Update README.md
Browse files
README.md
CHANGED
|
@@ -48,6 +48,11 @@ The specific release versions and download links are listed in the table below:
|
|
| 48 |
| 13B | [Baichuan2-13B-Base](https://huggingface.co/baichuan-inc/Baichuan2-13B-Base) | [Baichuan2-13B-Chat](https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat) | [Baichuan2-13B-Chat-4bits](https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat-4bits) |
|
| 49 |
|
| 50 |
# <span id="Start">快速开始/Quick Start</span>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
```python
|
| 52 |
import torch
|
| 53 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
|
|
| 48 |
| 13B | [Baichuan2-13B-Base](https://huggingface.co/baichuan-inc/Baichuan2-13B-Base) | [Baichuan2-13B-Chat](https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat) | [Baichuan2-13B-Chat-4bits](https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat-4bits) |
|
| 49 |
|
| 50 |
# <span id="Start">快速开始/Quick Start</span>
|
| 51 |
+
|
| 52 |
+
在Baichuan2系列模型中，我们为了加快推理速度使用了Pytorch2.0加入的新功能F.scaled_dot_product_attention，因此模型需要在Pytorch2.0环境下运行。
|
| 53 |
+
|
| 54 |
+
In the Baichuan 2 series models, we have utilized the new feature `F.scaled_dot_product_attention` introduced in PyTorch 2.0 to accelerate inference speed. Therefore, the model needs to be run in a PyTorch 2.0 environment.
|
| 55 |
+
|
| 56 |
```python
|
| 57 |
import torch
|
| 58 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|