from functools import partial import os import multiprocessing from datasets import load_dataset, load_from_disk import jsonlines """Get task list: !git clone https://github.com/bigscience-workshop/t-zero.git %cd t-zero !pip install -e .[seqio_tasks] !pip install -q py7zr import t0.seqio_tasks import seqio tasks = [task.name for task in seqio.MixtureRegistry.get('t0_train').tasks] print(tasks) After running the script, merge train & validation jsonls separately into two big files: cat folder_with_all_jsonl/*.jsonl > merged_file.jsonl """ TZERO_TASK_LIST = [ 'adversarial_qa_dbert_answer_the_following_q', 'adversarial_qa_dbert_based_on', 'adversarial_qa_dbert_generate_question', 'adversarial_qa_dbert_question_context_answer', 'adversarial_qa_dbert_tell_what_it_is', 'adversarial_qa_dbidaf_answer_the_following_q', 'adversarial_qa_dbidaf_based_on', 'adversarial_qa_dbidaf_generate_question', 'adversarial_qa_dbidaf_question_context_answer', 'adversarial_qa_dbidaf_tell_what_it_is', 'adversarial_qa_droberta_answer_the_following_q', 'adversarial_qa_droberta_based_on', 'adversarial_qa_droberta_generate_question', 'adversarial_qa_droberta_question_context_answer', 'adversarial_qa_droberta_tell_what_it_is', 'ag_news_classify', 'ag_news_classify_question_first', 'ag_news_classify_with_choices', 'ag_news_classify_with_choices_question_first', 'ag_news_recommend', 'ag_news_which_section', 'ag_news_which_section_choices', 'amazon_polarity_Is_this_product_review_positive', 'amazon_polarity_Is_this_review', 'amazon_polarity_Is_this_review_negative', 'amazon_polarity_User_recommend_this_product', 'amazon_polarity_convey_negative_or_positive_sentiment', 'amazon_polarity_flattering_or_not', 'amazon_polarity_negative_or_positive_tone', 'amazon_polarity_user_satisfied', 'amazon_polarity_would_you_buy', 'app_reviews_categorize_rating_using_review', 'app_reviews_convert_to_rating', 'app_reviews_convert_to_star_rating', 'app_reviews_generate_review', 'cnn_dailymail_3.0.0_2_or_3_sentences', 'cnn_dailymail_3.0.0_generate_story', 'cnn_dailymail_3.0.0_news_card_view', 'cnn_dailymail_3.0.0_news_stock', 'cnn_dailymail_3.0.0_news_summary', 'cnn_dailymail_3.0.0_spice_up_story', 'cnn_dailymail_3.0.0_sum_in_brief', 'cnn_dailymail_3.0.0_tldr_summary', 'cnn_dailymail_3.0.0_write_an_outline', 'common_gen_Example_prompt', 'common_gen_Given_concepts_type_1', 'common_gen_Given_concepts_type_2', 'common_gen_Put_together', 'common_gen_choice_in_concept_centric_sentence_generation', 'common_gen_random_task_template_prompt', 'common_gen_sentence_to_concepts', 'common_gen_topic_to_sentence', 'common_gen_topics_from_the_sentence', 'cos_e_v1.11_aligned_with_common_sense', 'cos_e_v1.11_description_question_option_id', 'cos_e_v1.11_description_question_option_text', 'cos_e_v1.11_explain_why_human', 'cos_e_v1.11_generate_explanation_given_text', 'cos_e_v1.11_i_think', 'cos_e_v1.11_question_description_option_id', 'cos_e_v1.11_question_description_option_text', 'cos_e_v1.11_question_option_description_id', 'cos_e_v1.11_question_option_description_text', 'cos_e_v1.11_rationale', 'cosmos_qa_context_answer_to_question', 'cosmos_qa_context_description_question_answer_id', 'cosmos_qa_context_description_question_answer_text', 'cosmos_qa_context_description_question_text', 'cosmos_qa_context_question_description_answer_id', 'cosmos_qa_context_question_description_answer_text', 'cosmos_qa_context_question_description_text', 'cosmos_qa_description_context_question_answer_id', 'cosmos_qa_description_context_question_answer_text', 'cosmos_qa_description_context_question_text', 'cosmos_qa_no_prompt_id', 'cosmos_qa_no_prompt_text', 'cosmos_qa_only_question_answer', 'dbpedia_14_given_a_choice_of_categories_', 'dbpedia_14_given_a_list_of_category_what_does_the_title_belong_to', 'dbpedia_14_given_list_what_category_does_the_paragraph_belong_to', 'dbpedia_14_pick_one_category_for_the_following_text', 'dream_answer_to_dialogue', 'dream_baseline', 'dream_generate_first_utterance', 'dream_generate_last_utterance', 'dream_read_the_following_conversation_and_answer_the_question', 'duorc_ParaphraseRC_answer_question', 'duorc_ParaphraseRC_build_story_around_qa', 'duorc_ParaphraseRC_decide_worth_it', 'duorc_ParaphraseRC_extract_answer', 'duorc_ParaphraseRC_generate_question', 'duorc_ParaphraseRC_generate_question_by_answer', 'duorc_ParaphraseRC_movie_director', 'duorc_ParaphraseRC_question_answering', 'duorc_ParaphraseRC_title_generation', 'duorc_SelfRC_answer_question', 'duorc_SelfRC_build_story_around_qa', 'duorc_SelfRC_decide_worth_it', 'duorc_SelfRC_extract_answer', 'duorc_SelfRC_generate_question', 'duorc_SelfRC_generate_question_by_answer', 'duorc_SelfRC_movie_director', 'duorc_SelfRC_question_answering', 'duorc_SelfRC_title_generation', 'gigaword_TLDR', 'gigaword_first_sentence_title', 'gigaword_generate_summary_for_this', 'gigaword_in_a_nutshell', 'gigaword_make_a_title', 'gigaword_reverse_writing', 'gigaword_write_a_title_for_this_sentence', 'gigaword_write_an_article', 'gigaword_write_its_sentence', 'glue_mrpc_equivalent', 'glue_mrpc_generate_paraphrase', 'glue_mrpc_generate_sentence', 'glue_mrpc_paraphrase', 'glue_mrpc_replace', 'glue_mrpc_same_thing', 'glue_mrpc_want_to_know', 'glue_qqp_answer', 'glue_qqp_duplicate', 'glue_qqp_duplicate_or_not', 'glue_qqp_meaning', 'glue_qqp_quora', 'glue_qqp_same_thing', 'imdb_Movie_Expressed_Sentiment', 'imdb_Movie_Expressed_Sentiment_2', 'imdb_Negation_template_for_positive_and_negative', 'imdb_Reviewer_Enjoyment', 'imdb_Reviewer_Enjoyment_Yes_No', 'imdb_Reviewer_Expressed_Sentiment', 'imdb_Reviewer_Opinion_bad_good_choices', 'imdb_Reviewer_Sentiment_Feeling', 'imdb_Sentiment_with_choices_', 'imdb_Text_Expressed_Sentiment', 'imdb_Writer_Expressed_Sentiment', 'kilt_tasks_hotpotqa_combining_facts', 'kilt_tasks_hotpotqa_complex_question', 'kilt_tasks_hotpotqa_final_exam', 'kilt_tasks_hotpotqa_formulate', 'kilt_tasks_hotpotqa_straighforward_qa', 'multi_news_distill', 'multi_news_expand_reverse_task_', 'multi_news_summarize', 'multi_news_summary_scenario', 'multi_news_synthesize', 'multi_news_what_are_the_key_points', 'paws_labeled_final_Concatenation', 'paws_labeled_final_Concatenation_no_label', 'paws_labeled_final_Meaning', 'paws_labeled_final_Meaning_no_label', 'paws_labeled_final_PAWS_ANLI_GPT3', 'paws_labeled_final_PAWS_ANLI_GPT3_no_label', 'paws_labeled_final_Rewrite', 'paws_labeled_final_Rewrite_no_label', 'paws_labeled_final_context_question', 'paws_labeled_final_context_question_no_label', 'paws_labeled_final_paraphrase_task', 'paws_labeled_final_task_description_no_label', 'qasc_is_correct_1', 'qasc_is_correct_2', 'qasc_qa_with_combined_facts_1', 'qasc_qa_with_separated_facts_1', 'qasc_qa_with_separated_facts_2', 'qasc_qa_with_separated_facts_3', 'qasc_qa_with_separated_facts_4', 'qasc_qa_with_separated_facts_5', 'quail_context_description_question_answer_id', 'quail_context_description_question_answer_text', 'quail_context_description_question_text', 'quail_context_question_answer_description_id', 'quail_context_question_answer_description_text', 'quail_context_question_description_answer_id', 'quail_context_question_description_answer_text', 'quail_context_question_description_text', 'quail_description_context_question_answer_id', 'quail_description_context_question_answer_text', 'quail_description_context_question_text', 'quail_no_prompt_id', 'quail_no_prompt_text', 'quarel_choose_between', 'quarel_do_not_use', 'quarel_heres_a_story', 'quarel_logic_test', 'quarel_testing_students', 'quartz_answer_question_based_on', 'quartz_answer_question_below', 'quartz_given_the_fact_answer_the_q', 'quartz_having_read_above_passage', 'quartz_paragraph_question_plain_concat', 'quartz_read_passage_below_choose', 'quartz_use_info_from_paragraph_question', 'quartz_use_info_from_question_paragraph', 'quoref_Answer_Friend_Question', 'quoref_Answer_Question_Given_Context', 'quoref_Answer_Test', 'quoref_Context_Contains_Answer', 'quoref_Find_Answer', 'quoref_Found_Context_Online', 'quoref_Given_Context_Answer_Question', 'quoref_Guess_Answer', 'quoref_Guess_Title_For_Context', 'quoref_Read_And_Extract_', 'quoref_What_Is_The_Answer', 'ropes_background_new_situation_answer', 'ropes_background_situation_middle', 'ropes_given_background_situation', 'ropes_new_situation_background_answer', 'ropes_plain_background_situation', 'ropes_plain_bottom_hint', 'ropes_plain_no_background', 'ropes_prompt_beginning', 'ropes_prompt_bottom_hint_beginning', 'ropes_prompt_bottom_no_hint', 'ropes_prompt_mix', 'ropes_read_background_situation', 'rotten_tomatoes_Movie_Expressed_Sentiment', 'rotten_tomatoes_Movie_Expressed_Sentiment_2', 'rotten_tomatoes_Reviewer_Enjoyment', 'rotten_tomatoes_Reviewer_Enjoyment_Yes_No', 'rotten_tomatoes_Reviewer_Expressed_Sentiment', 'rotten_tomatoes_Reviewer_Opinion_bad_good_choices', 'rotten_tomatoes_Reviewer_Sentiment_Feeling', 'rotten_tomatoes_Sentiment_with_choices_', 'rotten_tomatoes_Text_Expressed_Sentiment', 'rotten_tomatoes_Writer_Expressed_Sentiment', 'samsum_Generate_a_summary_for_this_dialogue', 'samsum_Given_the_above_dialogue_write_a_summary', 'samsum_Sum_up_the_following_dialogue', 'samsum_Summarize_', 'samsum_Summarize_this_dialogue_', 'samsum_To_sum_up_this_dialog', 'samsum_Write_a_dialogue_that_match_this_summary', 'sciq_Direct_Question', 'sciq_Direct_Question_Closed_Book_', 'sciq_Multiple_Choice', 'sciq_Multiple_Choice_Closed_Book_', 'sciq_Multiple_Choice_Question_First', 'social_i_qa_Check_if_a_random_answer_is_valid_or_not', 'social_i_qa_Generate_answer', 'social_i_qa_Generate_the_question_from_the_answer', 'social_i_qa_I_was_wondering', 'social_i_qa_Show_choices_and_generate_answer', 'social_i_qa_Show_choices_and_generate_index', 'trec_fine_grained_ABBR', 'trec_fine_grained_ABBR_context_first', 'trec_fine_grained_DESC', 'trec_fine_grained_DESC_context_first', 'trec_fine_grained_ENTY', 'trec_fine_grained_HUM', 'trec_fine_grained_HUM_context_first', 'trec_fine_grained_LOC', 'trec_fine_grained_LOC_context_first', 'trec_fine_grained_NUM', 'trec_fine_grained_NUM_context_first', 'trec_fine_grained_open', 'trec_fine_grained_open_context_first', 'trec_pick_the_best_descriptor', 'trec_trec1', 'trec_trec2', 'trec_what_category_best_describe', 'trec_which_category_best_describes', 'wiki_bio_comprehension', 'wiki_bio_guess_person', 'wiki_bio_key_content', 'wiki_bio_what_content', 'wiki_bio_who', 'wiki_hop_original_choose_best_object_affirmative_1', 'wiki_hop_original_choose_best_object_affirmative_2', 'wiki_hop_original_choose_best_object_affirmative_3', 'wiki_hop_original_choose_best_object_interrogative_1', 'wiki_hop_original_choose_best_object_interrogative_2', 'wiki_hop_original_explain_relation', 'wiki_hop_original_generate_object', 'wiki_hop_original_generate_subject', 'wiki_hop_original_generate_subject_and_object', 'wiki_qa_Decide_good_answer', 'wiki_qa_Direct_Answer_to_Question', 'wiki_qa_Generate_Question_from_Topic', 'wiki_qa_Is_This_True_', 'wiki_qa_Jeopardy_style', 'wiki_qa_Topic_Prediction_Answer_Only', 'wiki_qa_Topic_Prediction_Question_Only', 'wiki_qa_Topic_Prediction_Question_and_Answer_Pair', 'wiki_qa_automatic_system', 'wiki_qa_exercise', 'wiki_qa_found_on_google', 'wiqa_does_the_supposed_perturbation_have_an_effect', 'wiqa_effect_with_label_answer', 'wiqa_effect_with_string_answer', 'wiqa_what_is_the_final_step_of_the_following_process', 'wiqa_what_is_the_missing_first_step', 'wiqa_what_might_be_the_first_step_of_the_process', 'wiqa_what_might_be_the_last_step_of_the_process', 'wiqa_which_of_the_following_is_the_supposed_perturbation', 'xsum_DOC_boils_down_to_simple_idea_that', 'xsum_DOC_given_above_write_one_sentence', 'xsum_DOC_how_would_you_rephrase_few_words', 'xsum_DOC_tldr', 'xsum_DOC_write_summary_of_above', 'xsum_article_DOC_summary', 'xsum_college_roommate_asked_DOC_so_I_recap', 'xsum_read_below_DOC_write_abstract', 'xsum_summarize_DOC', 'xsum_summarize_this_DOC_summary', 'yelp_review_full_based_on_that', 'yelp_review_full_format_rating', 'yelp_review_full_format_score', 'yelp_review_full_format_star', 'yelp_review_full_on_a_scale', 'yelp_review_full_so_i_would', 'yelp_review_full_this_place' ] # Optonally download all first # for task_name in TZERO_TASK_LIST: # ds = load_dataset("bigscience/P3", task_name) def write_to_jsonl_hub(task_name, split): # Could also use ds.to_json() ds = load_dataset("bigscience/P3", task_name) if split in ds: with jsonlines.open(f'p3_{task_name}_{split}.jsonl', mode='w') as writer: for example in ds[split].select(range(len(ds[split]))): writer.write({ "inputs": example["inputs_pretokenized"], "targets": example["targets_pretokenized"] }) def write_to_jsonl_disk(task_name, split): ds = load_from_disk(f"{os.environ['six_ALL_CCFRSCRATCH']}/datasets/p3/{task_name}") if split in ds: with jsonlines.open(f'p3_{task_name}_{split}.jsonl', mode='w') as writer: for example in ds[split].select(range(len(ds[split]))): writer.write({ "inputs": example["inputs_pretokenized"], "targets": example["targets_pretokenized"] }) with multiprocessing.Pool(num_proc=multiprocessing.cpu_count()) as pool: pool.map(partial(write_to_jsonl_disk, split="train"), TZERO_TASK_LIST) pool.map(partial(write_to_jsonl_disk, split="validation"), TZERO_TASK_LIST)