Important

NeMo 2.0 is an experimental feature and currently released in the dev container only: nvcr.io/nvidia/nemo:dev. Please refer to NeMo 2.0 overview for information on getting started.

Task Decontamination

Base Class

class nemo_curator.tasks.DownstreamTask

class nemo_curator.tasks.import_task(task_path)

Module

class nemo_curator.TaskDecontamination(tasks: Union[nemo_curator.tasks.downstream_task.DownstreamTask, Iterable[nemo_curator.tasks.downstream_task.DownstreamTask]], text_field='text', max_ngram_size=13, max_matches=10, min_document_length=200, remove_char_each_side=200, max_splits=10, removed_dir=None)

prepare_task_ngram_count() → dict

Computes a dictionary of all ngrams in each task as keys and each value set to 0.

Tasks

class nemo_curator.tasks.Race(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.Squad(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.ArcEasy(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.ArcChallenge(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.OpenBookQA(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.BoolQ(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.Copa(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.RTE(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.MultiRC(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.WSC(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.CB(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.ANLI(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.Record(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.COQA(file_path, min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.TriviaQA(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.Quac(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.WebQA(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.Drop(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.WiC(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.MMLU(path=None, min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.BigBenchHard(path=None, min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.BigBenchLight(path=None, min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.Multilingual(path=None, min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.PIQA(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.Winogrande(min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.Lambada(file_path, min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.NumDasc(n, file_path, min_ngram_size=8, max_ngram_size=13)

class nemo_curator.tasks.StoryCloze(file_path, min_ngram_size=8, max_ngram_size=13)