import os import torch from dataclasses import dataclass from enum import Enum from src.envs import CACHE_PATH @dataclass class Task: benchmark: str metric: str col_name: str num_fewshot: int class Tasks(Enum): # task_key in the json file, metric_key in the json file, name to display in the leaderboard # task0 = Task("anli_r1", "acc", "ANLI") # task1 = Task("logiqa", "acc_norm", "LogiQA") task0 = Task("nq_open", "em", "NQ Open", 64) # 64, as in the ATLAS paper task1 = Task("triviaqa", "em", "TriviaQA", 64) # 64, as in the ATLAS paper # TruthfulQA is intended as a zero-shot benchmark [5, 47]. https://owainevans.github.io/pdfs/truthfulQA_lin_evans.pdf # task2 = Task("truthfulqa_gen", "rougeL_acc", "TruthfulQA Gen", 0) task3 = Task("truthfulqa_mc1", "acc", "TruthfulQA MC1", 0) task4 = Task("truthfulqa_mc2", "acc", "TruthfulQA MC2", 0) task5 = Task("halueval_qa", "acc", "HaluEval QA", 0) # task6 = Task("halueval_dialogue", "acc", "HaluEval Dialogue", 0) # task7 = Task("halueval_summarization", "acc", "HaluEval Summarization", 0) task8 = Task("xsum", "rougeL", "XSum", 2) task9 = Task("cnndm", "rougeL", "CNN/DM", 2) task10 = Task("memo-trap", "acc", "memo-trap", 0) task11 = Task("nq8", "em", "NQ Open 8", 8) task12 = Task("tqa8", "em", "TriviaQA 8", 8) task13 = Task("ifeval", "prompt_level_strict_acc", "IFEval", 0) # NUM_FEWSHOT = 64 # Change with your few shot EVAL_REQUESTS_PATH_BACKEND = os.path.join(CACHE_PATH, "eval-queue-bk") EVAL_RESULTS_PATH_BACKEND = os.path.join(CACHE_PATH, "eval-results-bk") DEVICE = "cuda" if torch.cuda.is_available() else 'cpu' LIMIT = None # Testing; needs to be None