From aac7f578e0a9e64129abb4c9d06659bb04e7eb19 Mon Sep 17 00:00:00 2001 From: whcao <41630003+HIT-cwh@users.noreply.github.com> Date: Mon, 29 Apr 2024 19:39:21 +0800 Subject: [PATCH] [Docs] Delete colab and add speed benchmark (#617) * delete colab and add speed benchmark * change speed benchmark figures * fix en readme --- README.md | 49 +++++++++++++++++++------------------------------ README_zh-CN.md | 49 +++++++++++++++++++------------------------------ 2 files changed, 38 insertions(+), 60 deletions(-) diff --git a/README.md b/README.md index f92acf36f..e0be695ef 100644 --- a/README.md +++ b/README.md @@ -23,6 +23,20 @@ English | [简体中文](README_zh-CN.md) +## 🚀 Speed Benchmark + +- Llama2 7B Training Speed + +
+ +
+ +- Llama2 70B Training Speed + +
+ +
+ ## 🎉 News - **\[2024/04\]** [LLaVA-Phi-3-mini](https://huggingface.co/xtuner/llava-phi-3-mini-hf) is released! Click [here](xtuner/configs/llava/phi3_mini_4k_instruct_clip_vit_large_p14_336) for details! @@ -65,31 +79,6 @@ XTuner is an efficient, flexible and full-featured toolkit for fine-tuning large - Support chatting with large models with pre-defined templates. - The output models can seamlessly integrate with deployment and server toolkit ([LMDeploy](https://github.com/InternLM/lmdeploy)), and large-scale evaluation toolkit ([OpenCompass](https://github.com/open-compass/opencompass), [VLMEvalKit](https://github.com/open-compass/VLMEvalKit)). -## 🌟 Demos - -- Ready-to-use models and datasets from XTuner API [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/17CSO7T8q6KePuvu684IiHl6_id-CjPjh?usp=sharing) - -- QLoRA Fine-tune [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1QAEZVBfQ7LZURkMUtaq0b-5nEQII9G9Z?usp=sharing) - -- Plugin-based Chat [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/144OuTVyT_GvFyDMtlSlTzcxYIfnRsklq?usp=sharing) - - - - - - - - - - -
Examples of Plugin-based Chat 🔥🔥🔥
- - - - - -
- ## 🔥 Supports @@ -112,13 +101,12 @@ XTuner is an efficient, flexible and full-featured toolkit for fine-tuning large
  • InternLM2
    • -
  • InternLM
    • -
  • Llama
    • +
  • Llama 3
  • Llama 2
    • +
  • Phi-3
  • ChatGLM2
  • ChatGLM3
  • Qwen
    • -
  • Baichuan
  • Baichuan2
  • Mixtral 8x7B
  • DeepSeek MoE
    • @@ -192,7 +180,7 @@ XTuner is an efficient, flexible and full-featured toolkit for fine-tuning large pip install -e '.[all]' ``` -### Fine-tune [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1QAEZVBfQ7LZURkMUtaq0b-5nEQII9G9Z?usp=sharing) +### Fine-tune XTuner supports the efficient fine-tune (*e.g.*, QLoRA) for LLMs. Dataset prepare guides can be found on [dataset_prepare.md](./docs/en/user_guides/dataset_prepare.md). @@ -235,7 +223,7 @@ XTuner supports the efficient fine-tune (*e.g.*, QLoRA) for LLMs. Dataset prepar xtuner convert pth_to_hf ${CONFIG_NAME_OR_PATH} ${PTH} ${SAVE_PATH} ``` -### Chat [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/144OuTVyT_GvFyDMtlSlTzcxYIfnRsklq?usp=sharing) +### Chat XTuner provides tools to chat with pretrained / fine-tuned LLMs. @@ -295,6 +283,7 @@ We appreciate all contributions to XTuner. Please refer to [CONTRIBUTING.md](.gi ## 🎖️ Acknowledgement - [Llama 2](https://github.com/facebookresearch/llama) +- [DeepSpeed](https://github.com/microsoft/DeepSpeed) - [QLoRA](https://github.com/artidoro/qlora) - [LMDeploy](https://github.com/InternLM/lmdeploy) - [LLaVA](https://github.com/haotian-liu/LLaVA) diff --git a/README_zh-CN.md b/README_zh-CN.md index dcc6649ff..c5037d28c 100644 --- a/README_zh-CN.md +++ b/README_zh-CN.md @@ -23,6 +23,20 @@ +## 🚀 Speed Benchmark + +- XTuner 与 LLaMA-Factory 在 Llama2-7B 模型上的训练效率对比 + +
    + +
    + +- XTuner 与 LLaMA-Factory 在 Llama2-70B 模型上的训练效率对比 + +
    + +
    + ## 🎉 更新 - **\[2024/04\]** 多模态大模型 [LLaVA-Phi-3-mini](https://huggingface.co/xtuner/llava-phi-3-mini-hf) 发布！快速开始请查阅此[文档](xtuner/configs/llava/phi3_mini_4k_instruct_clip_vit_large_p14_336)！ @@ -65,31 +79,6 @@ XTuner 是一个高效、灵活、全能的轻量化大模型微调工具库。 - 预定义众多开源对话模版，支持与开源或训练所得模型进行对话。 - 训练所得模型可无缝接入部署工具库 [LMDeploy](https://github.com/InternLM/lmdeploy)、大规模评测工具库 [OpenCompass](https://github.com/open-compass/opencompass) 及 [VLMEvalKit](https://github.com/open-compass/VLMEvalKit)。 -## 🌟 示例 - -- XTuner APIs所提供的开箱即用的模型与数据集 [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/17CSO7T8q6KePuvu684IiHl6_id-CjPjh?usp=sharing) - -- QLoRA 微调 [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1QAEZVBfQ7LZURkMUtaq0b-5nEQII9G9Z?usp=sharing) - -- 基于插件的对话 [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/144OuTVyT_GvFyDMtlSlTzcxYIfnRsklq?usp=sharing) - - - - - - - - - - -
    基于插件的对话 🔥🔥🔥
    - - - - - -
    - ## 🔥 支持列表 @@ -112,13 +101,12 @@ XTuner 是一个高效、灵活、全能的轻量化大模型微调工具库。
    • InternLM2
      • -
    • InternLM
      • -
    • Llama
      • +
    • Llama 3
    • Llama 2
      • +
    • Phi-3
    • ChatGLM2
    • ChatGLM3
    • Qwen
      • -
    • Baichuan
    • Baichuan2
    • Mixtral 8x7B
    • DeepSeek MoE
      • @@ -192,7 +180,7 @@ XTuner 是一个高效、灵活、全能的轻量化大模型微调工具库。 pip install -e '.[all]' ``` -### 微调 [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1QAEZVBfQ7LZURkMUtaq0b-5nEQII9G9Z?usp=sharing) +### 微调 XTuner 支持微调大语言模型。数据集预处理指南请查阅[文档](./docs/zh_cn/user_guides/dataset_prepare.md)。 @@ -235,7 +223,7 @@ XTuner 支持微调大语言模型。数据集预处理指南请查阅[文档](. xtuner convert pth_to_hf ${CONFIG_NAME_OR_PATH} ${PTH} ${SAVE_PATH} ``` -### 对话 [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/144OuTVyT_GvFyDMtlSlTzcxYIfnRsklq?usp=sharing) +### 对话 XTuner 提供与大语言模型对话的工具。 @@ -295,6 +283,7 @@ xtuner chat internlm/internlm2-chat-7b --visual-encoder openai/clip-vit-large-pa ## 🎖️ 致谢 - [Llama 2](https://github.com/facebookresearch/llama) +- [DeepSpeed](https://github.com/microsoft/DeepSpeed) - [QLoRA](https://github.com/artidoro/qlora) - [LMDeploy](https://github.com/InternLM/lmdeploy) - [LLaVA](https://github.com/haotian-liu/LLaVA)