Commit
•
f1d706b
1
Parent(s):
2eecc60
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
- zh
|
6 |
+
pipeline_tag: text-generation
|
7 |
+
base_model: UnicomLLM/Unichat-llama3-Chinese-8B
|
8 |
+
---
|
9 |
+
|
10 |
+
# Unichat-llama3-Chinese-8B- GGUF
|
11 |
+
- This is quantized version of [UnicomLLM/Unichat-llama3-Chinese-8B](https://huggingface.co/UnicomLLM/Unichat-llama3-Chinese-8B)
|
12 |
+
|
13 |
+
# Model Description (Translated)
|
14 |
+
|
15 |
+
- China Unicom AI Innovation Center released the industry's first llama3 Chinese instruction fine-tuning model (full parameter fine-tuning), uploaded at 22:00 on April 19, 2024
|
16 |
+
- This model is based on [**Meta Llama 3**](https://huggingface.co/collections/meta-llama/meta-llama-3-66214712577ca38149ebb2b6) , adds Chinese data for training, and achieves high-quality Chinese question and answer using the llama3 model.
|
17 |
+
- The model context maintains the native length of 8K, and a version that supports 64K length will be released later.
|
18 |
+
- Base model [**Meta-Llama-3-8B**](https://huggingface.co/meta-llama/Meta-Llama-3-8B)
|
19 |
+
|
20 |
+
|
21 |
+
### 📊 Data
|
22 |
+
|
23 |
+
- High-quality instruction data, covering multiple fields and industries, providing sufficient data support for model training
|
24 |
+
- Fine-tuning instruction data undergoes strict manual screening to ensure high-quality instruction data is used for model fine-tuning.
|
25 |
+
|
26 |
+
For more details on models, datasets and training please refer to:
|
27 |
+
* Github:[**Unichat-llama3-Chinese**](https://github.com/UnicomAI/Unichat-llama3-Chinese)
|
28 |
+
|