composer.datasets.c4#

C4 (Colossal Cleaned Common Crawl) dataset.

This dataset is a colossal, cleaned version of Common Crawlโ€™s web crawl corpus and it is based on the Common Crawl dataset.

Classes

C4Dataset

Builds a streaming, sharded, sized torch.utils.data.IterableDataset for the C4 (Colossal Cleaned Common Crawl) dataset.

StreamingC4

Implementation of the C4 (Colossal Cleaned Common Crawl) dataset using StreamingDataset.
class composer.datasets.c4.C4Dataset(split, num_samples, tokenizer_name, max_seq_len, group_method, shuffle=False, shuffle_buffer_size=10000, seed=5)[source]#

Bases: torch.utils.data.dataset.IterableDataset

Builds a streaming, sharded, sized torch.utils.data.IterableDataset for the C4 (Colossal Cleaned Common Crawl) dataset. Used for pretraining autoregressive or masked language models. Text samples are streamed directly from the cloud using HuggingFaceโ€™s C4 Dataset with streaming backend (See https://huggingface.co/datasets/c4 for more details). The text samples are then shuffled, tokenized, and grouped on- the-fly.

Parameters

  • split (str) โ€“ What split of the dataset to use. Either 'train' or 'validation'.

  • num_samples (int) โ€“ The number of post-processed token samples, used to set epoch size of the torch.data.utils.IterableDataset.

  • tokenizer_name (str) โ€“ The name of the HuggingFace tokenizer to preprocess text with.

  • max_seq_len (int) โ€“ The max sequence length of each token sample.

  • group_method (str) โ€“ How to group text samples into token samples. Either 'truncate' or 'concat'.

  • shuffle (bool) โ€“ Whether to shuffle the samples in the dataset. Currently, shards are assigned and consumed with deterministic per-device shard order, but shuffling affects the order of samples via (per-device) shuffle buffers. Default: False.

  • shuffle_buffer_size (int) โ€“ If shuffle=True, samples are read into a buffer of this size (per-device), and randomly sampled from there to produce shuffled samples. Default: 10000.

  • seed (int) โ€“ If shuffle=True, what seed to use for shuffling operations. Default: 5.

Returns

IterableDataset โ€“ A torch.utils.data.IterableDataset object.

class composer.datasets.c4.StreamingC4(remote, local, split, shuffle, tokenizer_name, max_seq_len, group_method='truncate', max_retries=2, timeout=120, batch_size=None)[source]#

Bases: composer.datasets.streaming.dataset.StreamingDataset

Implementation of the C4 (Colossal Cleaned Common Crawl) dataset using StreamingDataset.

Parameters

  • remote (str) โ€“ Remote directory (S3 or local filesystem) where dataset is stored.

  • local (str) โ€“ Local filesystem directory where dataset is cached during operation.

  • split (str) โ€“ The dataset split to use, either โ€˜trainโ€™ or โ€˜valโ€™.

  • shuffle (bool) โ€“ Whether to shuffle the samples in this dataset.

  • tokenizer_name (str) โ€“ The name of the HuggingFace tokenizer to use to tokenize samples.

  • max_seq_len (int) โ€“ The max sequence length of each token sample.

  • group_method (str) โ€“ How to group text samples into token samples. Currently only supporting 'truncate'.

  • max_retries (int) โ€“ Number of download re-attempts before giving up. Default: 2.

  • timeout (float) โ€“ How long to wait for shard to download before raising an exception. Default: 120 sec.

  • batch_size (Optional[int]) โ€“ Hint batch_size that will be used on each deviceโ€™s DataLoader. Default: None.