lm_eval/tasks/super_glue/rte/t5-prompt.yaml (22 lines of code) (raw):

tag: - super-glue-t5-prompt task: super_glue-rte-t5-prompt dataset_path: super_glue dataset_name: rte training_split: train validation_split: validation output_type: generate_until doc_to_text: "rte hypothesis: {{hypothesis}} premise: {{premise}}" doc_to_target: label doc_to_choice: ['entailment', 'not_entailment'] generation_kwargs: until: - "</s>" metric_list: - metric: exact_match aggregation: mean higher_is_better: true ignore_case: true ignore_punctuation: true metadata: version: 0.0