allenai/smashed

SMASHED is a toolkit designed to apply transformations to samples in datasets, such as fields extraction, tokenization, prompting, batching, and more. Supports datasets from Huggingface, torchdata iterables, or simple lists of dictionaries.

46
/ 100
Emerging

This tool helps machine learning engineers preprocess text data for language models. It takes raw text (like sentences or documents) as input and transforms it into numerical tokens, correctly formatted with padding and attention masks. It's designed for ML engineers building or fine-tuning models that consume large text datasets.

Used by 2 other packages. No commits in the last 6 months. Available on PyPI.

Use this if you need to efficiently prepare diverse text datasets for training or inference with transformer-based language models, handling tokenization, length constraints, and batching.

Not ideal if you're looking for a general-purpose data cleaning or ETL tool for non-textual data or for users who are not working with machine learning models.

natural-language-processing machine-learning-engineering text-preprocessing language-model-training
Stale 6m
Maintenance 0 / 25
Adoption 9 / 25
Maturity 25 / 25
Community 12 / 25

How are scores calculated?

Stars

35

Forks

5

Language

Python

License

Apache-2.0

Last pushed

May 24, 2024

Commits (30d)

0

Dependencies

7

Reverse dependents

2

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/allenai/smashed"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.