Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,32 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- datajuicer/redpajama-wiki-refined-by-data-juicer
|
| 5 |
+
- datajuicer/redpajama-arxiv-refined-by-data-juicer
|
| 6 |
+
- datajuicer/redpajama-c4-refined-by-data-juicer
|
| 7 |
+
- datajuicer/redpajama-book-refined-by-data-juicer
|
| 8 |
+
- datajuicer/redpajama-cc-2019-30-refined-by-data-juicer
|
| 9 |
+
- datajuicer/redpajama-cc-2020-05-refined-by-data-juicer
|
| 10 |
+
- datajuicer/redpajama-cc-2021-04-refined-by-data-juicer
|
| 11 |
+
- datajuicer/redpajama-cc-2022-05-refined-by-data-juicer
|
| 12 |
+
- datajuicer/redpajama-cc-2023-06-refined-by-data-juicer
|
| 13 |
+
- datajuicer/redpajama-pile-stackexchange-refined-by-data-juicer
|
| 14 |
+
- datajuicer/redpajama-stack-code-refined-by-data-juicer
|
| 15 |
+
- datajuicer/the-pile-nih-refined-by-data-juicer
|
| 16 |
+
- datajuicer/the-pile-europarl-refined-by-data-juicer
|
| 17 |
+
- datajuicer/the-pile-philpaper-refined-by-data-juicer
|
| 18 |
+
- datajuicer/the-pile-pubmed-abstracts-refined-by-data-juicer
|
| 19 |
+
- datajuicer/the-pile-pubmed-central-refined-by-data-juicer
|
| 20 |
+
- datajuicer/the-pile-freelaw-refined-by-data-juicer
|
| 21 |
+
- datajuicer/the-pile-hackernews-refined-by-data-juicer
|
| 22 |
---
|
| 23 |
+
|
| 24 |
+
This is a reference LLM from [Data-Juicer](https://github.com/alibaba/data-juicer).
|
| 25 |
+
|
| 26 |
+
The model architecture is LLaMA-1.3B and we adopt the [OpenLLaMA](https://github.com/openlm-research/open_llama) implementation.
|
| 27 |
+
The model is pre-trained on 150B tokens of Data-Juicer's refined RedPajama and Pile, and 4.7B tokens of Data-Juicer refined instruct data.
|
| 28 |
+
It achieves an average score of 36.76 over 16 HELM tasks, improved the OpenLLaMA-DJ-150B by 2.55 point.
|
| 29 |
+
|
| 30 |
+
For more details, please refer to our [paper](https://arxiv.org/abs/2309.02033).
|
| 31 |
+
|
| 32 |
+
|