metadata
license: apache-2.0
datasets:
- gair-prox/RedPajama-pro
language:
- en
tags:
- llama
RedPJ-ProX-1.7B
RedPJ-ProX-1.7B is a small language model. It was and trained on the RedPajama-V2-pro for 50B tokens.
Evaluations
ProX models are evaluated over 10 language model benchmarks in zero-shot setting.
| ArC-c | ARC-e | CSQA | HellaS | MMLU | OBQA | PiQA | SIQA | WinoG | SciQ | AVG | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| raw | 26.9 | 51.4 | 32.4 | 47.3 | 29.3 | 32.2 | 69.7 | 39.6 | 52.1 | 79.1 | 46.0 |
| ours | 31.1 | 60.7 | 29.8 | 51.0 | 31.7 | 33.2 | 70.9 | 39.2 | 53.3 | 79.1 | 48.0 |
Citation
@article{zhou2024programming,
title={Programming Every Example: Lifting Pre-training Data Quality like Experts at Scale},
author={Zhou, Fan and Wang, Zengzhi and Liu, Qian and Li, Junlong and Liu, Pengfei},
journal={arXiv preprint arXiv:2409.17115},
year={2024}
}