diff --git a/README.md b/README.md
index af255a1..a0e6757 100644
--- a/README.md
+++ b/README.md
@@ -1,8 +1,8 @@
# LaWGPT：基于中文法律知识的大语言模型
-
- 
+
+ 
@@ -24,28 +24,80 @@ LaWGPT 是一系列基于中文法律知识的开源大语言模型。
## 更新
-- 🔥🔥🔥 2023/05/13：公开发布 legal-base-7b，lawgpt-7b-beta1.0
- - legal-base-7b：基座 Chinese-LLaMA-7B ，基于 50w 中文裁判文书数据二次预训练
- - lawgpt-7b-beta1.0：基于 legal-base-7b，构造 30w 高质量法律问答数据集指令精调
-- 🔥🔥🔥 2023/04/12：内部测试 lawgpt-7b-alpha
- - lawgpt-7b-alpha：基座 Chinese-LLaMA-7B ，构造 30w 法律问答数据集指令精调
+- 🌟 2023/05/13：公开发布 
+
+ - **legal-base-7b**：法律基座模型，使用 50w 中文裁判文书数据二次预训练
+
+ - **lawgpt-7b-beta1.0**：法律对话模型，构造 30w 高质量法律问答数据集指令精调
+
+- 🌟 2023/04/12：内部测试 
+ - **lawgpt-7b-alpha**：在 Chinese-LLaMA-7B 的基础上直接构造 30w 法律问答数据集指令精调
## 快速开始
-**1. 准备代码，创建环境**
+1. 准备代码，创建环境
-```bash
-git clone git@github.com:pengxiao-song/LaWGPT.git
-cd LaWGPT
-conda env create -f environment.yml
-conda activate lawgpt
-```
-**2. 下载模型权重**
+ ```bash
+ git clone git@github.com:pengxiao-song/LaWGPT.git
+ cd LaWGPT
+ conda activate lawgpt
+ pip install -r requirements.txt
+ ```
+
+2. 合并模型权重
+
+ 由于 [LLaMA](https://github.com/facebookresearch/llama) 和 [Chinese-LLaMA](https://github.com/ymcui/Chinese-LLaMA-Alpaca) 均未开源模型权重。根据相应开源许可，**本项目只能发布 LoRA 权重**，无法发布完整的模型权重，请各位谅解。
+
+ 本项目给出[合并方式](https://github.com/pengxiao-song/LaWGPT/wiki/%E6%A8%A1%E5%9E%8B%E5%90%88%E5%B9%B6)，请各位获取原版权重后自行重构模型。
+
+
+3. 启动示例
+
+ 启动本地服务：
+
+ ```bash
+ conda activate lawgpt
+ cd LaWGPT
+ sh src/scripts/generate.sh
+ ```
+
+ 接入服务链接：
+
+ ```
+ Running on local URL: http://0.0.0.0:7862
+ Running on public URL: https://06e989c08fe171f47c.gradio.live
+ ```
+
+ 呈现效果：
+
+
+ 
+
-**3. 启动示例**
## 项目结构
+```bash
+LaWGPT
+├── assets # 项目静态资源
+├── data # 语料及精调数据
+├── tools # 数据清洗等工具
+├── README.md
+├── requirements.txt
+└── src # 源码
+ ├── finetune.py
+ ├── generate.py
+ ├── models # 基座模型及 Lora 权重
+ │ ├── base_models
+ │ └── lora_weights
+ ├── outputs
+ ├── scripts # 脚本文件
+ │ ├── finetune.sh # 指令微调
+ │ └── generate.sh # 服务创建
+ ├── templates
+ └── utils
+```
+
## 数据构建
@@ -60,18 +112,22 @@ conda activate lawgpt
1. 第一阶段：扩充法律领域词表，在大规模法律文书及法典数据上预训练 Chinese-LLaMA
2. 第二阶段：构造法律领域对话问答数据集，在预训练模型基础上指令精调
+### 二次训练流程
+
+### 指令精调步骤
+
+1. 参考 `data/example_instruction.json` 构造指令微调数据集
+2. 运行 `src/scripts/finetune.sh`
+
### 计算资源
8 张 Tesla V100-SXM2-32GB
-### 训练细节
-
-
## 模型评估
-评估工作正有序开展，敬请期待。
+### 输出示例
-## 局限性
+### 局限性
由于计算资源、数据规模等因素限制，当前阶段 LawGPT 存在诸多局限性：
@@ -97,9 +153,9 @@ conda activate lawgpt
## 问题反馈
-如有问题，请于 GitHub Issue 中提交。请礼貌讨论，构建和谐交流环境。
+如有问题，请在 GitHub Issue 中提交。请礼貌讨论，构建和谐社区。
-> **协作者科研之余全力推进项目进展，由于人力有限难以实时反馈，给诸君带来不便，敬请谅解！**
+> **协作者科研之余推进项目进展，由于人力有限难以实时反馈，给诸君带来不便，敬请谅解！**
## 致谢
diff --git a/assets/demo/demo.png b/assets/demo/demo.png
new file mode 100644
index 0000000..a0e738e
Binary files /dev/null and b/assets/demo/demo.png differ
diff --git a/assets/logo/lawgpt.jpeg b/assets/logo/lawgpt.jpeg
new file mode 100644
index 0000000..a5d492c
Binary files /dev/null and b/assets/logo/lawgpt.jpeg differ
diff --git a/assets/logo/lawgpt1.png b/assets/logo/lawgpt1.png
deleted file mode 100644
index b20cc64..0000000
Binary files a/assets/logo/lawgpt1.png and /dev/null differ
diff --git a/assets/logo/lawgpt2.jpeg b/assets/logo/lawgpt2.jpeg
deleted file mode 100644
index cdb702f..0000000
Binary files a/assets/logo/lawgpt2.jpeg and /dev/null differ
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..234d81b
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,12 @@
+accelerate
+appdirs
+bitsandbytes
+black
+black[jupyter]
+datasets
+fire
+git+https://github.com/huggingface/peft.git@e536616888d51b453ed354a6f1e243fecb02ea08
+git+https://github.com/huggingface/transformers.git
+gradio
+sentencepiece
+wandb
\ No newline at end of file
diff --git a/src/utils/merge.py b/src/utils/merge.py
index 7eeaae6..85d5320 100644
--- a/src/utils/merge.py
+++ b/src/utils/merge.py
@@ -48,4 +48,4 @@ deloreanized_sd = {
if "lora" not in k
}
-LlamaForCausalLM.save_pretrained(base_model, '../models/LawGPT_step_1', state_dict=deloreanized_sd, max_shard_size="400MB")
+LlamaForCausalLM.save_pretrained(base_model, '../models/legal-base-7b', state_dict=deloreanized_sd, max_shard_size="400MB")
diff --git a/scripts/clear_law.py b/tools/clear_law.py
similarity index 100%
rename from scripts/clear_law.py
rename to tools/clear_law.py
diff --git a/scripts/generate_instructions_knowledge.py b/tools/generate_instructions_knowledge.py
similarity index 100%
rename from scripts/generate_instructions_knowledge.py
rename to tools/generate_instructions_knowledge.py
diff --git a/scripts/generate_instructions_law.py b/tools/generate_instructions_law.py
similarity index 100%
rename from scripts/generate_instructions_law.py
rename to tools/generate_instructions_law.py
diff --git a/scripts/merge_vocabulary.py b/tools/merge_vocabulary.py
similarity index 100%
rename from scripts/merge_vocabulary.py
rename to tools/merge_vocabulary.py