RichardErkhov
commited on
Commit
•
aabf0df
1
Parent(s):
a44b22b
uploaded readme
Browse files
README.md
ADDED
@@ -0,0 +1,118 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Quantization made by Richard Erkhov.
|
2 |
+
|
3 |
+
[Github](https://github.com/RichardErkhov)
|
4 |
+
|
5 |
+
[Discord](https://discord.gg/pvy7H8DZMG)
|
6 |
+
|
7 |
+
[Request more models](https://github.com/RichardErkhov/quant_request)
|
8 |
+
|
9 |
+
|
10 |
+
typhoon-7b - GGUF
|
11 |
+
- Model creator: https://huggingface.co/scb10x/
|
12 |
+
- Original model: https://huggingface.co/scb10x/typhoon-7b/
|
13 |
+
|
14 |
+
|
15 |
+
| Name | Quant method | Size |
|
16 |
+
| ---- | ---- | ---- |
|
17 |
+
| [typhoon-7b.Q2_K.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q2_K.gguf) | Q2_K | 2.55GB |
|
18 |
+
| [typhoon-7b.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.IQ3_XS.gguf) | IQ3_XS | 2.83GB |
|
19 |
+
| [typhoon-7b.IQ3_S.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.IQ3_S.gguf) | IQ3_S | 2.98GB |
|
20 |
+
| [typhoon-7b.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q3_K_S.gguf) | Q3_K_S | 2.96GB |
|
21 |
+
| [typhoon-7b.IQ3_M.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.IQ3_M.gguf) | IQ3_M | 3.07GB |
|
22 |
+
| [typhoon-7b.Q3_K.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q3_K.gguf) | Q3_K | 3.29GB |
|
23 |
+
| [typhoon-7b.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q3_K_M.gguf) | Q3_K_M | 3.29GB |
|
24 |
+
| [typhoon-7b.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q3_K_L.gguf) | Q3_K_L | 3.57GB |
|
25 |
+
| [typhoon-7b.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.IQ4_XS.gguf) | IQ4_XS | 3.69GB |
|
26 |
+
| [typhoon-7b.Q4_0.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q4_0.gguf) | Q4_0 | 3.84GB |
|
27 |
+
| [typhoon-7b.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.IQ4_NL.gguf) | IQ4_NL | 3.89GB |
|
28 |
+
| [typhoon-7b.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q4_K_S.gguf) | Q4_K_S | 3.87GB |
|
29 |
+
| [typhoon-7b.Q4_K.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q4_K.gguf) | Q4_K | 4.09GB |
|
30 |
+
| [typhoon-7b.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q4_K_M.gguf) | Q4_K_M | 4.09GB |
|
31 |
+
| [typhoon-7b.Q4_1.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q4_1.gguf) | Q4_1 | 4.26GB |
|
32 |
+
| [typhoon-7b.Q5_0.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q5_0.gguf) | Q5_0 | 4.67GB |
|
33 |
+
| [typhoon-7b.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q5_K_S.gguf) | Q5_K_S | 4.67GB |
|
34 |
+
| [typhoon-7b.Q5_K.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q5_K.gguf) | Q5_K | 4.8GB |
|
35 |
+
| [typhoon-7b.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q5_K_M.gguf) | Q5_K_M | 4.8GB |
|
36 |
+
| [typhoon-7b.Q5_1.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q5_1.gguf) | Q5_1 | 5.09GB |
|
37 |
+
| [typhoon-7b.Q6_K.gguf](https://huggingface.co/RichardErkhov/scb10x_-_typhoon-7b-gguf/blob/main/typhoon-7b.Q6_K.gguf) | Q6_K | 5.55GB |
|
38 |
+
|
39 |
+
|
40 |
+
|
41 |
+
|
42 |
+
Original model description:
|
43 |
+
|
44 |
+
---
|
45 |
+
license: apache-2.0
|
46 |
+
language:
|
47 |
+
- th
|
48 |
+
library_name: transformers
|
49 |
+
pipeline_tag: text-generation
|
50 |
+
tags:
|
51 |
+
- pretrained
|
52 |
+
---
|
53 |
+
# Typhoon-7B: Thai Large Language Model (Pretrained)
|
54 |
+
|
55 |
+
**Typhoon-7B** is a *pretrained* Thai 🇹🇠large language model with 7 billion parameters, and it is based on Mistral-7B.
|
56 |
+
|
57 |
+
**Typhoon-7B** outperforms all open-source Thai language models at the time of writing as evaluated on Thai examination benchmarks, and its instruction-tuned variant achieves the best results in instruction-following tasks. Also, its performance in Thai is on par with GPT-3.5 while being 2.62 times more efficient in tokenizing Thai text.
|
58 |
+
|
59 |
+
**This is not an instruction-tuned model** - It may not be able to follow human instructions without using one/few-shot learning or instruction fine-tuning. The model does not have any moderation mechanisms, and may generate harmful or inappropriate responses.
|
60 |
+
|
61 |
+
The Instruct model (chat-model) will be released soon. The beta version register is open at https://opentyphoon.ai/ or follow us for future model release https://twitter.com/opentyphoon.
|
62 |
+
|
63 |
+
<div align="center">
|
64 |
+
<img src="https://storage.googleapis.com/scb10x-ai-lab-public/assets/typhoon_benchmark.png" alt="Typhoon benchmark" width="100%" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
|
65 |
+
</div>
|
66 |
+
|
67 |
+
For full details of this model, please read our [paper](https://arxiv.org/abs/2312.13951).
|
68 |
+
|
69 |
+
|
70 |
+
## Model Description
|
71 |
+
- **Model type**: A 7B pretrained decoder-only model
|
72 |
+
- **Requirement**: transformers 4.34.0 or newer.
|
73 |
+
- **Primary Language(s)**: Thai 🇹🇠and English 🇬🇧
|
74 |
+
- **License**: Apache-2.0 (Commercial)
|
75 |
+
|
76 |
+
## Performance on Thai Benchmark
|
77 |
+
|
78 |
+
| **Model** | **ONET** | **IC** | **TGAT** | **TPAT-1** | **A-Level** |
|
79 |
+
|---------------------|----------|--------|----------|------------|-------------|
|
80 |
+
| Typhoon-7B | 0.379 | 0.393 | 0.700 | 0.414 | 0.324 |
|
81 |
+
| SeaLLM-7B | 0.342 | 0.256 | 0.589 | 0.336 | 0.305 |
|
82 |
+
| OpenThaiGPT-beta-7B | 0.180 | 0.278 | 0.411 | 0.319 | 0.243 |
|
83 |
+
| WangChanGLM | 0.192 | 0.271 | 0.167 | 0.172 | 0.175 |
|
84 |
+
| SEA-LION-7B | 0.179 | 0.290 | 0.244 | 0.198 | 0.175 |
|
85 |
+
| Avg. Human | 0.318 | - | 0.472 | 0.406 | - |
|
86 |
+
|
87 |
+
## Intended Uses & Limitations
|
88 |
+
|
89 |
+
This model is a pretrained base model. Thus, it may not be able to follow human instructions without using one/few-shot learning or instruction fine-tuning. The model does not have any moderation mechanisms, and may generate harmful or inappropriate responses.
|
90 |
+
|
91 |
+
## Follow us
|
92 |
+
|
93 |
+
https://twitter.com/opentyphoon
|
94 |
+
|
95 |
+
## Support / Ask any question
|
96 |
+
|
97 |
+
https://discord.gg/CqyBscMFpg
|
98 |
+
|
99 |
+
## SCB10X AI Team
|
100 |
+
|
101 |
+
- Kunat Pipatanakul, Phatrasek Jirabovonvisut, Potsawee Manakul, Sittipong Sripaisarnmongkol, Ruangsak Patomwong, Pathomporn Chokchainant, Kasima Tharnpipitchai
|
102 |
+
- If you find Typhoon-7B useful for your work, please cite it using:
|
103 |
+
```
|
104 |
+
@article{pipatanakul2023typhoon,
|
105 |
+
title={Typhoon: Thai Large Language Models},
|
106 |
+
author={Kunat Pipatanakul and Phatrasek Jirabovonvisut and Potsawee Manakul and Sittipong Sripaisarnmongkol and Ruangsak Patomwong and Pathomporn Chokchainant and Kasima Tharnpipitchai},
|
107 |
+
year={2023},
|
108 |
+
journal={arXiv preprint arXiv:2312.13951},
|
109 |
+
url={https://arxiv.org/abs/2312.13951}
|
110 |
+
}
|
111 |
+
```
|
112 |
+
|
113 |
+
## Contact Us
|
114 |
+
|
115 |
+
- General & Collaboration: kasima@scb10x.com, pathomporn@scb10x.com
|
116 |
+
- Technical: kunat@scb10x.com
|
117 |
+
|
118 |
+
|