3 ms·
Is there any path towards folding tokens into the actual model? That is, continual training rather than the current "training first then just tokens after"
by Workaccount2 4y ago
Is there any path towards folding tokens into the actual model? That is, continual training rather than the current "training first then just tokens after"
- ilaksh 4y agoPaLM 2 on Vertex AI which Google just released yesterday has fine tuning the large models as a core part of their offering.