Commit
•
b53a98d
1
Parent(s):
b134e86
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
---
|
3 |
+
language:
|
4 |
+
- zh
|
5 |
+
tags:
|
6 |
+
- generation
|
7 |
+
- question answering
|
8 |
+
- instruction tuning
|
9 |
+
license: cc-by-nc-4.0
|
10 |
+
---
|
11 |
+
|
12 |
+
### Model Description
|
13 |
+
|
14 |
+
This HF repository contains base LLMs instruction tuned (SFT) with LoRA and then used to study whether monolingual or multilingual instruction tuning is more favourable.
|
15 |
+
* [GitHub](https://github.com/hplt-project/monolingual-multilingual-instruction-tuning/tree/main)
|
16 |
+
* [Paper](https://arxiv.org/abs/2309.08958)
|
17 |
+
|
18 |
+
#### Instruction tuning details
|
19 |
+
* Base model: [EleutherAI/pythia-6.9b-deduped](https://huggingface.co/EleutherAI/pythia-6.9b-deduped)
|
20 |
+
* Instruction tuning language: Chinese
|
21 |
+
* Training method: LoRA.
|
22 |
+
* LoRA details: rank=8, alpha=16, target modules={key, query, value}.
|
23 |
+
* Best checkpoint: best cross-entropy on a validation set, trained for 5 epochs.
|
24 |
+
* Dataset: machine-translated from [yahma/alpaca-cleaned](https://huggingface.co/datasets/yahma/alpaca-cleaned). You can download our data [HERE](https://github.com/hplt-project/monolingual-multilingual-instruction-tuning/tree/main/training-data).
|
25 |
+
|
26 |
+
#### Usage
|
27 |
+
The model checkpoint should be loaded with the base model together using `transformers` and `peft` libraries.
|
28 |
+
|
29 |
+
Please refer to our Github repository [HERE](https://github.com/hplt-project/monolingual-multilingual-instruction-tuning/tree/main/loraft) for inference and training instructions.
|
30 |
+
|
31 |
+
#### Citation
|
32 |
+
```
|
33 |
+
@inproceedings{chen-etal-2024-monolingual,
|
34 |
+
title="Monolingual or multilingual instruction tuning: Which makes a better {Alpaca}",
|
35 |
+
author="Pinzhen Chen and Shaoxiong Ji and Nikolay Bogoychev and Andrey Kutuzov and Barry Haddow and Kenneth Heafield",
|
36 |
+
year="2024",
|
37 |
+
booktitle = "Findings of the Association for Computational Linguistics: EACL 2024",
|
38 |
+
}
|
39 |
+
```
|
40 |
+
|