lm_eval/tasks/tinyBenchmarks/tinyHellaswag.yaml

task: tinyHellaswag
dataset_path: tinyBenchmarks/tinyHellaswag
dataset_name: null
output_type: multiple_choice
training_split: train
validation_split: validation
num_fewshot: 10
test_split: null
process_docs: !function utils_hellaswag.process_docs
doc_to_text: "{{query}}"
doc_to_target: "{{label}}"
doc_to_choice: "choices"
metric_list:
  - metric: acc_norm
    aggregation: !function agg_functions.agg_gpirt_hellaswag
    higher_is_better: true
metadata:
  version: 0.0