Idea provided by Alex Sebastian: https://x.com/alexbastian_ai/status/1886253431557587419
3.1.0
3.0.0
2.0.0
- We have a new strategy: `Adaptive temperature strategy`. - `on_probs` now also receives `filtered_logits`, which is used in `Adaptive temperature strategy`. All past strategies now receive this param, but not use it.
1.2.0
- Human Guidance Strategy now has a new param for auto-selecting the top token. - README.md enhancements
1.1.0
Adding 2 more strategies: - Debug Strategy - Human Guidance Strategy