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