Update README.md
Browse files
README.md
CHANGED
@@ -58,3 +58,61 @@ pipeline_tag: text-generation
|
|
58 |
</code></pre>
|
59 |
|
60 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
</code></pre>
|
59 |
|
60 |
|
61 |
+
Here’s the English version of the provided text:
|
62 |
+
|
63 |
+
---
|
64 |
+
|
65 |
+
## ❶ Model Description
|
66 |
+
|
67 |
+
**Model Name and Key Features**:
|
68 |
+
KTDSbaseLM v0.11 is based on the OpenChat 3.5 model, fine-tuned using the SFT method on the Mistral 7B model.
|
69 |
+
It is designed to understand Korean and various cultural contexts, utilizing data from 135 domains in Korean society.
|
70 |
+
The model supports tasks such as text generation, conversation inference, document summarization,
|
71 |
+
question answering, sentiment analysis, and other NLP tasks.
|
72 |
+
Its applications span fields like law, finance, science, education, business, and cultural research.
|
73 |
+
|
74 |
+
**Model Architecture**:
|
75 |
+
KTDSBaseLM v0.11 is a high-performance language model with 7 billion parameters based on the Mistral 7B model.
|
76 |
+
It uses OpenChat 3.5 as the foundation and is fine-tuned using SFT to excel in Korean language and culture.
|
77 |
+
The streamlined Mistral 7B architecture ensures fast inference and memory efficiency,
|
78 |
+
optimized for various NLP tasks like text generation, question answering, document summarization, and sentiment analysis.
|
79 |
+
|
80 |
+
---
|
81 |
+
|
82 |
+
## ❷ Training Data
|
83 |
+
|
84 |
+
KTDSbaseLM v0.11 was trained on 3.6GB of data, comprising 2.33 million Q&A instances.
|
85 |
+
This includes 1.33 million multiple-choice questions across 135 domains such as history,
|
86 |
+
finance, law, tax, and science, trained with the Chain of Thought method. Additionally,
|
87 |
+
1.3 million short-answer questions cover 100 domains including history, finance, and law.
|
88 |
+
|
89 |
+
**Training Instruction Dataset Format**:
|
90 |
+
`{"prompt": "prompt text", "completion": "ideal generated text"}`
|
91 |
+
|
92 |
+
---
|
93 |
+
|
94 |
+
## ❸ Use Cases
|
95 |
+
|
96 |
+
KTDSbaseLM v0.11 can be used across multiple fields, such as:
|
97 |
+
|
98 |
+
- **Education**: Answering questions and generating explanations for subjects like history, math, and science.
|
99 |
+
- **Business**: Providing responses and summaries for legal, financial, and tax-related queries.
|
100 |
+
- **Research and Culture**: Performing NLP tasks, sentiment analysis, document generation, and translation.
|
101 |
+
- **Customer Service**: Generating conversations and personalized responses for users.
|
102 |
+
|
103 |
+
This model is highly versatile in various NLP tasks.
|
104 |
+
|
105 |
+
---
|
106 |
+
|
107 |
+
## ❹ Limitations
|
108 |
+
|
109 |
+
KTDSBaseLM v0.11 is specialized in Korean language and culture.
|
110 |
+
However, it may lack accuracy in responding to topics outside its scope,
|
111 |
+
such as international or specialized data.
|
112 |
+
Additionally, it may have limited reasoning ability for complex logical problems and
|
113 |
+
may produce biased responses if trained on biased data.
|
114 |
+
|
115 |
+
---
|
116 |
+
|
117 |
+
## ❺ Usage Instructions
|
118 |
+
|