Update README.md
Browse files
README.md
CHANGED
@@ -28,6 +28,7 @@ SambaLingo-Hungarian-Base is a pretrained Bi-lingual Hungarian and English model
|
|
28 |
- **Language(s):** Hungarian, English
|
29 |
- **Finetuned from model:** [Llama 2](https://huggingface.co/meta-llama/Llama-2-7b-hf)
|
30 |
- **Try the chat version of this model**: [SambaLingo-chat-space](https://huggingface.co/spaces/sambanovasystems/SambaLingo-chat-space).
|
|
|
31 |
- **Blog Post**: [sambalingo-open-source-language-experts](https://sambanova.ai/blog/sambalingo-open-source-language-experts)
|
32 |
|
33 |
## Getting Started
|
@@ -53,15 +54,7 @@ All pre-training is done on the [Cultura-X](https://huggingface.co/datasets/uonl
|
|
53 |
We extended the vocabulary of the base llama model from 32,000 tokens to 57,000 tokens by adding up to 25,000 non-overlapping tokens from the new language.
|
54 |
|
55 |
## Evaluation
|
56 |
-
|
57 |
-
|------------------------------|-------------|-----------|-----------|----------|--------|
|
58 |
-
| Perplexity (Lower Is Better) | **1.614** | 1.720 | 3.284 | 3.687 | 2.488 |
|
59 |
-
| FLORES en->hu (8 shot, CHRF) | **0.496** | 0.422 | 0.113 | 0.019 | 0.200 |
|
60 |
-
| FLORES hu->en (8 shot, CHRF) | **0.558** | 0.459 | 0.162 | 0.146 | 0.174 |
|
61 |
-
| FLORES en->hu (8 shot, BLEU) | **0.164** | 0.098 | 0.002 | 0.001 | 0.010 |
|
62 |
-
| FLORES hu->en (8 shot, BLEU) | **0.261** | 0.154 | 0.004 | 0.004 | 0.009 |
|
63 |
-
| Belebele (3 shot) | **41.78%** | 27.67% | 26.78% | 24.00% | 24.56% |
|
64 |
-
| SIB-200 (3 shot) | **57.35%** | 51.96% | 35.78% | 41.18% | 41.18% |
|
65 |
|
66 |
|
67 |
## Uses
|
@@ -103,12 +96,12 @@ We would like to give a special thanks to the following groups:
|
|
103 |
|
104 |
## Cite SambaLingo
|
105 |
```
|
106 |
-
@
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
}
|
114 |
```
|
|
|
28 |
- **Language(s):** Hungarian, English
|
29 |
- **Finetuned from model:** [Llama 2](https://huggingface.co/meta-llama/Llama-2-7b-hf)
|
30 |
- **Try the chat version of this model**: [SambaLingo-chat-space](https://huggingface.co/spaces/sambanovasystems/SambaLingo-chat-space).
|
31 |
+
- **Paper:** [SambaLingo: Teaching Large Language Models New Languages](https://arxiv.org/abs/2404.05829)
|
32 |
- **Blog Post**: [sambalingo-open-source-language-experts](https://sambanova.ai/blog/sambalingo-open-source-language-experts)
|
33 |
|
34 |
## Getting Started
|
|
|
54 |
We extended the vocabulary of the base llama model from 32,000 tokens to 57,000 tokens by adding up to 25,000 non-overlapping tokens from the new language.
|
55 |
|
56 |
## Evaluation
|
57 |
+
For evaluation results see our paper: [SambaLingo: Teaching Large Language Models New Languages](https://arxiv.org/abs/2404.05829)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
|
59 |
|
60 |
## Uses
|
|
|
96 |
|
97 |
## Cite SambaLingo
|
98 |
```
|
99 |
+
@misc{csaki2024sambalingo,
|
100 |
+
title={SambaLingo: Teaching Large Language Models New Languages},
|
101 |
+
author={Zoltan Csaki and Bo Li and Jonathan Li and Qiantong Xu and Pian Pawakapan and Leon Zhang and Yun Du and Hengyu Zhao and Changran Hu and Urmish Thakker},
|
102 |
+
year={2024},
|
103 |
+
eprint={2404.05829},
|
104 |
+
archivePrefix={arXiv},
|
105 |
+
primaryClass={cs.CL}
|
106 |
}
|
107 |
```
|