mirror of https://github.com/InternLM/InternLM
update readme.md
YWMditto
parent
7efd96502a
commit
a3946235b2
|
|
@ -133,6 +133,12 @@ streamlit run web_demo.py
|
|||
|
||||
|
||||
|
||||
`web_demo_internlm.py` を使用して、InternLM 形式のモデルと直接対話できるようになりました。
|
||||
まず、モデルの重みを InternLM 形式でダウンロードし、`web_demo_internlm.py` の `ckpt_dir` を置き換えてください。 次のコマンドを実行して対話します。
|
||||
```python
|
||||
torchrun --master_port 12331 --nnodes=1 --node_rank=0 --nproc_per_node=1 -m streamlit run web_demo_internlm.py
|
||||
```
|
||||
|
||||
### デプロイ
|
||||
|
||||
[LMDeploy](https://github.com/InternLM/LMDeploy) を使って、InternLM をワンクリックでデプロイする。
|
||||
|
|
|
|||
|
|
@ -44,7 +44,7 @@ InternLM 是一个开源的轻量级训练框架，旨在支持大模型训练
|
|||
|
||||
## 更新
|
||||
|
||||
[20230920] InternLM-20B 已发布，包括基础版和对话版。
|
||||
[20230920] InternLM-20B 已发布，包括基础版和对话版。
|
||||
[20230822] InternLM-7B-Chat v1.1 已发布，增加了代码解释器和函数调用能力。您可以使用 [Lagent](https://github.com/InternLM/lagent) 进行尝试。
|
||||
|
||||
|
||||
|
|
@ -62,13 +62,13 @@ InternLM 是一个开源的轻量级训练框架，旨在支持大模型训练
|
|||
| **InternLM Chat 7B 8k** | [🤗internlm/internlm-chat-7b-8k](https://huggingface.co/internlm/internlm-chat-7b-8k) | [<img src="./doc/imgs/modelscope_logo.png" width="20px" /> Shanghai_AI_Laboratory/internlm-chat-7b-8k](https://modelscope.cn/models/Shanghai_AI_Laboratory/internlm-chat-7b-8k/summary) | [](https://openxlab.org.cn/models/detail/OpenLMLab/InternLM-chat-7b-8k) | 2023-07-06 |
|
||||
|
||||
|
||||
<details>
|
||||
<details>
|
||||
<summary> InternLM-20B </summary>
|
||||
|
||||
#### 简介
|
||||
InternLM-20B 在超过 **2.3T** Tokens 包含高质量英文、中文和代码的数据上进行预训练，其中 Chat 版本还经过了 SFT 和 RLHF 训练，使其能够更好、更安全地满足用户的需求。
|
||||
InternLM-20B 在超过 **2.3T** Tokens 包含高质量英文、中文和代码的数据上进行预训练，其中 Chat 版本还经过了 SFT 和 RLHF 训练，使其能够更好、更安全地满足用户的需求。
|
||||
|
||||
InternLM 20B 在模型结构上选择了深结构，InternLM-20B 的层数设定为60层，超过常规7B和13B模型所使用的32层或者40层。在参数受限的情况下，提高层数有利于提高模型的综合能力。此外，相较于InternLM-7B，InternLM-20B使用的预训练数据经过了更高质量的清洗，并补充了高知识密度和用于强化理解和推理能力的训练数据。因此，它在理解能力、推理能力、数学能力、编程能力等考验语言模型技术水平的方面都得到了显著提升。总体而言，InternLM-20B具有以下的特点：
|
||||
InternLM 20B 在模型结构上选择了深结构，InternLM-20B 的层数设定为60层，超过常规7B和13B模型所使用的32层或者40层。在参数受限的情况下，提高层数有利于提高模型的综合能力。此外，相较于InternLM-7B，InternLM-20B使用的预训练数据经过了更高质量的清洗，并补充了高知识密度和用于强化理解和推理能力的训练数据。因此，它在理解能力、推理能力、数学能力、编程能力等考验语言模型技术水平的方面都得到了显著提升。总体而言，InternLM-20B具有以下的特点：
|
||||
- 优异的综合性能
|
||||
- 很强的工具调用功能
|
||||
- 支持16k语境长度（通过推理时外推）
|
||||
|
|
@ -117,7 +117,7 @@ InternLM 20B 在模型结构上选择了深结构，InternLM-20B 的层数设定
|
|||
</details>
|
||||
|
||||
|
||||
<details>
|
||||
<details>
|
||||
<summary> InternLM-7B </summary>
|
||||
|
||||
#### 模型更新
|
||||
|
|
@ -194,7 +194,7 @@ for response, history in model.stream_chat(tokenizer, "你好", history=[]):
|
|||
length = len(response)
|
||||
```
|
||||
|
||||
### 通过 ModelScope 加载
|
||||
### 通过 ModelScope 加载
|
||||
|
||||
通过以下的代码从 ModelScope 加载 InternLM 模型 （可修改模型名称替换不同的模型）
|
||||
|
||||
|
|
@ -226,6 +226,12 @@ streamlit run web_demo.py
|
|||
|
||||
|
||||
|
||||
现在您可以使用 `web_demo_internlm.py` 直接与 InternLM 格式的模型进行交互。
|
||||
首先请下载 InternLM 格式的模型权重，然后替换 `web_demo_internlm.py` 中的 `ckpt_dir`。运行以下命令进行交互：
|
||||
````bash
|
||||
torchrun --master_port 12331 --nnodes=1 --node_rank=0 --nproc_per_node=1 -m streamlit run web_demo_internlm.py
|
||||
````
|
||||
|
||||
### 基于InternLM高性能部署
|
||||
|
||||
我们使用 [LMDeploy](https://github.com/InternLM/LMDeploy) 完成 InternLM 的一键部署。
|
||||
|
|
|
|||
13
README.md
13
README.md
|
|
@ -44,13 +44,13 @@ Based on the InternLM training framework, we have released two open-sourced pret
|
|||
|
||||
## News
|
||||
|
||||
[20230920] InternLM-20B is released with base and chat versions.
|
||||
[20230920] InternLM-20B is released with base and chat versions.
|
||||
[20230822] InternLM-7B-Chat v1.1 is released with code interpreter and function calling capability. You can try it with [Lagent](https://github.com/InternLM/lagent).
|
||||
|
||||
|
||||
## Model Zoo
|
||||
|
||||
Our models are released in three platforms: Transformers, ModelScope and OpenXLab.
|
||||
Our models are released in three platforms: Transformers, ModelScope and OpenXLab.
|
||||
|
||||
| Model | Transformers | ModelScope | OpenXLab | Release Date |
|
||||
|---------------------------|------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------|
|
||||
|
|
@ -113,7 +113,7 @@ Overall, InternLM-20B comprehensively outperforms open-source models in the 13B
|
|||
</details>
|
||||
|
||||
|
||||
<details>
|
||||
<details>
|
||||
<summary> InternLM-7B </summary>
|
||||
|
||||
#### News
|
||||
|
|
@ -222,6 +222,13 @@ The effect is as follows
|
|||
|
||||
|
||||
|
||||
Now you can interact with models directly in InternLM format using `web_demo_internlm.py`.
|
||||
First, please download the model weights in InternLM format, and then replace `ckpt_dir` in `web_demo_internlm.py`. Start by running the following command:
|
||||
```bash
|
||||
torchrun --master_port 12331 --nnodes=1 --node_rank=0 --nproc_per_node=1 -m streamlit run web_demo_internlm.py
|
||||
```
|
||||
|
||||
|
||||
### Deployment
|
||||
|
||||
We use [LMDeploy](https://github.com/InternLM/LMDeploy) to complete the one-click deployment of InternLM.
|
||||
|
|
|
|||
|
|
@ -53,9 +53,9 @@ system_meta_instruction = (
|
|||
- InternLM (书生·浦语) can understand and communicate fluently in the language chosen by the user such as English and 中文.
|
||||
"""
|
||||
)
|
||||
user_prompt = "<|User|>:{user}<eoh>\n"
|
||||
user_prompt = "<|User|>:{user}\n"
|
||||
robot_prompt = "<|Bot|>:{robot}<eoa>\n"
|
||||
cur_query_prompt = "<|User|>:{user}<eoh>\n<|Bot|>:"
|
||||
cur_query_prompt = "<|User|>:{user}\n<|Bot|>:"
|
||||
|
||||
|
||||
def combine_history(prompt):
|
||||
|
|
|
|||
|
|
@ -92,9 +92,9 @@ system_meta_instruction = (
|
|||
- InternLM (书生·浦语) can understand and communicate fluently in the language chosen by the user such as English and 中文.
|
||||
"""
|
||||
)
|
||||
user_prompt = "<|User|>:{user}<eoh>\n"
|
||||
user_prompt = "<|User|>:{user}\n"
|
||||
robot_prompt = "<|Bot|>:{robot}<eoa>\n"
|
||||
cur_query_prompt = "<|User|>:{user}<eoh>\n<|Bot|>:"
|
||||
cur_query_prompt = "<|User|>:{user}\n<|Bot|>:"
|
||||
|
||||
|
||||
def combine_history(prompt):
|
||||
|
|
|
|||
Loading…
Reference in New Issue