anyone speculate on the best way to update an LLM with new data? Do you fine-tune it on just the new data? Do you mix it with the old data and train for ano
by etaioinshrdlu 3y ago
Can anyone speculate on the best way to update an LLM with new data? Do you fine-tune it on just the new data? Do you mix it with the old data and train for another epoch?
- vincent_s 3y agoMost deep learning frameworks allow you to save and load model weights. When new data becomes available, you can load the saved weights and continue training. This is essentially what happens during fine-tuning. So, yes, they probably could just update GPT with new data. But there would be a difference in outcome compared to training a model from scratch with all the info available in 2023. Training from scratch would ensure that newer data has equal footing with older data. So a model trained from scratch in 2023 might have a more balanced understanding of both old and new data. Also there's a chance that the model might identify new patterns or nuances. When you just fine-tune the existing model there might be areas where the model's response leans more towards the knowledge from 2021. Having said that, training the model from scratch with all the new data would be incredibly resource-intensive, so I'm pretty sure OpenAI didn't go that route.