kuotient commited on
Commit
e644254
โ€ข
1 Parent(s): c64335a

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +86 -0
README.md ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-nc-sa-4.0
3
+ datasets:
4
+ - squarelike/sharegpt_deepl_ko_translation
5
+ language:
6
+ - ko
7
+ pipeline_tag: translation
8
+ tags:
9
+ - translate
10
+ ---
11
+ # **Seagull-13b-translation-AWQ ๐Ÿ“‡**
12
+ ![Seagull-typewriter](./Seagull-typewriter-pixelated.png)
13
+ ## This is quantized version of original model: Seagull-13b-translation.
14
+ **Seagull-13b-translation** is yet another translator model, but carefully considered the following issues from existing translation models.
15
+ - `newline` or `space` not matching the original text
16
+ - Using translated dataset with first letter removed for training
17
+ - Codes
18
+ - Markdown format
19
+ - LaTeX format
20
+ - etc
21
+
22
+ ์ด๋Ÿฐ ์ด์Šˆ๋“ค์„ ์ถฉ๋ถ„ํžˆ ์ฒดํฌํ•˜๊ณ  ํ•™์Šต์„ ์ง„ํ–‰ํ•˜์˜€์ง€๋งŒ, ๋ชจ๋ธ์„ ์‚ฌ์šฉํ•  ๋•Œ๋Š” ์ด๋Ÿฐ ๋ถ€๋ถ„์— ๋Œ€ํ•œ ๊ฒฐ๊ณผ๋ฅผ ๋ฉด๋ฐ€ํ•˜๊ฒŒ ์‚ดํŽด๋ณด๋Š” ๊ฒƒ์„ ์ถ”์ฒœํ•ฉ๋‹ˆ๋‹ค(์ฝ”๋“œ๊ฐ€ ํฌํ•จ๋œ ํ…์ŠคํŠธ ๋“ฑ).
23
+
24
+ > If you're interested in building large-scale language models to solve a wide variety of problems in a wide variety of domains, you should consider joining [Allganize](https://allganize.career.greetinghr.com/o/65146).
25
+ For a coffee chat or if you have any questions, please do not hesitate to contact me as well! - [email protected]
26
+
27
+ This model was created as a personal experiment, unrelated to the organization I work for.
28
+
29
+ # **License**
30
+ ## From original model author:
31
+ - Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Public License, under LLAMA 2 COMMUNITY LICENSE AGREEMENT
32
+ - Full License available at: https://huggingface.co/beomi/llama-2-koen-13b/blob/main/LICENSE
33
+
34
+ # **Model Details**
35
+ #### **Developed by**
36
+ Jisoo Kim(kuotient)
37
+ #### **Base Model**
38
+ [beomi/llama-2-koen-13b](https://huggingface.co/beomi/llama-2-koen-13b)
39
+ #### **Datasets**
40
+ - [sharegpt_deepl_ko_translation](https://huggingface.co/datasets/squarelike/sharegpt_deepl_ko_translation)
41
+ - AIHUB
42
+ - ๊ธฐ์ˆ ๊ณผํ•™ ๋ถ„์•ผ ํ•œ-์˜ ๋ฒˆ์—ญ ๋ณ‘๋ ฌ ๋ง๋ญ‰์น˜ ๋ฐ์ดํ„ฐ
43
+ - ์ผ์ƒ์ƒํ™œ ๋ฐ ๊ตฌ์–ด์ฒด ํ•œ-์˜ ๋ฒˆ์—ญ ๋ณ‘๋ ฌ ๋ง๋ญ‰์น˜ ๋ฐ์ดํ„ฐ
44
+
45
+ ## Usage
46
+ #### Format
47
+ It follows only **ChatML** format.
48
+
49
+ ```python
50
+ <|im_start|>system
51
+ ์ฃผ์–ด์ง„ ๋ฌธ์žฅ์„ ํ•œ๊ตญ์–ด๋กœ ๋ฒˆ์—ญํ•˜์„ธ์š”.<|im_end|>
52
+ <|im_start|>user
53
+ {instruction}<|im_end|>
54
+ <|im_start|>assistant
55
+ # Don't miss newline here
56
+ ```
57
+ ```python
58
+ <|im_start|>system
59
+ ์ฃผ์–ด์ง„ ๋ฌธ์žฅ์„ ์˜์–ด๋กœ ๋ฒˆ์—ญํ•˜์„ธ์š”.<|im_end|>
60
+ <|im_start|>user
61
+ {instruction}<|im_end|>
62
+ <|im_start|>assistant
63
+ # Don't miss newline here
64
+ ```
65
+
66
+ #### Example
67
+ **I highly recommend to use vllm. I will write a guide for quick and easy inference if requested.**
68
+
69
+ Since, chat_template already contains insturction format above.
70
+ You can use the code below.
71
+ ```python
72
+ from transformers import AutoModelForCausalLM, AutoTokenizer
73
+ device = "cuda" # the device to load the model onto
74
+ model = AutoModelForCausalLM.from_pretrained("kuotient/Seagull-13B-translation")
75
+ tokenizer = AutoTokenizer.from_pretrained("kuotient/Seagull-13B-translation")
76
+ messages = [
77
+ {"role": "user", "content": "๋ฐ”๋‚˜๋‚˜๋Š” ์›๋ž˜ ํ•˜์–€์ƒ‰์ด์•ผ?"},
78
+ ]
79
+ encodeds = tokenizer.apply_chat_template(messages, return_tensors="pt")
80
+
81
+ model_inputs = encodeds.to(device)
82
+ model.to(device)
83
+ generated_ids = model.generate(model_inputs, max_new_tokens=1000, do_sample=True)
84
+ decoded = tokenizer.batch_decode(generated_ids)
85
+ print(decoded[0])
86
+ ```