From 135c4e351252c67dd2ee3867235d88e5df98dbaf Mon Sep 17 00:00:00 2001
From: hiyouga <hiyouga@buaa.edu.cn>
Date: Tue, 2 Apr 2024 20:37:37 +0800
Subject: [PATCH] update readme

Former-commit-id: 11a6c1bad65a86b0f3d9c5e5df84d246d7d368df
---
 README.md                                     | 21 ++++++---
 README_zh.md                                  | 22 ++++++----
 examples/README.md                            | 43 +++++++++++++++++++
 .../extras/fsdp_qlora/{fsdp.sh => sft.sh}     |  0
 4 files changed, 71 insertions(+), 15 deletions(-)
 create mode 100644 examples/README.md
 rename examples/extras/fsdp_qlora/{fsdp.sh => sft.sh} (100%)

diff --git a/README.md b/README.md
index 99278fbc..c374ab38 100644
--- a/README.md
+++ b/README.md
@@ -245,8 +245,6 @@ You also can add a custom chat template to [template.py](src/llmtuner/data/templ
 
 </details>
 
-Please refer to [data/README.md](data/README.md) for details.
-
 Some datasets require confirmation before using them, so we recommend logging in with your Hugging Face account using these commands.
 
 ```bash
@@ -366,8 +364,18 @@ docker compose -f ./docker-compose.yml up -d
 
 See [examples](examples) for usage.
 
-> [!TIP]
-> Use `python src/train_bash.py -h` to display arguments description.
+Use `python src/train_bash.py -h` to display arguments description.
+
+### Deploy with OpenAI-style API and vLLM
+
+```bash
+CUDA_VISIBLE_DEVICES=0 API_PORT=8000 python src/api_demo.py \
+    --model_name_or_path path_to_model \
+    --adapter_name_or_path path_to_lora_adapter \
+    --template default \
+    --finetuning_type lora \
+    --infer_backend vllm
+```
 
 ### Use ModelScope Hub
 
@@ -381,6 +389,8 @@ Train the model by specifying a model ID of the ModelScope Hub as the `--model_n
 
 ## Projects using LLaMA Factory
 
+If you have a project that should be incorporated, please contact via email or create a pull request.
+
 <details><summary>Click to show</summary>
 
 1. Wang et al. ESRL: Efficient Sampling-based Reinforcement Learning for Sequence Generation. 2023. [[arxiv]](https://arxiv.org/abs/2308.02223)
@@ -414,9 +424,6 @@ Train the model by specifying a model ID of the ModelScope Hub as the `--model_n
 
 </details>
 
-> [!TIP]
-> If you have a project that should be incorporated, please contact via email or create a pull request.
-
 ## License
 
 This repository is licensed under the [Apache-2.0 License](LICENSE).
diff --git a/README_zh.md b/README_zh.md
index d6dea04d..6f060460 100644
--- a/README_zh.md
+++ b/README_zh.md
@@ -245,8 +245,6 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 
 </details>
 
-使用方法请参考 [data/README_zh.md](data/README_zh.md) 文件。
-
 部分数据集的使用需要确认，我们推荐使用下述命令登录您的 Hugging Face 账户。
 
 ```bash
@@ -337,7 +335,6 @@ CUDA_VISIBLE_DEVICES=0 python src/train_web.py
 
 ```bash
 docker build -f ./Dockerfile -t llama-factory:latest .
-
 docker run --gpus=all \
     -v ./hf_cache:/root/.cache/huggingface/ \
     -v ./data:/app/data \
@@ -367,8 +364,18 @@ docker compose -f ./docker-compose.yml up -d
 
 使用方法请参考 [examples](examples) 文件夹。
 
-> [!TIP]
-> 使用 `python src/train_bash.py -h` 查看参数文档。
+使用 `python src/train_bash.py -h` 查看参数文档。
+
+### 使用 OpenAI 风格 API 和 vLLM 部署
+
+```bash
+CUDA_VISIBLE_DEVICES=0 API_PORT=8000 python src/api_demo.py \
+    --model_name_or_path path_to_model \
+    --adapter_name_or_path path_to_lora_adapter \
+    --template default \
+    --finetuning_type lora \
+    --infer_backend vllm
+```
 
 ### 使用魔搭社区
 
@@ -382,6 +389,8 @@ export USE_MODELSCOPE_HUB=1 # Windows 使用 `set USE_MODELSCOPE_HUB=1`
 
 ## 使用了 LLaMA Factory 的项目
 
+如果您有项目希望添加至上述列表，请通过邮件联系或者创建一个 PR。
+
 <details><summary>点击显示</summary>
 
 1. Wang et al. ESRL: Efficient Sampling-based Reinforcement Learning for Sequence Generation. 2023. [[arxiv]](https://arxiv.org/abs/2308.02223)
@@ -415,9 +424,6 @@ export USE_MODELSCOPE_HUB=1 # Windows 使用 `set USE_MODELSCOPE_HUB=1`
 
 </details>
 
-> [!TIP]
-> 如果您有项目希望添加至上述列表，请通过邮件联系或者创建一个 PR。
-
 ## 协议
 
 本仓库的代码依照 [Apache-2.0](LICENSE) 协议开源。
diff --git a/examples/README.md b/examples/README.md
new file mode 100644
index 00000000..71c4b742
--- /dev/null
+++ b/examples/README.md
@@ -0,0 +1,43 @@
+We provide diverse examples about fine-tuning LLMs.
+
+```
+examples/
+├── lora_single_gpu/
+│   ├── pt.sh: Pre-training
+│   ├── sft.sh: Supervised fine-tuning
+│   ├── reward.sh: Reward modeling
+│   ├── ppo.sh: PPO training
+│   ├── dpo.sh: DPO training
+│   ├── orpo.sh: ORPO training
+│   ├── prepare.sh: Save tokenized dataset
+│   └── predict.sh: Batch prediction
+├── qlora_single_gpu/
+│   ├── bitsandbytes.sh
+│   ├── gptq.sh
+│   ├── awq.sh
+│   └── aqlm.sh
+├── lora_multi_gpu/
+│   ├── single_node.sh
+│   └── multi_node.sh
+├── full_multi_gpu/
+│   ├── single_node.sh
+│   └── multi_node.sh
+├── merge_lora/
+│   ├── merge.sh
+│   └── quantize.sh
+├── inference/
+│   ├── cli_demo.sh
+│   ├── api_demo.sh
+│   ├── web_demo.sh
+│   └── evaluate.sh
+└── extras/
+    ├── galore/
+    │   └── sft.sh
+    ├── loraplus/
+    │   └── sft.sh
+    ├── llama_pro/
+    │   ├── expand.sh
+    │   └── sft.sh
+    └── fsdp_qlora/
+        └── sft.sh
+```
diff --git a/examples/extras/fsdp_qlora/fsdp.sh b/examples/extras/fsdp_qlora/sft.sh
similarity index 100%
rename from examples/extras/fsdp_qlora/fsdp.sh
rename to examples/extras/fsdp_qlora/sft.sh