lm_eval/tasks/super_glue/cb/t5-prompt.yaml

tag:
  - super-glue-t5-prompt
task: super_glue-cb-t5-prompt
dataset_path: super_glue
dataset_name: cb
training_split: train
validation_split: validation
output_type: generate_until
doc_to_text: "cb hypothesis: {{hypothesis}} premise: {{premise}}"
doc_to_target: label
doc_to_choice: ['entailment', 'contradiction', 'neutral']
generation_kwargs:
  until:
    - "</s>"
metric_list:
  - metric: exact_match
    aggregation: mean
    higher_is_better: true
    ignore_case: true
    ignore_punctuation: true
  - metric: !function "t5_utils.mean_3class_f1"
    aggregation: !function "t5_utils.agg_mean_3class_f1"
    higher_is_better: true
metadata:
  version: 0.0