Update README.md
Browse files
README.md
CHANGED
|
@@ -3,23 +3,41 @@ language:
|
|
| 3 |
- zh
|
| 4 |
- en
|
| 5 |
license: apache-2.0
|
| 6 |
-
base_model: Qwen/Qwen2.5-0.5B
|
| 7 |
tags:
|
| 8 |
-
-
|
|
|
|
| 9 |
- lora
|
| 10 |
- fine-tuned
|
| 11 |
-
- dialogue
|
| 12 |
---
|
| 13 |
|
| 14 |
-
#
|
| 15 |
|
| 16 |
-
|
| 17 |
|
| 18 |
## 模型描述
|
| 19 |
-
|
| 20 |
- **基座模型**: Qwen2.5-0.5B
|
| 21 |
-
- **微调方法**: LoRA
|
| 22 |
- **训练数据**: 自定义对话数据集
|
| 23 |
-
- **适用任务**: 开放域对话
|
| 24 |
|
| 25 |
## 使用方法
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
- zh
|
| 4 |
- en
|
| 5 |
license: apache-2.0
|
| 6 |
+
base_model: Qwen/Qwen2.5-0.5B # 必须声明基座模型!
|
| 7 |
tags:
|
| 8 |
+
- text-generation
|
| 9 |
+
- dialogue
|
| 10 |
- lora
|
| 11 |
- fine-tuned
|
|
|
|
| 12 |
---
|
| 13 |
|
| 14 |
+
# test_1023
|
| 15 |
|
| 16 |
+
基于 [Qwen2.5-0.5B](https://huggingface.co/Qwen/Qwen2.5-0.5B) 使用LoRA微调的对话生成模型。
|
| 17 |
|
| 18 |
## 模型描述
|
|
|
|
| 19 |
- **基座模型**: Qwen2.5-0.5B
|
| 20 |
+
- **微调方法**: LoRA (rank=64)
|
| 21 |
- **训练数据**: 自定义对话数据集
|
| 22 |
+
- **适用任务**: 开放域对话、客服问答等
|
| 23 |
|
| 24 |
## 使用方法
|
| 25 |
+
python
|
| 26 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 27 |
+
model = AutoModelForCausalLM.from_pretrained("thdddd/test_1023")
|
| 28 |
+
tokenizer = AutoTokenizer.from_pretrained("thdddd/test_1023")
|
| 29 |
+
input_text = "你好,能推荐北京的美食吗?"
|
| 30 |
+
inputs = tokenizer(input_text, return_tensors="pt")
|
| 31 |
+
outputs = model.generate(**inputs, max_length=100)
|
| 32 |
+
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
| 33 |
+
|
| 34 |
+
## 训练参数
|
| 35 |
+
| 参数 | 值 |
|
| 36 |
+
|---------------|---------|
|
| 37 |
+
| 学习率 | 2e-5 |
|
| 38 |
+
| 训练轮数 | 3 |
|
| 39 |
+
| LoRA Rank | 64 |
|
| 40 |
+
| 批大小 | 8 |
|
| 41 |
+
|
| 42 |
+
## 许可证
|
| 43 |
+
Apache-2.0(与基座模型保持一致)
|