Instructions to use yass4/halt-cot with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use yass4/halt-cot with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("yass4/halt-cot", dtype="auto") - Notebooks
- Google Colab
- Kaggle
| """HALT-CoT: entropy-based early stopping for chain-of-thought reasoning.""" | |
| from .core import ( | |
| AnswerCandidate, | |
| AnswerDistribution, | |
| EntropyHaltingController, | |
| HaltCoTConfig, | |
| HaltCoTResult, | |
| HaltCoTStep, | |
| HaltDecision, | |
| answer_distribution_from_scores, | |
| entropy_from_probabilities, | |
| format_reasoning, | |
| integer_candidates, | |
| multiple_choice_candidates, | |
| normalize_candidates, | |
| numeric_candidates_from_texts, | |
| softmax, | |
| yes_no_candidates, | |
| ) | |
| __all__ = [ | |
| "AnswerCandidate", | |
| "AnswerDistribution", | |
| "EntropyHaltingController", | |
| "HaltCoTConfig", | |
| "HaltCoTResult", | |
| "HaltCoTStep", | |
| "HaltDecision", | |
| "answer_distribution_from_scores", | |
| "entropy_from_probabilities", | |
| "format_reasoning", | |
| "integer_candidates", | |
| "multiple_choice_candidates", | |
| "normalize_candidates", | |
| "numeric_candidates_from_texts", | |
| "softmax", | |
| "yes_no_candidates", | |
| ] | |