Spaces:
Sleeping
Sleeping
| def format_results(model_name: str, revision: str, precision: str, overall_js: float, overall_ci: tuple, **experiment_scores) -> dict: | |
| """ | |
| Formats the evaluation results into a structured dictionary. | |
| Args: | |
| model_name (str): The name of the evaluated model. | |
| revision (str): The revision hash of the model. | |
| precision (str): The precision with which the evaluation was run. | |
| overall_js (float): The overall average JS divergence. | |
| overall_ci (tuple): The confidence interval for the overall JS divergence. | |
| experiment_scores: Experiment-specific scores and confidence intervals (E1, E1_ci, E2, E2_ci, ...). | |
| Returns: | |
| dict: A dictionary containing the structured evaluation results. | |
| """ | |
| # Initialize the base structure | |
| results = { | |
| "config": { | |
| "model_dtype": precision, # Precision with which you ran the evaluation | |
| "model_name": model_name, # Name of the model | |
| #"model_sha": revision # Hash of the model | |
| }, | |
| "results": { | |
| "overall_js_divergence": overall_js, # Overall JS divergence | |
| "overall_confidence_interval": overall_ci, # Confidence interval for the overall JS divergence | |
| } | |
| } | |
| # Add experiment-specific results to the dictionary | |
| for exp_name, score in experiment_scores.items(): | |
| results["results"][exp_name] = score # Add each experiment score and its CI | |
| return results | |