Temperature, top-k, top-p
The model gives probabilities, you make the choice. The same model, the same prompt: these three numbers change the output completely.
2 steps
110 XP
A free account is needed
Start the lesson →
Sources
- Holtzman, A. et al. 2020 · The Curious Case of Neural Text Degeneration (nucleus sampling) · ICLR 2020
- Fan, A. et al. 2018 · Hierarchical Neural Story Generation (top-k sampling) · ACL 2018
- Hinton, Vinyals, Dean 2015 · Distilling the Knowledge in a Neural Network (sıcaklık kavramı) · arXiv:1503.02531
ML Academy · an interactive machine learning course that runs in your browser ·
All lessons