[Docs] Update readme

README.md

@@ -1,32 +1,186 @@
 # MMChat
 
+<div align="center">
+
+[![docs](https://readthedocs.org/projects/opencompass/badge)](https://opencompass.readthedocs.io/en)
+[![license](https://img.shields.io/github/license/InternLM/opencompass.svg)](https://github.com/InternLM/opencompass/blob/main/LICENSE)
+[![PyPI](https://badge.fury.io/py/opencompass.svg)](https://pypi.org/project/opencompass/)
+
+[📘Documentation](https://opencompass.readthedocs.io/en/latest/) |
+[🤔Reporting Issues](https://github.com/InternLM/opencompass/issues/new/choose)
+
+English | [简体中文](README_zh-CN.md)
+
+</div>
+
+## 🌟 Demos
+
+- QLoRA fine-tune for InternLM-7B [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1yzGeYXayLomNQjLD4vC6wgUHvei3ezt4?usp=sharing)
+- Chat with Llama2-7B-Plugins [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](<>)
+- Use MMChat in HuggingFace training pipeline [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1eBI9yiOkX-t7P-0-t9vS8y1x5KmWrkoU?usp=sharing)
+
+## 🧭 Introduction
+
+MMChat is a toolkit for quickly fine-tuning LLM, developed by the [MMRazor](https://github.com/open-mmlab/mmrazor) and [MMDeploy](https://github.com/open-mmlab/mmdeploy) teams. It has the following core features:
+
+- Embrace [HuggingFace](https://huggingface.co) and provide fast support for new models, datasets, and algorithms.
+- Provide a comprehensive solution and related models for [MOSS plugins datasets](https://github.com/OpenLMLab/MOSS/tree/main/SFT_data).
+- Support arbitrary combinations of multiple datasets during fine-tuning.
+- Compatible with [DeepSpeed](https://github.com/microsoft/DeepSpeed), enabling the efficient fine-tuning of LLM on multiple GPUs.
+- Support [QLoRA](http://arxiv.org/abs/2305.14314), enabling the efficient fine-tuning of LLM using free resources on Colab.
+
+> 💥 [MMRazor](https://github.com/open-mmlab/mmrazor) and [MMDeploy](https://github.com/open-mmlab/mmdeploy) teams have also collaborated in developing [LMDeploy](https://github.com/InternLM/lmdeploy), a toolkit for for compressing, deploying, and serving LLM. Welcome to subscribe to stay updated with our latest developments.
+
+## 🔥 Supports
+
+<table>
+<tbody>
+<tr align="center" valign="middle">
+<td>
+  <b>Models</b>
+</td>
+<td>
+  <b>Datasets</b>
+</td>
+<td>
+  <b>Strategies</b>
+</td>
+ <td>
+  <b>Algorithms</b>
+</td>
+</tr>
+<tr valign="top">
+<td align="left" valign="top">
+<ul>
+  <li><a href="configs/internlm/internlm_7b">InternLM</a></li>
+  <li><a href="configs/internlm/internlm_chat_7b">InternLM-Chat</a></li>
+  <li><a href="configs/llama/llama_7b">Llama</a></li>
+  <li><a href="configs/llama/llama2_7b">Llama2</a></li>
+  <li><a href="configs/llama/llama2_7b_chat">Llama2-Chat</a></li>
+  <li><a href="configs/qwen/qwen_7b">Qwen</a></li>
+  <li><a href="configs/qwen/qwen_7b_chat">Qwen-Chat</a></li>
+  <li><a href="configs/baichuan/baichuan_7b">Baichuan-7B</a></li>
+  <li><a href="configs/baichuan/baichuan_13b_base">Baichuan-13B-Base</a></li>
+  <li><a href="configs/baichuan/baichuan_13b_chat">Baichuan-13B-Chat</a></li>
+  <li>...</li>
+</ul>
+</td>
+<td>
+<ul>
+  <li><a href="configs/_base_/datasets/moss_003_sft_all.py">MOSS-003-SFT</a></li>
+  <li><a href="configs/_base_/datasets/arxiv.py">Arxiv GenTitle</a></li>
+  <li><a href="configs/_base_/datasets/open_orca.py">OpenOrca</a></li>
+  <li><a href="configs/_base_/datasets/alpaca.py">Alpaca en</a> / <a href="configs/_base_/datasets/alpaca_zh.py">zh</a></li>
+  <li><a href="configs/_base_/datasets/oasst1.py">oasst1</a></li>
+  <li><a href="configs/_base_/datasets/cmd.py">Chinese Medical Dialogue</a></li>
+  <li>...</li>
+</ul>
+</td>
+<td>
+<ul>
+  <li>(Distributed) Data Parallel</li>
+  <li><a href="https://github.com/microsoft/DeepSpeed">Deepspeed</a> 🚀</li>
+</ul>
+</td>
+<td>
+<ul>
+  <li>Full parameter fine-tune</li>
+  <li><a href="http://arxiv.org/abs/2106.09685">LoRA</a></li>
+  <li><a href="http://arxiv.org/abs/2305.14314">QLoRA</a></li>
+  <li>...</li>
+</ul>
+</td>
+</tr>
+</tbody>
+</table>
+
+## 🛠️ Quick Start
+
 ### Installation
 
-1. 安装 Pytorch (https://pytorch.org/get-started/locally/)
-2. 安装 MMChat
+Below are quick steps for installation:
 
+```shell
+conda create -n mmchat python=3.10
+conda activate mmchat
+git clone XXX
+cd MMChat
+pip install -v -e .
 ```
-pip install -r requirements.txt
-pip install -e .
-```
 
-### TODO
+### Chat [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](<>)
+
+<img src="https://github.com/InternLM/lmdeploy/assets/36994684/20159556-7711-4b0d-9568-6884998ad66a">
+
+MMChat provides the tools to chat with pretrained / fine-tuned LLMs.
+
+- For example, we can start the chat with Llama2-7B-Plugins by
+
+  ```shell
+  python ./tools/chat_hf.py meta-llama/Llama-2-7b --adapter XXX --bot-name Llama2 --prompt plugins --with-plugins --command-stop-word "<eoc>" --answer-stop-word "<eom>" --no-streamer
+  ```
+
+For more usages, please see [TODO](<>).
+
+### Fine-tune [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1yzGeYXayLomNQjLD4vC6wgUHvei3ezt4?usp=sharing)
+
+MMChat supports the efficient fine-tune (*e.g.*, QLoRA) for Large Language Models (LLM).
+
+Taking the QLoRA fine-tuning  as an example, we can start it by
+
+- For example, we can start the QLoRA fine-tuning of InternLM-7B with oasst1 dataset by
+  ```shell
+  # On a single GPU
+  python ./tools/train.py ./configs/internlm/internlm_7b/internlm_7b_qlora_oasst1.py
+  # On multiple GPUs
+  bash ./tools/dist_train.sh ./configs/internlm/internlm_7b/internlm_7b_qlora_oasst1.py ${GPU_NUM}
+  ```
+
+For more usages, please see [TODO](<>).
+
+### Deploy
+
+- **Step 0**, convert the pth adapter to HuggingFace adapter, by
+
+  ```shell
+  python ./tools/model_converters/adapter_pth2hf.py \
+  		${CONFIG_FILE} \
+  		${PATH_TO_PTH_ADAPTER} \
+  		${SAVE_PATH_TO_HF_ADAPTER}
+  ```
+
+- **Step 1**, merge the HuggingFace adapter to the pretrained LLM, by
+
+  ```shell
+  python ./tools/model_converters/merge_lora_hf.py \
+      ${NAME_OR_PATH_TO_HF_MODEL} \
+      ${NAME_OR_PATH_TO_HF_ADAPTER} \
+      ${SAVE_PATH}
+  ```
+
+- **Step 2**, deploy the merged LLM with any other framework, such as [LMDeploy](https://github.com/InternLM/lmdeploy) 🚀.
+
+  ```shell
+  pip install lmdeploy
+  python -m lmdeploy.pytorch.chat ${NAME_OR_PATH_TO_HF_MODEL} \
+      --max_new_tokens 256 \
+      --temperture 0.8 \
+      --top_p 0.95 \
+      --seed 0
+  ```
+
+  🎯 We are woking closely with [LMDeploy](https://github.com/InternLM/lmdeploy), to implement the deployment of **dialogues with plugins**!
+
+### Evaluation
+
+- We recommend using [OpenCompass](https://github.com/InternLM/opencompass),  a comprehensive and systematic LLM evaluation library, which currently supports 50+ datasets with about 300,000 questions.
 
-##### 基础功能
+## 🔜 Roadmap
 
-- [x] MMEngine 加载 HF Dataset
-- [x] MMEngine 加载 HF Model & Tokenizer
-- [ ] DeepSpeed Config
-- [ ] FSDP Config
-- [ ] MMLU 评测
-- [ ] PTH -> HF Checkpoint 转换脚本
-- [ ] Attention Score 可视化
+## 🎖️ Acknowledgement
 
-##### 算法
+- [Llama 2](https://github.com/facebookresearch/llama)
+- [QLoRA](http://arxiv.org/abs/2305.14314)
+- [LMDeploy](https://github.com/InternLM/lmdeploy)
 
-- [x] SFT
-- [ ] Alpaca SFT Setting & 精度对齐
-- [x] QLora SFT
-- [ ] QLora SFT Setting & 精度对齐
-- [ ] QLora RLHF
-- [ ] Distill Finetune
+## 🎫 License