Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,129 @@
|
|
1 |
-
---
|
2 |
-
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- ko
|
4 |
+
- uz
|
5 |
+
- en
|
6 |
+
- ru
|
7 |
+
- zh
|
8 |
+
- ja
|
9 |
+
- km
|
10 |
+
- my
|
11 |
+
- si
|
12 |
+
- tl
|
13 |
+
- th
|
14 |
+
- vi
|
15 |
+
- kk
|
16 |
+
- bn
|
17 |
+
- mn
|
18 |
+
- id
|
19 |
+
- ne
|
20 |
+
- pt
|
21 |
+
tags:
|
22 |
+
- translation
|
23 |
+
- multilingual
|
24 |
+
- korean
|
25 |
+
- uzbek
|
26 |
+
datasets:
|
27 |
+
- custom_parallel_corpus
|
28 |
+
license: mit
|
29 |
+
---
|
30 |
+
|
31 |
+
# QWEN2.5-7B-Bnk-7e
|
32 |
+
|
33 |
+
## Model Description
|
34 |
+
|
35 |
+
QWEN2.5-7B-Bnk-5e is a multilingual translation model based on the QWEN 2.5 architecture with 7 billion parameters. It specializes in translating multiple languages to Korean and Uzbek.
|
36 |
+
|
37 |
+
## Intended Uses & Limitations
|
38 |
+
|
39 |
+
The model is designed for translating text from various Asian and European languages to Korean and Uzbek. It can be used for tasks such as:
|
40 |
+
|
41 |
+
- Multilingual document translation
|
42 |
+
- Cross-lingual information retrieval
|
43 |
+
- Language learning applications
|
44 |
+
- International communication assistance
|
45 |
+
|
46 |
+
Please note that while the model strives for accuracy, it may not always produce perfect translations, especially for idiomatic expressions or highly context-dependent content.
|
47 |
+
|
48 |
+
## Training and Evaluation Data
|
49 |
+
|
50 |
+
The model was fine-tuned on a diverse dataset of parallel texts covering the supported languages. Evaluation was performed on held-out test sets for each language pair.
|
51 |
+
|
52 |
+
## Training Procedure
|
53 |
+
|
54 |
+
Fine-tuning was performed on the QWEN 2.5 7B base model using custom datasets for the specific language pairs.
|
55 |
+
|
56 |
+
## Supported Languages
|
57 |
+
|
58 |
+
The model supports translation from the following languages to Korean and Uzbek:
|
59 |
+
|
60 |
+
- Kazakh (kk)
|
61 |
+
- Russian (ru)
|
62 |
+
- Thai (th)
|
63 |
+
- Chinese (Simplified) (zh)
|
64 |
+
- Chinese (Traditional) (zh-tw, zh-hant)
|
65 |
+
- Bengali (bn)
|
66 |
+
- Mongolian (mn)
|
67 |
+
- Indonesian (id)
|
68 |
+
- Nepali (ne)
|
69 |
+
- English (en)
|
70 |
+
- Khmer (km)
|
71 |
+
- Portuguese (pt)
|
72 |
+
- Sinhala (si)
|
73 |
+
- Korean (ko)
|
74 |
+
- Tagalog (tl)
|
75 |
+
- Burmese (my)
|
76 |
+
- Vietnamese (vi)
|
77 |
+
- Japanese (ja)
|
78 |
+
|
79 |
+
|
80 |
+
|
81 |
+
## How to Use
|
82 |
+
|
83 |
+
```python
|
84 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
85 |
+
|
86 |
+
model_name = "FINGU-AI/QWEN2.5-7B-Bnk-5e"
|
87 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
88 |
+
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
|
89 |
+
|
90 |
+
# Example usage
|
91 |
+
source_text = "Hello, how are you?"
|
92 |
+
source_lang = "en"
|
93 |
+
target_lang = "ko" # or "uz" for Uzbek
|
94 |
+
|
95 |
+
input_text = f"Translate from {source_lang} to {target_lang}: {source_text}"
|
96 |
+
input_ids = tokenizer(input_text, return_tensors="pt").input_ids
|
97 |
+
|
98 |
+
outputs = model.generate(input_ids, max_length=100)
|
99 |
+
translated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
100 |
+
print(translated_text)
|
101 |
+
```
|
102 |
+
## Performance
|
103 |
+
|
104 |
+
|
105 |
+
## Limitations
|
106 |
+
|
107 |
+
- The model's performance may vary across different language pairs and domains.
|
108 |
+
- It may struggle with very colloquial or highly specialized text.
|
109 |
+
- The model may not always capture cultural nuances or context-dependent meanings accurately.
|
110 |
+
|
111 |
+
## Ethical Considerations
|
112 |
+
|
113 |
+
- The model should not be used for generating or propagating harmful, biased, or misleading content.
|
114 |
+
- Users should be aware of potential biases in the training data that may affect translations.
|
115 |
+
- The model's outputs should not be considered as certified translations for official or legal purposes without human verification.
|
116 |
+
|
117 |
+
|
118 |
+
## Citation
|
119 |
+
|
120 |
+
|
121 |
+
```bibtex
|
122 |
+
@misc{fingu2023qwen25,
|
123 |
+
author = {FINGU AI and AI Team},
|
124 |
+
title = {QWEN2.5-7B-Bnk-7e: A Multilingual Translation Model},
|
125 |
+
year = {2024},
|
126 |
+
publisher = {Hugging Face},
|
127 |
+
journal = {Hugging Face Model Hub},
|
128 |
+
howpublished = {\url{https://huggingface.co/FINGU-AI/QWEN2.5-7B-Bnk-5e}}
|
129 |
+
}
|