Files changed (1) hide show
  1. README.md +28 -20
README.md CHANGED
@@ -1,3 +1,4 @@
 
1
  ---
2
  license: apache-2.0
3
  language:
@@ -7,40 +8,47 @@ pipeline_tag: text-generation
7
  tags:
8
  - pretrained
9
  ---
10
- # Model Card for Typhoon-7B
11
 
12
- **Typhoon 7B** is a pretrained Thai language adaption of Mistral-7B with 7 billion parameters.
13
 
14
- **Typhoon 7B** outperforms all open-source Thai language models as of this publishing, and its performance is on par with GPT-3.5 while being 2.62 times more efficient.
15
 
16
 
17
  <div align="center">
18
- <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'"/>
19
  </div>
20
 
21
- For full details of this model please read our [paper]() and [release blog post]().
22
-
23
-
24
- ## Requirements
25
-
26
- Transformers, 4.34.0 or newer.
27
 
28
 
29
- ## Model date
 
 
 
 
30
 
31
- **Typhoon 7B** was trained at December, 2023.
32
 
 
 
 
 
 
 
 
 
33
 
34
- ## License
35
- Apache-2.0 (Commercial)
36
 
37
-
38
- ## Notice
39
-
40
- Typhoon 7B is a pretrained base model; it cannot understand human instructions without using a few-shot or fine-tune approach on an instruct dataset,
41
- and does not have any moderation mechanisms.
42
 
43
 
44
  ## SCB10X AI Team
45
 
46
- Kunat Pipatanakul, Phatrasek Jirabovonvisut, Potsawee Manakul, Sittipong Sripaisarnmongkol, Ruangsak Patomwong, Pathomporn Chokchainant, Kasima Tharnpipitchai
 
 
 
 
 
 
1
+
2
  ---
3
  license: apache-2.0
4
  language:
 
8
  tags:
9
  - pretrained
10
  ---
11
+ # Typhoon-7B: Thai Large Language Model
12
 
13
+ **Typhoon-7B** is a *pretrained* Thai 🇹🇭 large language model with 7 billion parameters, and it is based on Mistral-7B.
14
 
15
+ **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.
16
 
17
 
18
  <div align="center">
19
+ <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'"/>
20
  </div>
21
 
22
+ For full details of this model, please read our [paper]() and [release blog post]().
 
 
 
 
 
23
 
24
 
25
+ ## Model Description
26
+ - **Model type**: A 7B pretrained decoder-only model
27
+ - **Requirement**: transformers 4.34.0 or newer.
28
+ - **Primary Language(s)**: Thai 🇹🇭 and English 🇬🇧
29
+ - **License**: Apache-2.0 (Commercial)
30
 
31
+ ## Performance on Thai Benchmark
32
 
33
+ | **Model** | **ONET** | **IC** | **TGAT** | **TPAT-1** | **A-Level** |
34
+ |---------------------|----------|--------|----------|------------|-------------|
35
+ | Typhoon-7B | 0.379 | 0.393 | 0.700 | 0.414 | 0.324 |
36
+ | SeaLLM-7B | 0.342 | 0.256 | 0.589 | 0.336 | 0.305 |
37
+ | OpenThaiGPT-beta-7B | 0.180 | 0.278 | 0.411 | 0.319 | 0.243 |
38
+ | WangChanGLM | 0.192 | 0.271 | 0.167 | 0.172 | 0.175 |
39
+ | SEA-LION-7B | 0.179 | 0.290 | 0.244 | 0.198 | 0.175 |
40
+ | Avg. Human | 0.318 | - | 0.472 | 0.406 | - |
41
 
42
+ ## Intended Uses & Limitations
 
43
 
44
+ 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.
 
 
 
 
45
 
46
 
47
  ## SCB10X AI Team
48
 
49
+ - Kunat Pipatanakul, Phatrasek Jirabovonvisut, Potsawee Manakul, Sittipong Sripaisarnmongkol, Ruangsak Patomwong, Pathomporn Chokchainant, Kasima Tharnpipitchai
50
+ - Corresponding Author: kasima@scb10x.com
51
+ - If you find Typhoon-7B useful for your work, please cite it using:
52
+ ```
53
+
54
+ ```