Skip to content

GitLab

Commit 12d5cbac authored by chenzk's avatar chenzk
Browse files

v1.0

parents
Pipeline #1780 canceled with stages
Show whitespace changes
Inline Side-by-side
Showing with 4241 additions and 0 deletions
.dockerignore 0 → 100644
View file @ 12d5cbac
.vscode
.git
.github
.venv
cache
data
docker
saves
hf_cache
output
.dockerignore
.gitattributes
.gitignore
.env.local 0 → 100644
View file @ 12d5cbac
# Note: actually we do not support .env, just for reference
# api
API_HOST=0.0.0.0
API_PORT=8000
API_KEY=
API_MODEL_NAME=gpt-3.5-turbo
FASTAPI_ROOT_PATH=
# general
DISABLE_VERSION_CHECK=
FORCE_CHECK_IMPORTS=
LLAMAFACTORY_VERBOSITY=
USE_MODELSCOPE_HUB=
RECORD_VRAM=
# torchrun
FORCE_TORCHRUN=
MASTER_ADDR=
MASTER_PORT=
NNODES=
RANK=
NPROC_PER_NODE=
# wandb
WANDB_DISABLED=
WANDB_PROJECT=huggingface
WANDB_API_KEY=
# gradio ui
GRADIO_SHARE=False
GRADIO_SERVER_NAME=0.0.0.0
GRADIO_SERVER_PORT=
GRADIO_ROOT_PATH=
# setup
ENABLE_SHORT_CONSOLE=1
# reserved (do not use)
LLAMABOARD_ENABLED=
LLAMABOARD_WORKDIR=
.gitattributes 0 → 100644
View file @ 12d5cbac
# Auto detect text files and perform LF normalization
* text=auto
.gitignore 0 → 100644
View file @ 12d5cbac
# Byte-compiled / optimized / DLL files
__pycache__/
*.py[cod]
*$py.class
# C extensions
*.so
# Distribution / packaging
.Python
build/
develop-eggs/
dist/
downloads/
eggs/
.eggs/
lib/
lib64/
parts/
sdist/
var/
wheels/
share/python-wheels/
*.egg-info/
.installed.cfg
*.egg
MANIFEST
# PyInstaller
# Usually these files are written by a python script from a template
# before PyInstaller builds the exe, so as to inject date/other infos into it.
*.manifest
*.spec
# Installer logs
pip-log.txt
pip-delete-this-directory.txt
# Unit test / coverage reports
htmlcov/
.tox/
.nox/
.coverage
.coverage.*
.cache
nosetests.xml
coverage.xml
*.cover
*.py,cover
.hypothesis/
.pytest_cache/
cover/
# Translations
*.mo
*.pot
# Django stuff:
*.log
local_settings.py
db.sqlite3
db.sqlite3-journal
# Flask stuff:
instance/
.webassets-cache
# Scrapy stuff:
.scrapy
# Sphinx documentation
docs/_build/
# PyBuilder
.pybuilder/
target/
# Jupyter Notebook
.ipynb_checkpoints
# IPython
profile_default/
ipython_config.py
# pyenv
# For a library or package, you might want to ignore these files since the code is
# intended to run in multiple environments; otherwise, check them in:
# .python-version
# pipenv
# According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
# However, in case of collaboration, if having platform-specific dependencies or dependencies
# having no cross-platform support, pipenv may install dependencies that don't work, or not
# install all needed dependencies.
#Pipfile.lock
# poetry
# Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
# This is especially recommended for binary packages to ensure reproducibility, and is more
# commonly ignored for libraries.
# https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
#poetry.lock
# pdm
# Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
#pdm.lock
# pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
# in version control.
# https://pdm.fming.dev/#use-with-ide
.pdm.toml
# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
__pypackages__/
# Celery stuff
celerybeat-schedule
celerybeat.pid
# SageMath parsed files
*.sage.py
# Environments
.env
.venv
env/
venv/
ENV/
env.bak/
venv.bak/
# Spyder project settings
.spyderproject
.spyproject
# Rope project settings
.ropeproject
# mkdocs documentation
/site
# mypy
.mypy_cache/
.dmypy.json
dmypy.json
# Pyre type checker
.pyre/
# pytype static type analyzer
.pytype/
# Cython debug symbols
cython_debug/
# PyCharm
# JetBrains specific template is maintained in a separate JetBrains.gitignore that can
# be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
# and can be added to the global gitignore or merged into this file. For a more nuclear
# option (not recommended) you can uncomment the following to ignore the entire idea folder.
.idea/
# custom .gitignore
ms_cache/
hf_cache/
cache/
config/
saves/
output/
wandb/
CITATION.cff 0 → 100644
View file @ 12d5cbac
cff-version: 1.2.0
date-released: 2024-03
message: "If you use this software, please cite it as below."
authors:
- family-names: "Zheng"
given-names: "Yaowei"
- family-names: "Zhang"
given-names: "Richong"
- family-names: "Zhang"
given-names: "Junhao"
- family-names: "Ye"
given-names: "Yanhan"
- family-names: "Luo"
given-names: "Zheyan"
- family-names: "Feng"
given-names: "Zhangchi"
- family-names: "Ma"
given-names: "Yongqiang"
title: "LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models"
url: "https://arxiv.org/abs/2403.13372"
preferred-citation:
type: conference-paper
conference:
name: "Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)"
authors:
- family-names: "Zheng"
given-names: "Yaowei"
- family-names: "Zhang"
given-names: "Richong"
- family-names: "Zhang"
given-names: "Junhao"
- family-names: "Ye"
given-names: "Yanhan"
- family-names: "Luo"
given-names: "Zheyan"
- family-names: "Feng"
given-names: "Zhangchi"
- family-names: "Ma"
given-names: "Yongqiang"
title: "LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models"
url: "https://arxiv.org/abs/2403.13372"
year: 2024
publisher: "Association for Computational Linguistics"
address: "Bangkok, Thailand"
LICENSE 0 → 100644
View file @ 12d5cbac
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
MANIFEST.in 0 → 100644
View file @ 12d5cbac
include LICENSE requirements.txt
Makefile 0 → 100644
View file @ 12d5cbac
.PHONY: quality style test
check_dirs := scripts src tests setup.py
quality:
ruff check $(check_dirs)
ruff format --check $(check_dirs)
style:
ruff check $(check_dirs) --fix
ruff format $(check_dirs)
test:
CUDA_VISIBLE_DEVICES= pytest tests/
README.md 0 → 100644
View file @ 12d5cbac
# Llama3.2
Llama 3.2支持12.8万个token的上下文长度，在图像识别和其他视觉理解任务上能够与OpenAI的GPT 4o-mini相媲美，其中3B计算量的模型方便用于边缘端部署，Llama 3.2 3B模型在IFEval评测中达到了Llama 3.1 8B的水平。
## 论文
`Open and Efficient Foundation Language Models`
- https://arxiv.org/pdf/2405.14458
## 模型结构
Llama 3.2继续沿用Decoder-only结构，支持大型上下文窗口（最多 128K 个标记），其GQA在推理过程提速比较明显。
<div align=center>
<img src="./doc/llama3.png"/>
</div>
## 算法原理
通过剪枝（Pruning）和知识蒸馏，从Llama 3.1的8B和70B分别获得Llama 3.2的1B和3B模型，实现了在边缘设备端高效运行的能力，在后期训练中，采用与Llama 3.1类似的流程，通过多轮对预训练模型的对齐，包括监督微调、拒绝采样和直接偏好优化，生成最终的对话模型，另外，开始支持11B和90B多模态模型。
<div align=center>
<img src="./doc/distillation.png"/>
</div>
## 环境配置
```
mv llama-factory-llama3.2_pytorch LLaMA-Factory-Llama3.2 # 去框架名后缀
```
### Docker（方法一）
```
docker pull image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.3.0-ubuntu22.04-dtk24.04.2-py3.10
# <your IMAGE ID>为以上拉取的docker的镜像ID替换，本镜像为：83714c19d308
docker run -it --shm-size=64G -v $PWD/LLaMA-Factory-Llama3.2:/home/LLaMA-Factory-Llama3.2 -v /opt/hyhal:/opt/hyhal:ro --privileged=true --device=/dev/kfd --device=/dev/dri/ --group-add video --name llama32 <your IMAGE ID> bash
cd /home/LLaMA-Factory-Llama3.2
pip install -r requirements.txt # requirements.txt
pip install -e ".[torch,metrics]" # 安装llamafactory-cli=0.9.1
```
### Dockerfile（方法二）
```
cd LLaMA-Factory-Llama3.2/docker
docker build --no-cache -t llama32:latest .
docker run --shm-size=64G --name llama32 -v /opt/hyhal:/opt/hyhal:ro --privileged=true --device=/dev/kfd --device=/dev/dri/ --group-add video -v $PWD/../../LLaMA-Factory-Llama3.2:/home/LLaMA-Factory-Llama3.2 -it llama32 bash
# 若遇到Dockerfile启动的方式安装环境需要长时间等待，可注释掉里面的pip安装，启动容器后再安装python库：pip install -r requirements.txt。
# 在镜像里安装llamafactory-cli=0.9.1
cd /home/LLaMA-Factory-Llama3.2
pip install -e ".[torch,metrics]"
```
### Anaconda（方法三）
1、关于本项目DCU显卡所需的特殊深度学习库可从光合开发者社区下载安装：
- https://developer.hpccube.com/tool/
```
DTK驱动:dtk24.04.2
python:python3.10
torch:2.3.0
torchvision:0.18.1
torchaudio:2.1.2
triton:2.1.0
flash-attn:2.0.4
deepspeed:0.14.2
apex:1.3.0
xformers:0.0.25
```
`Tips：以上dtk驱动、python、torch等DCU相关工具版本需要严格一一对应。`
2、其它非特殊库参照requirements.txt安装
```
pip install -r requirements.txt # requirements.txt
pip install -e ".[torch,metrics]" # 安装llamafactory-cli=0.9.1
```
## 数据集
`identity`、`alpaca_en_demo`
数据集快速下载中心：[SCNet AIDatasets](http://113.200.138.88:18080/aidatasets) 。
项目中已提供用于试验训练的迷你数据集:[`identity`](./data/identity.json) 、 [`alpaca_en_demo`](./data/alpaca_en_demo.json)，训练数据目录结构如下：
```
/home/LLaMA-Factory-Llama3.2/data
├── identity.json
├── alpaca_en_demo.json
├── ...
```
用户自定义产品场景训练数据的制作方式请参照[`data/README`](./data/README.md)中的格式介绍，并更新[data/dataset_info](data/dataset_info.json)进行使用。例如微调数据集格式：
```
# 以下内容放入文件data.json
[
{
"instruction": "人类指令（必填）",
"input": "人类输入（选填）",
"output": "模型回答（必填）",
"system": "系统提示词（选填）",
"history": [
["第一轮指令（选填）", "第一轮回答（选填）"],
["第二轮指令（选填）", "第二轮回答（选填）"]
]
}
]
```
对于上述格式的数据，dataset_info.json 中的数据集描述应为：
```
"数据集名称": {
"file_name": "data.json",
"columns": {
"prompt": "instruction",
"query": "input",
"response": "output",
"system": "system",
"history": "history"
}
}
```
## 训练
### 单机多卡
```
# 微调
# 修改llama3_lora_sft.yaml中的model_name_or_path: meta-llama/Llama-3.2-3B-Instruct
cd /home/LLaMA-Factory-Llama3.2
llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml #本README以Llama-3.2-3B-Instruct进行示例，其它llama模型可以此类推进行使用。
# 微调完成后，利用以下命令将微调结果与下载的原始预训练权重合并成最终权重进行发布：
llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
```
更多资料可参考源项目的[`README_origin`](./README_origin.md)
## 推理
```
# 方法一：pytorch 推理
llamafactory-cli chat examples/inference/llama3_lora_sft.yaml
# 方法二：vllm 推理
python infer_vllm.py # 目前vllm官网发布的vllm=0.5.0暂未兼容llama3.2，待vllm兼容后，方可从光合开发者社区下载新的适配版本利用此命令进行vllm推理。
```
## result
`User: `
```
世界上效果最好的对话大模型是哪个？
```
`Assistant:`
```
目前，世界上效果最好的对话大模型是{{name}}。
```
### 精度
DCU与GPU精度一致，推理框架：pytorch。
## 应用场景
### 算法类别
`对话问答`
### 热点应用行业
`制造,广媒,金融,能源,医疗,家居,教育`
## 预训练权重
预训练权重快速下载中心：[SCNet AIModels](http://113.200.138.88:18080/aimodels) ，项目中的预训练权重可从快速下载通道下载：[Llama-3.2-3B-Instruct](http://113.200.138.88:18080/aimodels/meta-llama/Llama-3.2-3B-Instruct.git) 。
Hugging Face下载地址为：[Llama-3.2-3B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct)
## 源码仓库及问题反馈
- http://developer.hpccube.com/codes/modelzoo/yolov10_pytorch.git
## 参考资料
- https://github.com/hiyouga/LLaMA-Factory.git
- https://ai.meta.com/blog/llama-3-2-connect-2024-vision-edge-mobile-devices/
README_origin.md 0 → 100644
View file @ 12d5cbac
![# LLaMA Factory](assets/logo.png)
[![GitHub Repo stars](https://img.shields.io/github/stars/hiyouga/LLaMA-Factory?style=social)](https://github.com/hiyouga/LLaMA-Factory/stargazers)
[![GitHub Code License](https://img.shields.io/github/license/hiyouga/LLaMA-Factory)](LICENSE)
[![GitHub last commit](https://img.shields.io/github/last-commit/hiyouga/LLaMA-Factory)](https://github.com/hiyouga/LLaMA-Factory/commits/main)
[![PyPI](https://img.shields.io/pypi/v/llamafactory)](https://pypi.org/project/llamafactory/)
[![Citation](https://img.shields.io/badge/citation-91-green)](#projects-using-llama-factory)
[![GitHub pull request](https://img.shields.io/badge/PRs-welcome-blue)](https://github.com/hiyouga/LLaMA-Factory/pulls)
[![Discord](https://dcbadge.vercel.app/api/server/rKfvV9r9FK?compact=true&style=flat)](https://discord.gg/rKfvV9r9FK)
[![Twitter](https://img.shields.io/twitter/follow/llamafactory_ai)](https://twitter.com/llamafactory_ai)
[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing)
[![Open in DSW](https://gallery.pai-ml.com/assets/open-in-dsw.svg)](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory)
[![Spaces](https://img.shields.io/badge/🤗-Open%20in%20Spaces-blue)](https://huggingface.co/spaces/hiyouga/LLaMA-Board)
[![Studios](https://img.shields.io/badge/ModelScope-Open%20in%20Studios-blue)](https://modelscope.cn/studios/hiyouga/LLaMA-Board)
[![GitHub Tread](https://trendshift.io/api/badge/repositories/4535)](https://trendshift.io/repositories/4535)
👋 Join our [WeChat](assets/wechat.jpg) or [NPU user group](assets/wechat_npu.jpg).
\[ English | [中文](README_zh.md) \]
**Fine-tuning a large language model can be easy as...**
https://github.com/user-attachments/assets/7c96b465-9df7-45f4-8053-bf03e58386d3
Choose your path:
- **Colab**: https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing
- **PAI-DSW**: [Llama3 Example](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory) | [Qwen2-VL Example](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory_qwen2vl)
- **Local machine**: Please refer to [usage](#getting-started)
- **Documentation (WIP)**: https://llamafactory.readthedocs.io/zh-cn/latest/
> [!NOTE]
> Except for the above links, all other websites are unauthorized third-party websites. Please carefully use them.
## Table of Contents
- [Features](#features)
- [Benchmark](#benchmark)
- [Changelog](#changelog)
- [Supported Models](#supported-models)
- [Supported Training Approaches](#supported-training-approaches)
- [Provided Datasets](#provided-datasets)
- [Requirement](#requirement)
- [Getting Started](#getting-started)
- [Projects using LLaMA Factory](#projects-using-llama-factory)
- [License](#license)
- [Citation](#citation)
- [Acknowledgement](#acknowledgement)
## Features
- **Various models**: LLaMA, LLaVA, Mistral, Mixtral-MoE, Qwen, Qwen2-VL, Yi, Gemma, Baichuan, ChatGLM, Phi, etc.
- **Integrated methods**: (Continuous) pre-training, (multimodal) supervised fine-tuning, reward modeling, PPO, DPO, KTO, ORPO, etc.
- **Scalable resources**: 16-bit full-tuning, freeze-tuning, LoRA and 2/3/4/5/6/8-bit QLoRA via AQLM/AWQ/GPTQ/LLM.int8/HQQ/EETQ.
- **Advanced algorithms**: [GaLore](https://github.com/jiaweizzhao/GaLore), [BAdam](https://github.com/Ledzy/BAdam), [Adam-mini](https://github.com/zyushun/Adam-mini), DoRA, LongLoRA, LLaMA Pro, Mixture-of-Depths, LoRA+, LoftQ, PiSSA and Agent tuning.
- **Practical tricks**: [FlashAttention-2](https://github.com/Dao-AILab/flash-attention), [Unsloth](https://github.com/unslothai/unsloth), [Liger Kernel](https://github.com/linkedin/Liger-Kernel), RoPE scaling, NEFTune and rsLoRA.
- **Experiment monitors**: LlamaBoard, TensorBoard, Wandb, MLflow, etc.
- **Faster inference**: OpenAI-style API, Gradio UI and CLI with vLLM worker.
## Benchmark
Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/ptuning), LLaMA Factory's LoRA tuning offers up to **3.7 times faster** training speed with a better Rouge score on the advertising text generation task. By leveraging 4-bit quantization technique, LLaMA Factory's QLoRA further improves the efficiency regarding the GPU memory.
![benchmark](assets/benchmark.svg)
<details><summary>Definitions</summary>
- **Training Speed**: the number of training samples processed per second during the training. (bs=4, cutoff_len=1024)
- **Rouge Score**: Rouge-2 score on the development set of the [advertising text generation](https://aclanthology.org/D19-1321.pdf) task. (bs=4, cutoff_len=1024)
- **GPU Memory**: Peak GPU memory usage in 4-bit quantized training. (bs=1, cutoff_len=1024)
- We adopt `pre_seq_len=128` for ChatGLM's P-Tuning and `lora_rank=32` for LLaMA Factory's LoRA tuning.
</details>
## Changelog
[24/10/09] We supported downloading pre-trained models and datasets from the **[Modelers Hub](https://modelers.cn/models)**. See [this tutorial](#download-from-modelers-hub) for usage.
[24/09/19] We support fine-tuning the **[Qwen2.5](https://qwenlm.github.io/blog/qwen2.5/)** models.
[24/08/30] We support fine-tuning the **[Qwen2-VL](https://qwenlm.github.io/blog/qwen2-vl/)** models. Thank [@simonJJJ](https://github.com/simonJJJ)'s PR.
[24/08/27] We support **[Liger Kernel](https://github.com/linkedin/Liger-Kernel)**. Try `enable_liger_kernel: true` for efficient training.
[24/08/09] We support **[Adam-mini](https://github.com/zyushun/Adam-mini)** optimizer. See [examples](examples/README.md) for usage. Thank [@relic-yuexi](https://github.com/relic-yuexi)'s PR.
<details><summary>Full Changelog</summary>
[24/07/04] We support [contamination-free packed training](https://github.com/MeetKai/functionary/tree/main/functionary/train/packing). Use `neat_packing: true` to activate it. Thank [@chuan298](https://github.com/chuan298)'s PR.
[24/06/16] We support **[PiSSA](https://arxiv.org/abs/2404.02948)** algorithm. See [examples](examples/README.md) for usage.
[24/06/07] We supported fine-tuning the **[Qwen2](https://qwenlm.github.io/blog/qwen2/)** and **[GLM-4](https://github.com/THUDM/GLM-4)** models.
[24/05/26] We supported **[SimPO](https://arxiv.org/abs/2405.14734)** algorithm for preference learning. See [examples](examples/README.md) for usage.
[24/05/20] We supported fine-tuning the **PaliGemma** series models. Note that the PaliGemma models are pre-trained models, you need to fine-tune them with `paligemma` template for chat completion.
[24/05/18] We supported **[KTO](https://arxiv.org/abs/2402.01306)** algorithm for preference learning. See [examples](examples/README.md) for usage.
[24/05/14] We supported training and inference on the Ascend NPU devices. Check [installation](#installation) section for details.
[24/04/26] We supported fine-tuning the **LLaVA-1.5** multimodal LLMs. See [examples](examples/README.md) for usage.
[24/04/22] We provided a **[Colab notebook](https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing)** for fine-tuning the Llama-3 model on a free T4 GPU. Two Llama-3-derived models fine-tuned using LLaMA Factory are available at Hugging Face, check [Llama3-8B-Chinese-Chat](https://huggingface.co/shenzhi-wang/Llama3-8B-Chinese-Chat) and [Llama3-Chinese](https://huggingface.co/zhichen/Llama3-Chinese) for details.
[24/04/21] We supported **[Mixture-of-Depths](https://arxiv.org/abs/2404.02258)** according to [AstraMindAI's implementation](https://github.com/astramind-ai/Mixture-of-depths). See [examples](examples/README.md) for usage.
[24/04/16] We supported **[BAdam](https://arxiv.org/abs/2404.02827)** optimizer. See [examples](examples/README.md) for usage.
[24/04/16] We supported **[unsloth](https://github.com/unslothai/unsloth)**'s long-sequence training (Llama-2-7B-56k within 24GB). It achieves **117%** speed and **50%** memory compared with FlashAttention-2, more benchmarks can be found in [this page](https://github.com/hiyouga/LLaMA-Factory/wiki/Performance-comparison).
[24/03/31] We supported **[ORPO](https://arxiv.org/abs/2403.07691)**. See [examples](examples/README.md) for usage.
[24/03/21] Our paper "[LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models](https://arxiv.org/abs/2403.13372)" is available at arXiv!
[24/03/20] We supported **FSDP+QLoRA** that fine-tunes a 70B model on 2x24GB GPUs. See [examples](examples/README.md) for usage.
[24/03/13] We supported **[LoRA+](https://arxiv.org/abs/2402.12354)**. See [examples](examples/README.md) for usage.
[24/03/07] We supported **[GaLore](https://arxiv.org/abs/2403.03507)** optimizer. See [examples](examples/README.md) for usage.
[24/03/07] We integrated **[vLLM](https://github.com/vllm-project/vllm)** for faster and concurrent inference. Try `infer_backend: vllm` to enjoy **270%** inference speed.
[24/02/28] We supported weight-decomposed LoRA (**[DoRA](https://arxiv.org/abs/2402.09353)**). Try `use_dora: true` to activate DoRA training.
[24/02/15] We supported **block expansion** proposed by [LLaMA Pro](https://github.com/TencentARC/LLaMA-Pro). See [examples](examples/README.md) for usage.
[24/02/05] Qwen1.5 (Qwen2 beta version) series models are supported in LLaMA-Factory. Check this [blog post](https://qwenlm.github.io/blog/qwen1.5/) for details.
[24/01/18] We supported **agent tuning** for most models, equipping model with tool using abilities by fine-tuning with `dataset: glaive_toolcall_en`.
[23/12/23] We supported **[unsloth](https://github.com/unslothai/unsloth)**'s implementation to boost LoRA tuning for the LLaMA, Mistral and Yi models. Try `use_unsloth: true` argument to activate unsloth patch. It achieves **170%** speed in our benchmark, check [this page](https://github.com/hiyouga/LLaMA-Factory/wiki/Performance-comparison) for details.
[23/12/12] We supported fine-tuning the latest MoE model **[Mixtral 8x7B](https://huggingface.co/mistralai/Mixtral-8x7B-v0.1)** in our framework. See hardware requirement [here](#hardware-requirement).
[23/12/01] We supported downloading pre-trained models and datasets from the **[ModelScope Hub](https://modelscope.cn/models)**. See [this tutorial](#download-from-modelscope-hub) for usage.
[23/10/21] We supported **[NEFTune](https://arxiv.org/abs/2310.05914)** trick for fine-tuning. Try `neftune_noise_alpha: 5` argument to activate NEFTune.
[23/09/27] We supported **$S^2$-Attn** proposed by [LongLoRA](https://github.com/dvlab-research/LongLoRA) for the LLaMA models. Try `shift_attn: true` argument to enable shift short attention.
[23/09/23] We integrated MMLU, C-Eval and CMMLU benchmarks in this repo. See [examples](examples/README.md) for usage.
[23/09/10] We supported **[FlashAttention-2](https://github.com/Dao-AILab/flash-attention)**. Try `flash_attn: fa2` argument to enable FlashAttention-2 if you are using RTX4090, A100 or H100 GPUs.
[23/08/12] We supported **RoPE scaling** to extend the context length of the LLaMA models. Try `rope_scaling: linear` argument in training and `rope_scaling: dynamic` argument at inference to extrapolate the position embeddings.
[23/08/11] We supported **[DPO training](https://arxiv.org/abs/2305.18290)** for instruction-tuned models. See [examples](examples/README.md) for usage.
[23/07/31] We supported **dataset streaming**. Try `streaming: true` and `max_steps: 10000` arguments to load your dataset in streaming mode.
[23/07/29] We released two instruction-tuned 13B models at Hugging Face. See these Hugging Face Repos ([LLaMA-2](https://huggingface.co/hiyouga/Llama-2-Chinese-13b-chat) / [Baichuan](https://huggingface.co/hiyouga/Baichuan-13B-sft)) for details.
[23/07/18] We developed an **all-in-one Web UI** for training, evaluation and inference. Try `train_web.py` to fine-tune models in your Web browser. Thank [@KanadeSiina](https://github.com/KanadeSiina) and [@codemayq](https://github.com/codemayq) for their efforts in the development.
[23/07/09] We released **[FastEdit](https://github.com/hiyouga/FastEdit)** ⚡🩹, an easy-to-use package for editing the factual knowledge of large language models efficiently. Please follow [FastEdit](https://github.com/hiyouga/FastEdit) if you are interested.
[23/06/29] We provided a **reproducible example** of training a chat model using instruction-following datasets, see [Baichuan-7B-sft](https://huggingface.co/hiyouga/Baichuan-7B-sft) for details.
[23/06/22] We aligned the [demo API](src/api_demo.py) with the [OpenAI's](https://platform.openai.com/docs/api-reference/chat) format where you can insert the fine-tuned model in **arbitrary ChatGPT-based applications**.
[23/06/03] We supported quantized training and inference (aka **[QLoRA](https://github.com/artidoro/qlora)**). See [examples](examples/README.md) for usage.
</details>
## Supported Models
| Model | Model size | Template |
| ----------------------------------------------------------------- | -------------------------------- | ---------------- |
| [Baichuan 2](https://huggingface.co/baichuan-inc) | 7B/13B | baichuan2 |
| [BLOOM/BLOOMZ](https://huggingface.co/bigscience) | 560M/1.1B/1.7B/3B/7.1B/176B | - |
| [ChatGLM3](https://huggingface.co/THUDM) | 6B | chatglm3 |
| [Command R](https://huggingface.co/CohereForAI) | 35B/104B | cohere |
| [DeepSeek (Code/MoE)](https://huggingface.co/deepseek-ai) | 7B/16B/67B/236B | deepseek |
| [Falcon](https://huggingface.co/tiiuae) | 7B/11B/40B/180B | falcon |
| [Gemma/Gemma 2/CodeGemma](https://huggingface.co/google) | 2B/7B/9B/27B | gemma |
| [GLM-4](https://huggingface.co/THUDM) | 9B | glm4 |
| [InternLM2/InternLM2.5](https://huggingface.co/internlm) | 7B/20B | intern2 |
| [Llama](https://github.com/facebookresearch/llama) | 7B/13B/33B/65B | - |
| [Llama 2](https://huggingface.co/meta-llama) | 7B/13B/70B | llama2 |
| [Llama 3-3.2](https://huggingface.co/meta-llama) | 1B/3B/8B/70B | llama3 |
| [LLaVA-1.5](https://huggingface.co/llava-hf) | 7B/13B | llava |
| [LLaVA-NeXT](https://huggingface.co/llava-hf) | 7B/8B/13B/34B/72B/110B | llava_next |
| [LLaVA-NeXT-Video](https://huggingface.co/llava-hf) | 7B/34B | llava_next_video |
| [MiniCPM](https://huggingface.co/openbmb) | 1B/2B/4B | cpm/cpm3 |
| [Mistral/Mixtral](https://huggingface.co/mistralai) | 7B/8x7B/8x22B | mistral |
| [OLMo](https://huggingface.co/allenai) | 1B/7B | - |
| [PaliGemma](https://huggingface.co/google) | 3B | paligemma |
| [Phi-1.5/Phi-2](https://huggingface.co/microsoft) | 1.3B/2.7B | - |
| [Phi-3](https://huggingface.co/microsoft) | 4B/7B/14B | phi |
| [Qwen (1-2.5) (Code/Math/MoE)](https://huggingface.co/Qwen) | 0.5B/1.5B/3B/7B/14B/32B/72B/110B | qwen |
| [Qwen2-VL](https://huggingface.co/Qwen) | 2B/7B/72B | qwen2_vl |
| [StarCoder 2](https://huggingface.co/bigcode) | 3B/7B/15B | - |
| [XVERSE](https://huggingface.co/xverse) | 7B/13B/65B | xverse |
| [Yi/Yi-1.5 (Code)](https://huggingface.co/01-ai) | 1.5B/6B/9B/34B | yi |
| [Yi-VL](https://huggingface.co/01-ai) | 6B/34B | yi_vl |
| [Yuan 2](https://huggingface.co/IEITYuan) | 2B/51B/102B | yuan |
> [!NOTE]
> For the "base" models, the `template` argument can be chosen from `default`, `alpaca`, `vicuna` etc. But make sure to use the **corresponding template** for the "instruct/chat" models.
>
> Remember to use the **SAME** template in training and inference.
Please refer to [constants.py](src/llamafactory/extras/constants.py) for a full list of models we supported.
You also can add a custom chat template to [template.py](src/llamafactory/data/template.py).
## Supported Training Approaches
| Approach | Full-tuning | Freeze-tuning | LoRA | QLoRA |
| ---------------------- | ------------------ | ------------------ | ------------------ | ------------------ |
| Pre-Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| Supervised Fine-Tuning | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| Reward Modeling | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| PPO Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| DPO Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| KTO Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| ORPO Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| SimPO Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
> [!TIP]
> The implementation details of PPO can be found in [this blog](https://newfacade.github.io/notes-on-reinforcement-learning/17-ppo-trl.html).
## Provided Datasets
<details><summary>Pre-training datasets</summary>
- [Wiki Demo (en)](data/wiki_demo.txt)
- [RefinedWeb (en)](https://huggingface.co/datasets/tiiuae/falcon-refinedweb)
- [RedPajama V2 (en)](https://huggingface.co/datasets/togethercomputer/RedPajama-Data-V2)
- [Wikipedia (en)](https://huggingface.co/datasets/olm/olm-wikipedia-20221220)
- [Wikipedia (zh)](https://huggingface.co/datasets/pleisto/wikipedia-cn-20230720-filtered)
- [Pile (en)](https://huggingface.co/datasets/EleutherAI/pile)
- [SkyPile (zh)](https://huggingface.co/datasets/Skywork/SkyPile-150B)
- [FineWeb (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb)
- [FineWeb-Edu (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu)
- [The Stack (en)](https://huggingface.co/datasets/bigcode/the-stack)
- [StarCoder (en)](https://huggingface.co/datasets/bigcode/starcoderdata)
</details>
<details><summary>Supervised fine-tuning datasets</summary>
- [Identity (en&zh)](data/identity.json)
- [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)
- [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca-3)
- [Alpaca GPT4 (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
- [Glaive Function Calling V2 (en&zh)](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2)
- [LIMA (en)](https://huggingface.co/datasets/GAIR/lima)
- [Guanaco Dataset (multilingual)](https://huggingface.co/datasets/JosephusCheung/GuanacoDataset)
- [BELLE 2M (zh)](https://huggingface.co/datasets/BelleGroup/train_2M_CN)
- [BELLE 1M (zh)](https://huggingface.co/datasets/BelleGroup/train_1M_CN)
- [BELLE 0.5M (zh)](https://huggingface.co/datasets/BelleGroup/train_0.5M_CN)
- [BELLE Dialogue 0.4M (zh)](https://huggingface.co/datasets/BelleGroup/generated_chat_0.4M)
- [BELLE School Math 0.25M (zh)](https://huggingface.co/datasets/BelleGroup/school_math_0.25M)
- [BELLE Multiturn Chat 0.8M (zh)](https://huggingface.co/datasets/BelleGroup/multiturn_chat_0.8M)
- [UltraChat (en)](https://github.com/thunlp/UltraChat)
- [OpenPlatypus (en)](https://huggingface.co/datasets/garage-bAInd/Open-Platypus)
- [CodeAlpaca 20k (en)](https://huggingface.co/datasets/sahil2801/CodeAlpaca-20k)
- [Alpaca CoT (multilingual)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT)
- [OpenOrca (en)](https://huggingface.co/datasets/Open-Orca/OpenOrca)
- [SlimOrca (en)](https://huggingface.co/datasets/Open-Orca/SlimOrca)
- [MathInstruct (en)](https://huggingface.co/datasets/TIGER-Lab/MathInstruct)
- [Firefly 1.1M (zh)](https://huggingface.co/datasets/YeungNLP/firefly-train-1.1M)
- [Wiki QA (en)](https://huggingface.co/datasets/wiki_qa)
- [Web QA (zh)](https://huggingface.co/datasets/suolyer/webqa)
- [WebNovel (zh)](https://huggingface.co/datasets/zxbsmk/webnovel_cn)
- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
- [deepctrl (en&zh)](https://www.modelscope.cn/datasets/deepctrl/deepctrl-sft-data)
- [Advertise Generating (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)
- [ShareGPT Hyperfiltered (en)](https://huggingface.co/datasets/totally-not-an-llm/sharegpt-hyperfiltered-3k)
- [ShareGPT4 (en&zh)](https://huggingface.co/datasets/shibing624/sharegpt_gpt4)
- [UltraChat 200k (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k)
- [AgentInstruct (en)](https://huggingface.co/datasets/THUDM/AgentInstruct)
- [LMSYS Chat 1M (en)](https://huggingface.co/datasets/lmsys/lmsys-chat-1m)
- [Evol Instruct V2 (en)](https://huggingface.co/datasets/WizardLM/WizardLM_evol_instruct_V2_196k)
- [Cosmopedia (en)](https://huggingface.co/datasets/HuggingFaceTB/cosmopedia)
- [STEM (zh)](https://huggingface.co/datasets/hfl/stem_zh_instruction)
- [Ruozhiba (zh)](https://huggingface.co/datasets/hfl/ruozhiba_gpt4_turbo)
- [Neo-sft (zh)](https://huggingface.co/datasets/m-a-p/neo_sft_phase2)
- [WebInstructSub (en)](https://huggingface.co/datasets/TIGER-Lab/WebInstructSub)
- [Magpie-Pro-300K-Filtered (en)](https://huggingface.co/datasets/Magpie-Align/Magpie-Pro-300K-Filtered)
- [Magpie-ultra-v0.1 (en)](https://huggingface.co/datasets/argilla/magpie-ultra-v0.1)
- [LLaVA mixed (en&zh)](https://huggingface.co/datasets/BUAADreamer/llava-en-zh-300k)
- [Pokemon-gpt4o-captions (en&zh)](https://huggingface.co/datasets/jugg1024/pokemon-gpt4o-captions)
- [Open Assistant (de)](https://huggingface.co/datasets/mayflowergmbh/oasst_de)
- [Dolly 15k (de)](https://huggingface.co/datasets/mayflowergmbh/dolly-15k_de)
- [Alpaca GPT4 (de)](https://huggingface.co/datasets/mayflowergmbh/alpaca-gpt4_de)
- [OpenSchnabeltier (de)](https://huggingface.co/datasets/mayflowergmbh/openschnabeltier_de)
- [Evol Instruct (de)](https://huggingface.co/datasets/mayflowergmbh/evol-instruct_de)
- [Dolphin (de)](https://huggingface.co/datasets/mayflowergmbh/dolphin_de)
- [Booksum (de)](https://huggingface.co/datasets/mayflowergmbh/booksum_de)
- [Airoboros (de)](https://huggingface.co/datasets/mayflowergmbh/airoboros-3.0_de)
- [Ultrachat (de)](https://huggingface.co/datasets/mayflowergmbh/ultra-chat_de)
</details>
<details><summary>Preference datasets</summary>
- [DPO mixed (en&zh)](https://huggingface.co/datasets/hiyouga/DPO-En-Zh-20k)
- [UltraFeedback (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized)
- [RLHF-V (en)](https://huggingface.co/datasets/openbmb/RLHF-V-Dataset)
- [VLFeedback (en)](https://huggingface.co/datasets/Zhihui/VLFeedback)
- [Orca DPO Pairs (en)](https://huggingface.co/datasets/Intel/orca_dpo_pairs)
- [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
- [Orca DPO (de)](https://huggingface.co/datasets/mayflowergmbh/intel_orca_dpo_pairs_de)
- [KTO mixed (en)](https://huggingface.co/datasets/argilla/kto-mix-15k)
</details>
Some datasets require confirmation before using them, so we recommend logging in with your Hugging Face account using these commands.
```bash
pip install --upgrade huggingface_hub
huggingface-cli login
```
## Requirement
| Mandatory | Minimum | Recommend |
| ------------ | ------- | --------- |
| python | 3.8 | 3.11 |
| torch | 1.13.1 | 2.4.0 |
| transformers | 4.41.2 | 4.43.4 |
| datasets | 2.16.0 | 2.20.0 |
| accelerate | 0.30.1 | 0.32.0 |
| peft | 0.11.1 | 0.12.0 |
| trl | 0.8.6 | 0.9.6 |
| Optional | Minimum | Recommend |
| ------------ | ------- | --------- |
| CUDA | 11.6 | 12.2 |
| deepspeed | 0.10.0 | 0.14.0 |
| bitsandbytes | 0.39.0 | 0.43.1 |
| vllm | 0.4.3 | 0.5.0 |
| flash-attn | 2.3.0 | 2.6.3 |
### Hardware Requirement
\* *estimated*
| Method | Bits | 7B | 13B | 30B | 70B | 110B | 8x7B | 8x22B |
| ----------------- | ---- | ----- | ----- | ----- | ------ | ------ | ----- | ------ |
| Full | AMP | 120GB | 240GB | 600GB | 1200GB | 2000GB | 900GB | 2400GB |
| Full | 16 | 60GB | 120GB | 300GB | 600GB | 900GB | 400GB | 1200GB |
| Freeze | 16 | 20GB | 40GB | 80GB | 200GB | 360GB | 160GB | 400GB |
| LoRA/GaLore/BAdam | 16 | 16GB | 32GB | 64GB | 160GB | 240GB | 120GB | 320GB |
| QLoRA | 8 | 10GB | 20GB | 40GB | 80GB | 140GB | 60GB | 160GB |
| QLoRA | 4 | 6GB | 12GB | 24GB | 48GB | 72GB | 30GB | 96GB |
| QLoRA | 2 | 4GB | 8GB | 16GB | 24GB | 48GB | 18GB | 48GB |
## Getting Started
### Installation
> [!IMPORTANT]
> Installation is mandatory.
```bash
git clone --depth 1 https://github.com/hiyouga/LLaMA-Factory.git
cd LLaMA-Factory
pip install -e ".[torch,metrics]"
```
Extra dependencies available: torch, torch-npu, metrics, deepspeed, liger-kernel, bitsandbytes, hqq, eetq, gptq, awq, aqlm, vllm, galore, badam, adam-mini, qwen, modelscope, openmind, quality
> [!TIP]
> Use `pip install --no-deps -e .` to resolve package conflicts.
<details><summary>For Windows users</summary>
If you want to enable the quantized LoRA (QLoRA) on the Windows platform, you need to install a pre-built version of `bitsandbytes` library, which supports CUDA 11.1 to 12.2, please select the appropriate [release version](https://github.com/jllllll/bitsandbytes-windows-webui/releases/tag/wheels) based on your CUDA version.
```bash
pip install https://github.com/jllllll/bitsandbytes-windows-webui/releases/download/wheels/bitsandbytes-0.41.2.post2-py3-none-win_amd64.whl
```
To enable FlashAttention-2 on the Windows platform, you need to install the precompiled `flash-attn` library, which supports CUDA 12.1 to 12.2. Please download the corresponding version from [flash-attention](https://github.com/bdashore3/flash-attention/releases) based on your requirements.
</details>
<details><summary>For Ascend NPU users</summary>
To install LLaMA Factory on Ascend NPU devices, please specify extra dependencies: `pip install -e ".[torch-npu,metrics]"`. Additionally, you need to install the **[Ascend CANN Toolkit and Kernels](https://www.hiascend.com/developer/download/community/result?module=cann)**. Please follow the [installation tutorial](https://www.hiascend.com/document/detail/en/CANNCommunityEdition/600alphaX/softwareinstall/instg/atlasdeploy_03_0031.html) or use the following commands:
```bash
# replace the url according to your CANN version and devices
# install CANN Toolkit
wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C17SPC701/Ascend-cann-toolkit_8.0.RC1.alpha001_linux-"$(uname -i)".run
bash Ascend-cann-toolkit_8.0.RC1.alpha001_linux-"$(uname -i)".run --install
# install CANN Kernels
wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C17SPC701/Ascend-cann-kernels-910b_8.0.RC1.alpha001_linux.run
bash Ascend-cann-kernels-910b_8.0.RC1.alpha001_linux.run --install
# set env variables
source /usr/local/Ascend/ascend-toolkit/set_env.sh
```
| Requirement | Minimum | Recommend |
| ------------ | ------- | ----------- |
| CANN | 8.0.RC1 | 8.0.RC1 |
| torch | 2.1.0 | 2.1.0 |
| torch-npu | 2.1.0 | 2.1.0.post3 |
| deepspeed | 0.13.2 | 0.13.2 |
Remember to use `ASCEND_RT_VISIBLE_DEVICES` instead of `CUDA_VISIBLE_DEVICES` to specify the device to use.
If you cannot infer model on NPU devices, try setting `do_sample: false` in the configurations.
Download the pre-built Docker images: [32GB](http://mirrors.cn-central-221.ovaijisuan.com/detail/130.html) | [64GB](http://mirrors.cn-central-221.ovaijisuan.com/detail/131.html)
</details>
### Data Preparation
Please refer to [data/README.md](data/README.md) for checking the details about the format of dataset files. You can either use datasets on HuggingFace / ModelScope / Modelers hub or load the dataset in local disk.
> [!NOTE]
> Please update `data/dataset_info.json` to use your custom dataset.
### Quickstart
Use the following 3 commands to run LoRA **fine-tuning**, **inference** and **merging** of the Llama3-8B-Instruct model, respectively.
```bash
llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
llamafactory-cli chat examples/inference/llama3_lora_sft.yaml
llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
```
See [examples/README.md](examples/README.md) for advanced usage (including distributed training).
> [!TIP]
> Use `llamafactory-cli help` to show help information.
### Fine-Tuning with LLaMA Board GUI (powered by [Gradio](https://github.com/gradio-app/gradio))
```bash
llamafactory-cli webui
```
### Build Docker
For CUDA users:
```bash
cd docker/docker-cuda/
docker compose up -d
docker compose exec llamafactory bash
```
For Ascend NPU users:
```bash
cd docker/docker-npu/
docker compose up -d
docker compose exec llamafactory bash
```
For AMD ROCm users:
```bash
cd docker/docker-rocm/
docker compose up -d
docker compose exec llamafactory bash
```
<details><summary>Build without Docker Compose</summary>
For CUDA users:
```bash
docker build -f ./docker/docker-cuda/Dockerfile \
--build-arg INSTALL_BNB=false \
--build-arg INSTALL_VLLM=false \
--build-arg INSTALL_DEEPSPEED=false \
--build-arg INSTALL_FLASHATTN=false \
--build-arg PIP_INDEX=https://pypi.org/simple \
-t llamafactory:latest .
docker run -dit --gpus=all \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-p 7860:7860 \
-p 8000:8000 \
--shm-size 16G \
--name llamafactory \
llamafactory:latest
docker exec -it llamafactory bash
```
For Ascend NPU users:
```bash
# Choose docker image upon your environment
docker build -f ./docker/docker-npu/Dockerfile \
--build-arg INSTALL_DEEPSPEED=false \
--build-arg PIP_INDEX=https://pypi.org/simple \
-t llamafactory:latest .
# Change `device` upon your resources
docker run -dit \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-v /usr/local/dcmi:/usr/local/dcmi \
-v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
-v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
-v /etc/ascend_install.info:/etc/ascend_install.info \
-p 7860:7860 \
-p 8000:8000 \
--device /dev/davinci0 \
--device /dev/davinci_manager \
--device /dev/devmm_svm \
--device /dev/hisi_hdc \
--shm-size 16G \
--name llamafactory \
llamafactory:latest
docker exec -it llamafactory bash
```
For AMD ROCm users:
```bash
docker build -f ./docker/docker-rocm/Dockerfile \
--build-arg INSTALL_BNB=false \
--build-arg INSTALL_VLLM=false \
--build-arg INSTALL_DEEPSPEED=false \
--build-arg INSTALL_FLASHATTN=false \
--build-arg PIP_INDEX=https://pypi.org/simple \
-t llamafactory:latest .
docker run -dit \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-v ./saves:/app/saves \
-p 7860:7860 \
-p 8000:8000 \
--device /dev/kfd \
--device /dev/dri \
--shm-size 16G \
--name llamafactory \
llamafactory:latest
docker exec -it llamafactory bash
```
</details>
<details><summary>Details about volume</summary>
- `hf_cache`: Utilize Hugging Face cache on the host machine. Reassignable if a cache already exists in a different directory.
- `ms_cache`: Similar to Hugging Face cache but for ModelScope users.
- `om_cache`: Similar to Hugging Face cache but for Modelers users.
- `data`: Place datasets on this dir of the host machine so that they can be selected on LLaMA Board GUI.
- `output`: Set export dir to this location so that the merged result can be accessed directly on the host machine.
</details>
### Deploy with OpenAI-style API and vLLM
```bash
API_PORT=8000 llamafactory-cli api examples/inference/llama3_vllm.yaml
```
> [!TIP]
> Visit [this page](https://platform.openai.com/docs/api-reference/chat/create) for API document.
### Download from ModelScope Hub
If you have trouble with downloading models and datasets from Hugging Face, you can use ModelScope.
```bash
export USE_MODELSCOPE_HUB=1 # `set USE_MODELSCOPE_HUB=1` for Windows
```
Train the model by specifying a model ID of the ModelScope Hub as the `model_name_or_path`. You can find a full list of model IDs at [ModelScope Hub](https://modelscope.cn/models), e.g., `LLM-Research/Meta-Llama-3-8B-Instruct`.
### Download from Modelers Hub
You can also use Modelers Hub to download models and datasets.
```bash
export USE_OPENMIND_HUB=1 # `set USE_OPENMIND_HUB=1` for Windows
```
Train the model by specifying a model ID of the Modelers Hub as the `model_name_or_path`. You can find a full list of model IDs at [Modelers Hub](https://modelers.cn/models), e.g., `TeleAI/TeleChat-7B-pt`.
### Use W&B Logger
To use [Weights & Biases](https://wandb.ai) for logging experimental results, you need to add the following arguments to yaml files.
```yaml
report_to: wandb
run_name: test_run # optional
```
Set `WANDB_API_KEY` to [your key](https://wandb.ai/authorize) when launching training tasks to log in with your W&B account.
## 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)
1. Yu et al. Open, Closed, or Small Language Models for Text Classification? 2023. [[arxiv]](https://arxiv.org/abs/2308.10092)
1. Wang et al. UbiPhysio: Support Daily Functioning, Fitness, and Rehabilitation with Action Understanding and Feedback in Natural Language. 2023. [[arxiv]](https://arxiv.org/abs/2308.10526)
1. Luceri et al. Leveraging Large Language Models to Detect Influence Campaigns in Social Media. 2023. [[arxiv]](https://arxiv.org/abs/2311.07816)
1. Zhang et al. Alleviating Hallucinations of Large Language Models through Induced Hallucinations. 2023. [[arxiv]](https://arxiv.org/abs/2312.15710)
1. Wang et al. Know Your Needs Better: Towards Structured Understanding of Marketer Demands with Analogical Reasoning Augmented LLMs. KDD 2024. [[arxiv]](https://arxiv.org/abs/2401.04319)
1. Wang et al. CANDLE: Iterative Conceptualization and Instantiation Distillation from Large Language Models for Commonsense Reasoning. ACL 2024. [[arxiv]](https://arxiv.org/abs/2401.07286)
1. Choi et al. FACT-GPT: Fact-Checking Augmentation via Claim Matching with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2402.05904)
1. Zhang et al. AutoMathText: Autonomous Data Selection with Language Models for Mathematical Texts. 2024. [[arxiv]](https://arxiv.org/abs/2402.07625)
1. Lyu et al. KnowTuning: Knowledge-aware Fine-tuning for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11176)
1. Yang et al. LaCo: Large Language Model Pruning via Layer Collaps. 2024. [[arxiv]](https://arxiv.org/abs/2402.11187)
1. Bhardwaj et al. Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task Arithmetic. 2024. [[arxiv]](https://arxiv.org/abs/2402.11746)
1. Yang et al. Enhancing Empathetic Response Generation by Augmenting LLMs with Small-scale Empathetic Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11801)
1. Yi et al. Generation Meets Verification: Accelerating Large Language Model Inference with Smart Parallel Auto-Correct Decoding. ACL 2024 Findings. [[arxiv]](https://arxiv.org/abs/2402.11809)
1. Cao et al. Head-wise Shareable Attention for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11819)
1. Zhang et al. Enhancing Multilingual Capabilities of Large Language Models through Self-Distillation from Resource-Rich Languages. 2024. [[arxiv]](https://arxiv.org/abs/2402.12204)
1. Kim et al. Efficient and Effective Vocabulary Expansion Towards Multilingual Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.14714)
1. Yu et al. KIEval: A Knowledge-grounded Interactive Evaluation Framework for Large Language Models. ACL 2024. [[arxiv]](https://arxiv.org/abs/2402.15043)
1. Huang et al. Key-Point-Driven Data Synthesis with its Enhancement on Mathematical Reasoning. 2024. [[arxiv]](https://arxiv.org/abs/2403.02333)
1. Duan et al. Negating Negatives: Alignment without Human Positive Samples via Distributional Dispreference Optimization. 2024. [[arxiv]](https://arxiv.org/abs/2403.03419)
1. Xie and Schwertfeger. Empowering Robotics with Large Language Models: osmAG Map Comprehension with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2403.08228)
1. Wu et al. Large Language Models are Parallel Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2403.09073)
1. Zhang et al. EDT: Improving Large Language Models' Generation by Entropy-based Dynamic Temperature Sampling. 2024. [[arxiv]](https://arxiv.org/abs/2403.14541)
1. Weller et al. FollowIR: Evaluating and Teaching Information Retrieval Models to Follow Instructions. 2024. [[arxiv]](https://arxiv.org/abs/2403.15246)
1. Hongbin Na. CBT-LLM: A Chinese Large Language Model for Cognitive Behavioral Therapy-based Mental Health Question Answering. COLING 2024. [[arxiv]](https://arxiv.org/abs/2403.16008)
1. Zan et al. CodeS: Natural Language to Code Repository via Multi-Layer Sketch. 2024. [[arxiv]](https://arxiv.org/abs/2403.16443)
1. Liu et al. Extensive Self-Contrast Enables Feedback-Free Language Model Alignment. 2024. [[arxiv]](https://arxiv.org/abs/2404.00604)
1. Luo et al. BAdam: A Memory Efficient Full Parameter Training Method for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.02827)
1. Du et al. Chinese Tiny LLM: Pretraining a Chinese-Centric Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2404.04167)
1. Ma et al. Parameter Efficient Quasi-Orthogonal Fine-Tuning via Givens Rotation. ICML 2024. [[arxiv]](https://arxiv.org/abs/2404.04316)
1. Liu et al. Dynamic Generation of Personalities with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.07084)
1. Shang et al. How Far Have We Gone in Stripped Binary Code Understanding Using Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.09836)
1. Huang et al. LLMTune: Accelerate Database Knob Tuning with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.11581)
1. Deng et al. Text-Tuple-Table: Towards Information Integration in Text-to-Table Generation via Global Tuple Extraction. 2024. [[arxiv]](https://arxiv.org/abs/2404.14215)
1. Acikgoz et al. Hippocrates: An Open-Source Framework for Advancing Large Language Models in Healthcare. 2024. [[arxiv]](https://arxiv.org/abs/2404.16621)
1. Zhang et al. Small Language Models Need Strong Verifiers to Self-Correct Reasoning. ACL 2024 Findings. [[arxiv]](https://arxiv.org/abs/2404.17140)
1. Zhou et al. FREB-TQA: A Fine-Grained Robustness Evaluation Benchmark for Table Question Answering. NAACL 2024. [[arxiv]](https://arxiv.org/abs/2404.18585)
1. Xu et al. Large Language Models for Cyber Security: A Systematic Literature Review. 2024. [[arxiv]](https://arxiv.org/abs/2405.04760)
1. Dammu et al. "They are uncultured": Unveiling Covert Harms and Social Threats in LLM Generated Conversations. 2024. [[arxiv]](https://arxiv.org/abs/2405.05378)
1. Yi et al. A safety realignment framework via subspace-oriented model fusion for large language models. 2024. [[arxiv]](https://arxiv.org/abs/2405.09055)
1. Lou et al. SPO: Multi-Dimensional Preference Sequential Alignment With Implicit Reward Modeling. 2024. [[arxiv]](https://arxiv.org/abs/2405.12739)
1. Zhang et al. Getting More from Less: Large Language Models are Good Spontaneous Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2405.13816)
1. Zhang et al. TS-Align: A Teacher-Student Collaborative Framework for Scalable Iterative Finetuning of Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2405.20215)
1. Zihong Chen. Sentence Segmentation and Sentence Punctuation Based on XunziALLM. 2024. [[paper]](https://aclanthology.org/2024.lt4hala-1.30)
1. Gao et al. The Best of Both Worlds: Toward an Honest and Helpful Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2406.00380)
1. Wang and Song. MARS: Benchmarking the Metaphysical Reasoning Abilities of Language Models with a Multi-task Evaluation Dataset. 2024. [[arxiv]](https://arxiv.org/abs/2406.02106)
1. Hu et al. Computational Limits of Low-Rank Adaptation (LoRA) for Transformer-Based Models. 2024. [[arxiv]](https://arxiv.org/abs/2406.03136)
1. Ge et al. Time Sensitive Knowledge Editing through Efficient Finetuning. ACL 2024. [[arxiv]](https://arxiv.org/abs/2406.04496)
1. Tan et al. Peer Review as A Multi-Turn and Long-Context Dialogue with Role-Based Interactions. 2024. [[arxiv]](https://arxiv.org/abs/2406.05688)
1. Song et al. Turbo Sparse: Achieving LLM SOTA Performance with Minimal Activated Parameters. 2024. [[arxiv]](https://arxiv.org/abs/2406.05955)
1. Gu et al. RWKV-CLIP: A Robust Vision-Language Representation Learner. 2024. [[arxiv]](https://arxiv.org/abs/2406.06973)
1. Chen et al. Advancing Tool-Augmented Large Language Models: Integrating Insights from Errors in Inference Trees. 2024. [[arxiv]](https://arxiv.org/abs/2406.07115)
1. Zhu et al. Are Large Language Models Good Statisticians?. 2024. [[arxiv]](https://arxiv.org/abs/2406.07815)
1. Li et al. Know the Unknown: An Uncertainty-Sensitive Method for LLM Instruction Tuning. 2024. [[arxiv]](https://arxiv.org/abs/2406.10099)
1. Ding et al. IntentionQA: A Benchmark for Evaluating Purchase Intention Comprehension Abilities of Language Models in E-commerce. 2024. [[arxiv]](https://arxiv.org/abs/2406.10173)
1. He et al. COMMUNITY-CROSS-INSTRUCT: Unsupervised Instruction Generation for Aligning Large Language Models to Online Communities. 2024. [[arxiv]](https://arxiv.org/abs/2406.12074)
1. Lin et al. FVEL: Interactive Formal Verification Environment with Large Language Models via Theorem Proving. 2024. [[arxiv]](https://arxiv.org/abs/2406.14408)
1. Treutlein et al. Connecting the Dots: LLMs can Infer and Verbalize Latent Structure from Disparate Training Data. 2024. [[arxiv]](https://arxiv.org/abs/2406.14546)
1. Feng et al. SS-Bench: A Benchmark for Social Story Generation and Evaluation. 2024. [[arxiv]](https://arxiv.org/abs/2406.15695)
1. Feng et al. Self-Constructed Context Decompilation with Fined-grained Alignment Enhancement. 2024. [[arxiv]](https://arxiv.org/abs/2406.17233)
1. Liu et al. Large Language Models for Cuffless Blood Pressure Measurement From Wearable Biosignals. 2024. [[arxiv]](https://arxiv.org/abs/2406.18069)
1. Iyer et al. Exploring Very Low-Resource Translation with LLMs: The University of Edinburgh's Submission to AmericasNLP 2024 Translation Task. AmericasNLP 2024. [[paper]](https://aclanthology.org/2024.americasnlp-1.25)
1. Li et al. Calibrating LLMs with Preference Optimization on Thought Trees for Generating Rationale in Science Question Scoring. 2024. [[arxiv]](https://arxiv.org/abs/2406.19949)
1. Yang et al. Financial Knowledge Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2407.00365)
1. Lin et al. DogeRM: Equipping Reward Models with Domain Knowledge through Model Merging. 2024. [[arxiv]](https://arxiv.org/abs/2407.01470)
1. Bako et al. Evaluating the Semantic Profiling Abilities of LLMs for Natural Language Utterances in Data Visualization. 2024. [[arxiv]](https://arxiv.org/abs/2407.06129)
1. Huang et al. RoLoRA: Fine-tuning Rotated Outlier-free LLMs for Effective Weight-Activation Quantization. 2024. [[arxiv]](https://arxiv.org/abs/2407.08044)
1. Jiang et al. LLM-Collaboration on Automatic Science Journalism for the General Audience. 2024. [[arxiv]](https://arxiv.org/abs/2407.09756)
1. Inouye et al. Applied Auto-tuning on LoRA Hyperparameters. 2024. [[paper]](https://scholarcommons.scu.edu/cseng_senior/272/)
1. Qi et al. Research on Tibetan Tourism Viewpoints information generation system based on LLM. 2024. [[arxiv]](https://arxiv.org/abs/2407.13561)
1. Xu et al. Course-Correction: Safety Alignment Using Synthetic Preferences. 2024. [[arxiv]](https://arxiv.org/abs/2407.16637)
1. Sun et al. LAMBDA: A Large Model Based Data Agent. 2024. [[arxiv]](https://arxiv.org/abs/2407.17535)
1. Zhu et al. CollectiveSFT: Scaling Large Language Models for Chinese Medical Benchmark with Collective Instructions in Healthcare. 2024. [[arxiv]](https://arxiv.org/abs/2407.19705)
1. Yu et al. Correcting Negative Bias in Large Language Models through Negative Attention Score Alignment. 2024. [[arxiv]](https://arxiv.org/abs/2408.00137)
1. Xie et al. The Power of Personalized Datasets: Advancing Chinese Composition Writing for Elementary School through Targeted Model Fine-Tuning. IALP 2024. [[paper]](https://www.asianlp.sg/conferences/ialp2024/proceedings/papers/IALP2024_P055.pdf)
1. Liu et al. Instruct-Code-Llama: Improving Capabilities of Language Model in Competition Level Code Generation by Online Judge Feedback. ICIC 2024. [[paper]](https://link.springer.com/chapter/10.1007/978-981-97-5669-8_11)
1. Wang et al. Cybernetic Sentinels: Unveiling the Impact of Safety Data Selection on Model Security in Supervised Fine-Tuning. ICIC 2024. [[paper]](https://link.springer.com/chapter/10.1007/978-981-97-5669-8_23)
1. Xia et al. Understanding the Performance and Estimating the Cost of LLM Fine-Tuning. 2024. [[arxiv]](https://arxiv.org/abs/2408.04693)
1. Zeng et al. Perceive, Reflect, and Plan: Designing LLM Agent for Goal-Directed City Navigation without Instructions. 2024. [[arxiv]](https://arxiv.org/abs/2408.04168)
1. Xia et al. Using Pre-trained Language Model for Accurate ESG Prediction. FinNLP 2024. [[paper]](https://aclanthology.org/2024.finnlp-2.1/)
1. Liang et al. I-SHEEP: Self-Alignment of LLM from Scratch through an Iterative Self-Enhancement Paradigm. 2024. [[arxiv]](https://arxiv.org/abs/2408.08072)
1. **[StarWhisper](https://github.com/Yu-Yang-Li/StarWhisper)**: A large language model for Astronomy, based on ChatGLM2-6B and Qwen-14B.
1. **[DISC-LawLLM](https://github.com/FudanDISC/DISC-LawLLM)**: A large language model specialized in Chinese legal domain, based on Baichuan-13B, is capable of retrieving and reasoning on legal knowledge.
1. **[Sunsimiao](https://github.com/X-D-Lab/Sunsimiao)**: A large language model specialized in Chinese medical domain, based on Baichuan-7B and ChatGLM-6B.
1. **[CareGPT](https://github.com/WangRongsheng/CareGPT)**: A series of large language models for Chinese medical domain, based on LLaMA2-7B and Baichuan-13B.
1. **[MachineMindset](https://github.com/PKU-YuanGroup/Machine-Mindset/)**: A series of MBTI Personality large language models, capable of giving any LLM 16 different personality types based on different datasets and training methods.
1. **[Luminia-13B-v3](https://huggingface.co/Nekochu/Luminia-13B-v3)**: A large language model specialized in generate metadata for stable diffusion. [[🤗Demo]](https://huggingface.co/spaces/Nekochu/Luminia-13B_SD_Prompt)
1. **[Chinese-LLaVA-Med](https://github.com/BUAADreamer/Chinese-LLaVA-Med)**: A multimodal large language model specialized in Chinese medical domain, based on LLaVA-1.5-7B.
1. **[AutoRE](https://github.com/THUDM/AutoRE)**: A document-level relation extraction system based on large language models.
1. **[NVIDIA RTX AI Toolkit](https://github.com/NVIDIA/RTX-AI-Toolkit)**: SDKs for fine-tuning LLMs on Windows PC for NVIDIA RTX.
1. **[LazyLLM](https://github.com/LazyAGI/LazyLLM)**: An easy and lazy way for building multi-agent LLMs applications and supports model fine-tuning via LLaMA Factory.
</details>
## License
This repository is licensed under the [Apache-2.0 License](LICENSE).
Please follow the model licenses to use the corresponding model weights: [Baichuan 2](https://huggingface.co/baichuan-inc/Baichuan2-7B-Base/blob/main/Community%20License%20for%20Baichuan%202%20Model.pdf) / [BLOOM](https://huggingface.co/spaces/bigscience/license) / [ChatGLM3](https://github.com/THUDM/ChatGLM3/blob/main/MODEL_LICENSE) / [Command R](https://cohere.com/c4ai-cc-by-nc-license) / [DeepSeek](https://github.com/deepseek-ai/DeepSeek-LLM/blob/main/LICENSE-MODEL) / [Falcon](https://huggingface.co/tiiuae/falcon-180B/blob/main/LICENSE.txt) / [Gemma](https://ai.google.dev/gemma/terms) / [GLM-4](https://huggingface.co/THUDM/glm-4-9b/blob/main/LICENSE) / [InternLM2](https://github.com/InternLM/InternLM#license) / [Llama](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md) / [Llama 2 (LLaVA-1.5)](https://ai.meta.com/llama/license/) / [Llama 3](https://llama.meta.com/llama3/license/) / [MiniCPM](https://github.com/OpenBMB/MiniCPM/blob/main/MiniCPM%20Model%20License.md) / [Mistral](LICENSE) / [OLMo](LICENSE) / [Phi-1.5/Phi-2](https://huggingface.co/microsoft/phi-1_5/resolve/main/Research%20License.docx) / [Phi-3](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/blob/main/LICENSE) / [Qwen](https://github.com/QwenLM/Qwen/blob/main/Tongyi%20Qianwen%20LICENSE%20AGREEMENT) / [StarCoder 2](https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement) / [XVERSE](https://github.com/xverse-ai/XVERSE-13B/blob/main/MODEL_LICENSE.pdf) / [Yi](https://huggingface.co/01-ai/Yi-6B/blob/main/LICENSE) / [Yi-1.5](LICENSE) / [Yuan 2](https://github.com/IEIT-Yuan/Yuan-2.0/blob/main/LICENSE-Yuan)
## Citation
If this work is helpful, please kindly cite as:
```bibtex
@inproceedings{zheng2024llamafactory,
title={LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models},
author={Yaowei Zheng and Richong Zhang and Junhao Zhang and Yanhan Ye and Zheyan Luo and Zhangchi Feng and Yongqiang Ma},
booktitle={Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
address={Bangkok, Thailand},
publisher={Association for Computational Linguistics},
year={2024},
url={http://arxiv.org/abs/2403.13372}
}
```
## Acknowledgement
This repo benefits from [PEFT](https://github.com/huggingface/peft), [TRL](https://github.com/huggingface/trl), [QLoRA](https://github.com/artidoro/qlora) and [FastChat](https://github.com/lm-sys/FastChat). Thanks for their wonderful works.
## Star History
![Star History Chart](https://api.star-history.com/svg?repos=hiyouga/LLaMA-Factory&type=Date)
README_zh.md 0 → 100644
View file @ 12d5cbac
![# LLaMA Factory](assets/logo.png)
[![GitHub Repo stars](https://img.shields.io/github/stars/hiyouga/LLaMA-Factory?style=social)](https://github.com/hiyouga/LLaMA-Factory/stargazers)
[![GitHub Code License](https://img.shields.io/github/license/hiyouga/LLaMA-Factory)](LICENSE)
[![GitHub last commit](https://img.shields.io/github/last-commit/hiyouga/LLaMA-Factory)](https://github.com/hiyouga/LLaMA-Factory/commits/main)
[![PyPI](https://img.shields.io/pypi/v/llamafactory)](https://pypi.org/project/llamafactory/)
[![Citation](https://img.shields.io/badge/citation-91-green)](#使用了-llama-factory-的项目)
[![GitHub pull request](https://img.shields.io/badge/PRs-welcome-blue)](https://github.com/hiyouga/LLaMA-Factory/pulls)
[![Discord](https://dcbadge.vercel.app/api/server/rKfvV9r9FK?compact=true&style=flat)](https://discord.gg/rKfvV9r9FK)
[![Twitter](https://img.shields.io/twitter/follow/llamafactory_ai)](https://twitter.com/llamafactory_ai)
[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1d5KQtbemerlSDSxZIfAaWXhKr30QypiK?usp=sharing)
[![Open in DSW](https://gallery.pai-ml.com/assets/open-in-dsw.svg)](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory)
[![Spaces](https://img.shields.io/badge/🤗-Open%20in%20Spaces-blue)](https://huggingface.co/spaces/hiyouga/LLaMA-Board)
[![Studios](https://img.shields.io/badge/ModelScope-Open%20in%20Studios-blue)](https://modelscope.cn/studios/hiyouga/LLaMA-Board)
[![GitHub Tread](https://trendshift.io/api/badge/repositories/4535)](https://trendshift.io/repositories/4535)
👋 加入我们的[微信群](assets/wechat.jpg)或 [NPU 用户群](assets/wechat_npu.jpg)。
\[ [English](README.md) | 中文 \]
**微调大模型可以像这样轻松…**
https://github.com/user-attachments/assets/e6ce34b0-52d5-4f3e-a830-592106c4c272
选择你的打开方式：
- **Colab**：https://colab.research.google.com/drive/1d5KQtbemerlSDSxZIfAaWXhKr30QypiK?usp=sharing
- **PAI-DSW**：[Llama3 案例](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory) | [Qwen2-VL 案例](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory_qwen2vl)
- **本地机器**：请见[如何使用](#如何使用)
- **入门教程**：https://zhuanlan.zhihu.com/p/695287607
- **框架文档**：https://llamafactory.readthedocs.io/zh-cn/latest/
> [!NOTE]
> 除上述链接以外的其他网站均为未经许可的第三方网站，请小心甄别。
## 目录
- [项目特色](#项目特色)
- [性能指标](#性能指标)
- [更新日志](#更新日志)
- [模型](#模型)
- [训练方法](#训练方法)
- [数据集](#数据集)
- [软硬件依赖](#软硬件依赖)
- [如何使用](#如何使用)
- [使用了 LLaMA Factory 的项目](#使用了-llama-factory-的项目)
- [协议](#协议)
- [引用](#引用)
- [致谢](#致谢)
## 项目特色
- **多种模型**：LLaMA、LLaVA、Mistral、Mixtral-MoE、Qwen、Qwen2-VL、Yi、Gemma、Baichuan、ChatGLM、Phi 等等。
- **集成方法**：（增量）预训练、（多模态）指令监督微调、奖励模型训练、PPO 训练、DPO 训练、KTO 训练、ORPO 训练等等。
- **多种精度**：16 比特全参数微调、冻结微调、LoRA 微调和基于 AQLM/AWQ/GPTQ/LLM.int8/HQQ/EETQ 的 2/3/4/5/6/8 比特 QLoRA 微调。
- **先进算法**：[GaLore](https://github.com/jiaweizzhao/GaLore)、[BAdam](https://github.com/Ledzy/BAdam)、[Adam-mini](https://github.com/zyushun/Adam-mini)、DoRA、LongLoRA、LLaMA Pro、Mixture-of-Depths、LoRA+、LoftQ、PiSSA 和 Agent 微调。
- **实用技巧**：[FlashAttention-2](https://github.com/Dao-AILab/flash-attention)、[Unsloth](https://github.com/unslothai/unsloth)、[Liger Kernel](https://github.com/linkedin/Liger-Kernel)、RoPE scaling、NEFTune 和 rsLoRA。
- **实验监控**：LlamaBoard、TensorBoard、Wandb、MLflow 等等。
- **极速推理**：基于 vLLM 的 OpenAI 风格 API、浏览器界面和命令行接口。
## 性能指标
与 ChatGLM 官方的 [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/ptuning) 微调相比，LLaMA Factory 的 LoRA 微调提供了 **3.7 倍**的加速比，同时在广告文案生成任务上取得了更高的 Rouge 分数。结合 4 比特量化技术，LLaMA Factory 的 QLoRA 微调进一步降低了 GPU 显存消耗。
![benchmark](assets/benchmark.svg)
<details><summary>变量定义</summary>
- **Training Speed**: 训练阶段每秒处理的样本数量。（批处理大小=4，截断长度=1024）
- **Rouge Score**: [广告文案生成](https://aclanthology.org/D19-1321.pdf)任务验证集上的 Rouge-2 分数。（批处理大小=4，截断长度=1024）
- **GPU Memory**: 4 比特量化训练的 GPU 显存峰值。（批处理大小=1，截断长度=1024）
- 我们在 ChatGLM 的 P-Tuning 中采用 `pre_seq_len=128`，在 LLaMA Factory 的 LoRA 微调中采用 `lora_rank=32`。
</details>
## 更新日志
[24/10/09] 我们支持了从 **[魔乐社区](https://modelers.cn/models)** 下载预训练模型和数据集。详细用法请参照 [此教程](#从魔乐社区下载)。
[24/09/19] 我们支持了 **[Qwen2.5](https://qwenlm.github.io/blog/qwen2.5/)** 模型的微调。
[24/08/30] 我们支持了 **[Qwen2-VL](https://qwenlm.github.io/blog/qwen2-vl/)** 模型的微调。感谢 [@simonJJJ](https://github.com/simonJJJ) 的 PR。
[24/08/27] 我们支持了 **[Liger Kernel](https://github.com/linkedin/Liger-Kernel)**。请使用 `enable_liger_kernel: true` 来加速训练。
[24/08/09] 我们支持了 **[Adam-mini](https://github.com/zyushun/Adam-mini)** 优化器。详细用法请参照 [examples](examples/README_zh.md)。感谢 [@relic-yuexi](https://github.com/relic-yuexi) 的 PR。
<details><summary>展开日志</summary>
[24/07/04] 我们支持了[无污染打包训练](https://github.com/MeetKai/functionary/tree/main/functionary/train/packing)。请使用 `neat_packing: true` 参数。感谢 [@chuan298](https://github.com/chuan298) 的 PR。
[24/06/16] 我们支持了 **[PiSSA](https://arxiv.org/abs/2404.02948)** 算法。详细用法请参照 [examples](examples/README_zh.md)。
[24/06/07] 我们支持了 **[Qwen2](https://qwenlm.github.io/blog/qwen2/)** 和 **[GLM-4](https://github.com/THUDM/GLM-4)** 模型的微调。
[24/05/26] 我们支持了 **[SimPO](https://arxiv.org/abs/2405.14734)** 偏好对齐算法。详细用法请参照 [examples](examples/README_zh.md)。
[24/05/20] 我们支持了 **PaliGemma** 系列模型的微调。注意 PaliGemma 是预训练模型，你需要使用 `paligemma` 模板进行微调使其获得对话能力。
[24/05/18] 我们支持了 **[KTO](https://arxiv.org/abs/2402.01306)** 偏好对齐算法。详细用法请参照 [examples](examples/README_zh.md)。
[24/05/14] 我们支持了昇腾 NPU 设备的训练和推理。详情请查阅[安装](#安装-llama-factory)部分。
[24/04/26] 我们支持了多模态模型 **LLaVA-1.5** 的微调。详细用法请参照 [examples](examples/README_zh.md)。
[24/04/22] 我们提供了在免费 T4 GPU 上微调 Llama-3 模型的 **[Colab 笔记本](https://colab.research.google.com/drive/1d5KQtbemerlSDSxZIfAaWXhKr30QypiK?usp=sharing)**。Hugging Face 社区公开了两个利用 LLaMA Factory 微调的 Llama-3 模型，详情请见 [Llama3-8B-Chinese-Chat](https://huggingface.co/shenzhi-wang/Llama3-8B-Chinese-Chat) 和 [Llama3-Chinese](https://huggingface.co/zhichen/Llama3-Chinese)。
[24/04/21] 我们基于 [AstraMindAI 的仓库](https://github.com/astramind-ai/Mixture-of-depths)支持了 **[混合深度训练](https://arxiv.org/abs/2404.02258)**。详细用法请参照 [examples](examples/README_zh.md)。
[24/04/16] 我们支持了 **[BAdam](https://arxiv.org/abs/2404.02827)** 优化器。详细用法请参照 [examples](examples/README_zh.md)。
[24/04/16] 我们支持了 **[unsloth](https://github.com/unslothai/unsloth)** 的长序列训练（24GB 可训练 Llama-2-7B-56k）。该方法相比 FlashAttention-2 提供了 **117%** 的训练速度和 **50%** 的显存节约。更多数据请见[此页面](https://github.com/hiyouga/LLaMA-Factory/wiki/Performance-comparison)。
[24/03/31] 我们支持了 **[ORPO](https://arxiv.org/abs/2403.07691)**。详细用法请参照 [examples](examples/README_zh.md)。
[24/03/21] 我们的论文 "[LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models](https://arxiv.org/abs/2403.13372)" 可在 arXiv 上查看！
[24/03/20] 我们支持了能在 2x24GB GPU 上微调 70B 模型的 **FSDP+QLoRA**。详细用法请参照 [examples](examples/README_zh.md)。
[24/03/13] 我们支持了 **[LoRA+](https://arxiv.org/abs/2402.12354)**。详细用法请参照 [examples](examples/README_zh.md)。
[24/03/07] 我们支持了 **[GaLore](https://arxiv.org/abs/2403.03507)** 优化器。详细用法请参照 [examples](examples/README_zh.md)。
[24/03/07] 我们集成了 **[vLLM](https://github.com/vllm-project/vllm)** 以实现极速并发推理。请使用 `infer_backend: vllm` 来获得 **270%** 的推理速度。
[24/02/28] 我们支持了 **[DoRA](https://arxiv.org/abs/2402.09353)** 微调。请使用 `use_dora: true` 参数进行 DoRA 微调。
[24/02/15] 我们支持了 [LLaMA Pro](https://github.com/TencentARC/LLaMA-Pro) 提出的**块扩展**方法。详细用法请参照 [examples](examples/README_zh.md)。
[24/02/05] Qwen1.5（Qwen2 测试版）系列模型已在 LLaMA-Factory 中实现微调支持。详情请查阅该[博客页面](https://qwenlm.github.io/zh/blog/qwen1.5/)。
[24/01/18] 我们针对绝大多数模型实现了 **Agent 微调**，微调时指定 `dataset: glaive_toolcall_zh` 即可使模型获得工具调用能力。
[23/12/23] 我们针对 LLaMA, Mistral 和 Yi 模型支持了 **[unsloth](https://github.com/unslothai/unsloth)** 的 LoRA 训练加速。请使用 `use_unsloth: true` 参数启用 unsloth 优化。该方法可提供 **170%** 的训练速度，详情请查阅[此页面](https://github.com/hiyouga/LLaMA-Factory/wiki/Performance-comparison)。
[23/12/12] 我们支持了微调最新的混合专家模型 **[Mixtral 8x7B](https://huggingface.co/mistralai/Mixtral-8x7B-v0.1)**。硬件需求请查阅[此处](#硬件依赖)。
[23/12/01] 我们支持了从 **[魔搭社区](https://modelscope.cn/models)** 下载预训练模型和数据集。详细用法请参照 [此教程](#从魔搭社区下载)。
[23/10/21] 我们支持了 **[NEFTune](https://arxiv.org/abs/2310.05914)** 训练技巧。请使用 `neftune_noise_alpha: 5` 参数启用 NEFTune。
[23/09/27] 我们针对 LLaMA 模型支持了 [LongLoRA](https://github.com/dvlab-research/LongLoRA) 提出的 **$S^2$-Attn**。请使用 `shift_attn: true` 参数以启用该功能。
[23/09/23] 我们在项目中集成了 MMLU、C-Eval 和 CMMLU 评估集。详细用法请参照 [examples](examples/README_zh.md)。
[23/09/10] 我们支持了 **[FlashAttention-2](https://github.com/Dao-AILab/flash-attention)**。如果您使用的是 RTX4090、A100 或 H100 GPU，请使用 `flash_attn: fa2` 参数以启用 FlashAttention-2。
[23/08/12] 我们支持了 **RoPE 插值**来扩展 LLaMA 模型的上下文长度。请使用 `rope_scaling: linear` 参数训练模型或使用 `rope_scaling: dynamic` 参数评估模型。
[23/08/11] 我们支持了指令模型的 **[DPO 训练](https://arxiv.org/abs/2305.18290)**。详细用法请参照 [examples](examples/README_zh.md)。
[23/07/31] 我们支持了**数据流式加载**。请使用 `streaming: true` 和 `max_steps: 10000` 参数来流式加载数据集。
[23/07/29] 我们在 Hugging Face 发布了两个 13B 指令微调模型。详细内容请查阅我们的 Hugging Face 项目（[LLaMA-2](https://huggingface.co/hiyouga/Llama-2-Chinese-13b-chat) / [Baichuan](https://huggingface.co/hiyouga/Baichuan-13B-sft)）。
[23/07/18] 我们开发了支持训练和测试的**浏览器一体化界面**。请使用 `train_web.py` 在您的浏览器中微调模型。感谢 [@KanadeSiina](https://github.com/KanadeSiina) 和 [@codemayq](https://github.com/codemayq) 在该功能开发中付出的努力。
[23/07/09] 我们开源了 **[FastEdit](https://github.com/hiyouga/FastEdit)** ⚡🩹，一个简单易用的、能迅速编辑大模型事实记忆的工具包。如果您感兴趣请关注我们的 [FastEdit](https://github.com/hiyouga/FastEdit) 项目。
[23/06/29] 我们提供了一个**可复现的**指令模型微调示例，详细内容请查阅 [Baichuan-7B-sft](https://huggingface.co/hiyouga/Baichuan-7B-sft)。
[23/06/22] 我们对齐了[示例 API](src/api_demo.py) 与 [OpenAI API](https://platform.openai.com/docs/api-reference/chat) 的格式，您可以将微调模型接入**任意基于 ChatGPT 的应用**中。
[23/06/03] 我们实现了 4 比特的 LoRA 训练（也称 **[QLoRA](https://github.com/artidoro/qlora)**）。详细用法请参照 [examples](examples/README_zh.md)。
</details>
## 模型
| 模型名 | 模型大小 | Template |
| ----------------------------------------------------------------- | -------------------------------- | ---------------- |
| [Baichuan 2](https://huggingface.co/baichuan-inc) | 7B/13B | baichuan2 |
| [BLOOM/BLOOMZ](https://huggingface.co/bigscience) | 560M/1.1B/1.7B/3B/7.1B/176B | - |
| [ChatGLM3](https://huggingface.co/THUDM) | 6B | chatglm3 |
| [Command R](https://huggingface.co/CohereForAI) | 35B/104B | cohere |
| [DeepSeek (Code/MoE)](https://huggingface.co/deepseek-ai) | 7B/16B/67B/236B | deepseek |
| [Falcon](https://huggingface.co/tiiuae) | 7B/11B/40B/180B | falcon |
| [Gemma/Gemma 2/CodeGemma](https://huggingface.co/google) | 2B/7B/9B/27B | gemma |
| [GLM-4](https://huggingface.co/THUDM) | 9B | glm4 |
| [InternLM2/InternLM2.5](https://huggingface.co/internlm) | 7B/20B | intern2 |
| [Llama](https://github.com/facebookresearch/llama) | 7B/13B/33B/65B | - |
| [Llama 2](https://huggingface.co/meta-llama) | 7B/13B/70B | llama2 |
| [Llama 3-3.2](https://huggingface.co/meta-llama) | 1B/3B/8B/70B | llama3 |
| [LLaVA-1.5](https://huggingface.co/llava-hf) | 7B/13B | llava |
| [LLaVA-NeXT](https://huggingface.co/llava-hf) | 7B/8B/13B/34B/72B/110B | llava_next |
| [LLaVA-NeXT-Video](https://huggingface.co/llava-hf) | 7B/34B | llava_next_video |
| [MiniCPM](https://huggingface.co/openbmb) | 1B/2B/4B | cpm/cpm3 |
| [Mistral/Mixtral](https://huggingface.co/mistralai) | 7B/8x7B/8x22B | mistral |
| [OLMo](https://huggingface.co/allenai) | 1B/7B | - |
| [PaliGemma](https://huggingface.co/google) | 3B | paligemma |
| [Phi-1.5/Phi-2](https://huggingface.co/microsoft) | 1.3B/2.7B | - |
| [Phi-3](https://huggingface.co/microsoft) | 4B/7B/14B | phi |
| [Qwen (1-2.5) (Code/Math/MoE)](https://huggingface.co/Qwen) | 0.5B/1.5B/3B/7B/14B/32B/72B/110B | qwen |
| [Qwen2-VL](https://huggingface.co/Qwen) | 2B/7B/72B | qwen2_vl |
| [StarCoder 2](https://huggingface.co/bigcode) | 3B/7B/15B | - |
| [XVERSE](https://huggingface.co/xverse) | 7B/13B/65B | xverse |
| [Yi/Yi-1.5 (Code)](https://huggingface.co/01-ai) | 1.5B/6B/9B/34B | yi |
| [Yi-VL](https://huggingface.co/01-ai) | 6B/34B | yi_vl |
| [Yuan 2](https://huggingface.co/IEITYuan) | 2B/51B/102B | yuan |
> [!NOTE]
> 对于所有“基座”（Base）模型，`template` 参数可以是 `default`, `alpaca`, `vicuna` 等任意值。但“对话”（Instruct/Chat）模型请务必使用**对应的模板**。
>
> 请务必在训练和推理时采用**完全一致**的模板。
项目所支持模型的完整列表请参阅 [constants.py](src/llamafactory/extras/constants.py)。
您也可以在 [template.py](src/llamafactory/data/template.py) 中添加自己的对话模板。
## 训练方法
| 方法 | 全参数训练 | 部分参数训练 | LoRA | QLoRA |
| ---------------------- | ------------------ | ------------------ | ------------------ | ------------------ |
| 预训练 | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| 指令监督微调 | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| 奖励模型训练 | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| PPO 训练 | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| DPO 训练 | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| KTO 训练 | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| ORPO 训练 | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
| SimPO 训练 | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
> [!TIP]
> 有关 PPO 的实现细节，请参考[此博客](https://newfacade.github.io/notes-on-reinforcement-learning/17-ppo-trl.html)。
## 数据集
<details><summary>预训练数据集</summary>
- [Wiki Demo (en)](data/wiki_demo.txt)
- [RefinedWeb (en)](https://huggingface.co/datasets/tiiuae/falcon-refinedweb)
- [RedPajama V2 (en)](https://huggingface.co/datasets/togethercomputer/RedPajama-Data-V2)
- [Wikipedia (en)](https://huggingface.co/datasets/olm/olm-wikipedia-20221220)
- [Wikipedia (zh)](https://huggingface.co/datasets/pleisto/wikipedia-cn-20230720-filtered)
- [Pile (en)](https://huggingface.co/datasets/EleutherAI/pile)
- [SkyPile (zh)](https://huggingface.co/datasets/Skywork/SkyPile-150B)
- [FineWeb (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb)
- [FineWeb-Edu (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu)
- [The Stack (en)](https://huggingface.co/datasets/bigcode/the-stack)
- [StarCoder (en)](https://huggingface.co/datasets/bigcode/starcoderdata)
</details>
<details><summary>指令微调数据集</summary>
- [Identity (en&zh)](data/identity.json)
- [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)
- [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca-3)
- [Alpaca GPT4 (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
- [Glaive Function Calling V2 (en&zh)](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2)
- [LIMA (en)](https://huggingface.co/datasets/GAIR/lima)
- [Guanaco Dataset (multilingual)](https://huggingface.co/datasets/JosephusCheung/GuanacoDataset)
- [BELLE 2M (zh)](https://huggingface.co/datasets/BelleGroup/train_2M_CN)
- [BELLE 1M (zh)](https://huggingface.co/datasets/BelleGroup/train_1M_CN)
- [BELLE 0.5M (zh)](https://huggingface.co/datasets/BelleGroup/train_0.5M_CN)
- [BELLE Dialogue 0.4M (zh)](https://huggingface.co/datasets/BelleGroup/generated_chat_0.4M)
- [BELLE School Math 0.25M (zh)](https://huggingface.co/datasets/BelleGroup/school_math_0.25M)
- [BELLE Multiturn Chat 0.8M (zh)](https://huggingface.co/datasets/BelleGroup/multiturn_chat_0.8M)
- [UltraChat (en)](https://github.com/thunlp/UltraChat)
- [OpenPlatypus (en)](https://huggingface.co/datasets/garage-bAInd/Open-Platypus)
- [CodeAlpaca 20k (en)](https://huggingface.co/datasets/sahil2801/CodeAlpaca-20k)
- [Alpaca CoT (multilingual)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT)
- [OpenOrca (en)](https://huggingface.co/datasets/Open-Orca/OpenOrca)
- [SlimOrca (en)](https://huggingface.co/datasets/Open-Orca/SlimOrca)
- [MathInstruct (en)](https://huggingface.co/datasets/TIGER-Lab/MathInstruct)
- [Firefly 1.1M (zh)](https://huggingface.co/datasets/YeungNLP/firefly-train-1.1M)
- [Wiki QA (en)](https://huggingface.co/datasets/wiki_qa)
- [Web QA (zh)](https://huggingface.co/datasets/suolyer/webqa)
- [WebNovel (zh)](https://huggingface.co/datasets/zxbsmk/webnovel_cn)
- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
- [deepctrl (en&zh)](https://www.modelscope.cn/datasets/deepctrl/deepctrl-sft-data)
- [Advertise Generating (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)
- [ShareGPT Hyperfiltered (en)](https://huggingface.co/datasets/totally-not-an-llm/sharegpt-hyperfiltered-3k)
- [ShareGPT4 (en&zh)](https://huggingface.co/datasets/shibing624/sharegpt_gpt4)
- [UltraChat 200k (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k)
- [AgentInstruct (en)](https://huggingface.co/datasets/THUDM/AgentInstruct)
- [LMSYS Chat 1M (en)](https://huggingface.co/datasets/lmsys/lmsys-chat-1m)
- [Evol Instruct V2 (en)](https://huggingface.co/datasets/WizardLM/WizardLM_evol_instruct_V2_196k)
- [Cosmopedia (en)](https://huggingface.co/datasets/HuggingFaceTB/cosmopedia)
- [STEM (zh)](https://huggingface.co/datasets/hfl/stem_zh_instruction)
- [Ruozhiba (zh)](https://huggingface.co/datasets/hfl/ruozhiba_gpt4_turbo)
- [Neo-sft (zh)](https://huggingface.co/datasets/m-a-p/neo_sft_phase2)
- [WebInstructSub (en)](https://huggingface.co/datasets/TIGER-Lab/WebInstructSub)
- [Magpie-Pro-300K-Filtered (en)](https://huggingface.co/datasets/Magpie-Align/Magpie-Pro-300K-Filtered)
- [Magpie-ultra-v0.1 (en)](https://huggingface.co/datasets/argilla/magpie-ultra-v0.1)
- [LLaVA mixed (en&zh)](https://huggingface.co/datasets/BUAADreamer/llava-en-zh-300k)
- [Pokemon-gpt4o-captions (en&zh)](https://huggingface.co/datasets/jugg1024/pokemon-gpt4o-captions)
- [Open Assistant (de)](https://huggingface.co/datasets/mayflowergmbh/oasst_de)
- [Dolly 15k (de)](https://huggingface.co/datasets/mayflowergmbh/dolly-15k_de)
- [Alpaca GPT4 (de)](https://huggingface.co/datasets/mayflowergmbh/alpaca-gpt4_de)
- [OpenSchnabeltier (de)](https://huggingface.co/datasets/mayflowergmbh/openschnabeltier_de)
- [Evol Instruct (de)](https://huggingface.co/datasets/mayflowergmbh/evol-instruct_de)
- [Dolphin (de)](https://huggingface.co/datasets/mayflowergmbh/dolphin_de)
- [Booksum (de)](https://huggingface.co/datasets/mayflowergmbh/booksum_de)
- [Airoboros (de)](https://huggingface.co/datasets/mayflowergmbh/airoboros-3.0_de)
- [Ultrachat (de)](https://huggingface.co/datasets/mayflowergmbh/ultra-chat_de)
</details>
<details><summary>偏好数据集</summary>
- [DPO mixed (en&zh)](https://huggingface.co/datasets/hiyouga/DPO-En-Zh-20k)
- [UltraFeedback (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized)
- [RLHF-V (en)](https://huggingface.co/datasets/openbmb/RLHF-V-Dataset)
- [VLFeedback (en)](https://huggingface.co/datasets/Zhihui/VLFeedback)
- [Orca DPO Pairs (en)](https://huggingface.co/datasets/Intel/orca_dpo_pairs)
- [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
- [Orca DPO (de)](https://huggingface.co/datasets/mayflowergmbh/intel_orca_dpo_pairs_de)
- [KTO mixed (en)](https://huggingface.co/datasets/argilla/kto-mix-15k)
</details>
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
pip install --upgrade huggingface_hub
huggingface-cli login
```
## 软硬件依赖
| 必需项 | 至少 | 推荐 |
| ------------ | ------- | --------- |
| python | 3.8 | 3.11 |
| torch | 1.13.1 | 2.4.0 |
| transformers | 4.41.2 | 4.43.4 |
| datasets | 2.16.0 | 2.20.0 |
| accelerate | 0.30.1 | 0.32.0 |
| peft | 0.11.1 | 0.12.0 |
| trl | 0.8.6 | 0.9.6 |
| 可选项 | 至少 | 推荐 |
| ------------ | ------- | --------- |
| CUDA | 11.6 | 12.2 |
| deepspeed | 0.10.0 | 0.14.0 |
| bitsandbytes | 0.39.0 | 0.43.1 |
| vllm | 0.4.3 | 0.5.0 |
| flash-attn | 2.3.0 | 2.6.3 |
### 硬件依赖
\* *估算值*
| 方法 | 精度 | 7B | 13B | 30B | 70B | 110B | 8x7B | 8x22B |
| ----------------- | ---- | ----- | ----- | ----- | ------ | ------ | ----- | ------ |
| Full | AMP | 120GB | 240GB | 600GB | 1200GB | 2000GB | 900GB | 2400GB |
| Full | 16 | 60GB | 120GB | 300GB | 600GB | 900GB | 400GB | 1200GB |
| Freeze | 16 | 20GB | 40GB | 80GB | 200GB | 360GB | 160GB | 400GB |
| LoRA/GaLore/BAdam | 16 | 16GB | 32GB | 64GB | 160GB | 240GB | 120GB | 320GB |
| QLoRA | 8 | 10GB | 20GB | 40GB | 80GB | 140GB | 60GB | 160GB |
| QLoRA | 4 | 6GB | 12GB | 24GB | 48GB | 72GB | 30GB | 96GB |
| QLoRA | 2 | 4GB | 8GB | 16GB | 24GB | 48GB | 18GB | 48GB |
## 如何使用
### 安装 LLaMA Factory
> [!IMPORTANT]
> 此步骤为必需。
```bash
git clone --depth 1 https://github.com/hiyouga/LLaMA-Factory.git
cd LLaMA-Factory
pip install -e ".[torch,metrics]"
```
可选的额外依赖项：torch、torch-npu、metrics、deepspeed、liger-kernel、bitsandbytes、hqq、eetq、gptq、awq、aqlm、vllm、galore、badam、adam-mini、qwen、modelscope、openmind、quality
> [!TIP]
> 遇到包冲突时，可使用 `pip install --no-deps -e .` 解决。
<details><summary>Windows 用户指南</summary>
如果要在 Windows 平台上开启量化 LoRA（QLoRA），需要安装预编译的 `bitsandbytes` 库, 支持 CUDA 11.1 到 12.2, 请根据您的 CUDA 版本情况选择适合的[发布版本](https://github.com/jllllll/bitsandbytes-windows-webui/releases/tag/wheels)。
```bash
pip install https://github.com/jllllll/bitsandbytes-windows-webui/releases/download/wheels/bitsandbytes-0.41.2.post2-py3-none-win_amd64.whl
```
如果要在 Windows 平台上开启 FlashAttention-2，需要安装预编译的 `flash-attn` 库，支持 CUDA 12.1 到 12.2，请根据需求到 [flash-attention](https://github.com/bdashore3/flash-attention/releases) 下载对应版本安装。
</details>
<details><summary>昇腾 NPU 用户指南</summary>
在昇腾 NPU 设备上安装 LLaMA Factory 时，需要指定额外依赖项，使用 `pip install -e ".[torch-npu,metrics]"` 命令安装。此外，还需要安装 **[Ascend CANN Toolkit 与 Kernels](https://www.hiascend.com/developer/download/community/result?module=cann)**，安装方法请参考[安装教程](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC2alpha002/quickstart/quickstart/quickstart_18_0004.html)或使用以下命令：
```bash
# 请替换 URL 为 CANN 版本和设备型号对应的 URL
# 安装 CANN Toolkit
wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C17SPC701/Ascend-cann-toolkit_8.0.RC1.alpha001_linux-"$(uname -i)".run
bash Ascend-cann-toolkit_8.0.RC1.alpha001_linux-"$(uname -i)".run --install
# 安装 CANN Kernels
wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C17SPC701/Ascend-cann-kernels-910b_8.0.RC1.alpha001_linux.run
bash Ascend-cann-kernels-910b_8.0.RC1.alpha001_linux.run --install
# 设置环境变量
source /usr/local/Ascend/ascend-toolkit/set_env.sh
```
| 依赖项 | 至少 | 推荐 |
| ------------ | ------- | ----------- |
| CANN | 8.0.RC1 | 8.0.RC1 |
| torch | 2.1.0 | 2.1.0 |
| torch-npu | 2.1.0 | 2.1.0.post3 |
| deepspeed | 0.13.2 | 0.13.2 |
请使用 `ASCEND_RT_VISIBLE_DEVICES` 而非 `CUDA_VISIBLE_DEVICES` 来指定运算设备。
如果遇到无法正常推理的情况，请尝试设置 `do_sample: false`。
下载预构建 Docker 镜像：[32GB](http://mirrors.cn-central-221.ovaijisuan.com/detail/130.html) | [64GB](http://mirrors.cn-central-221.ovaijisuan.com/detail/131.html)
</details>
### 数据准备
关于数据集文件的格式，请参考 [data/README_zh.md](data/README_zh.md) 的内容。你可以使用 HuggingFace / ModelScope / Modelers 上的数据集或加载本地数据集。
> [!NOTE]
> 使用自定义数据集时，请更新 `data/dataset_info.json` 文件。
### 快速开始
下面三行命令分别对 Llama3-8B-Instruct 模型进行 LoRA **微调**、**推理**和**合并**。
```bash
llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
llamafactory-cli chat examples/inference/llama3_lora_sft.yaml
llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
```
高级用法请参考 [examples/README_zh.md](examples/README_zh.md)（包括多 GPU 微调）。
> [!TIP]
> 使用 `llamafactory-cli help` 显示帮助信息。
### LLaMA Board 可视化微调（由 [Gradio](https://github.com/gradio-app/gradio) 驱动）
```bash
llamafactory-cli webui
```
### 构建 Docker
CUDA 用户：
```bash
cd docker/docker-cuda/
docker compose up -d
docker compose exec llamafactory bash
```
昇腾 NPU 用户：
```bash
cd docker/docker-npu/
docker compose up -d
docker compose exec llamafactory bash
```
AMD ROCm 用户：
```bash
cd docker/docker-rocm/
docker compose up -d
docker compose exec llamafactory bash
```
<details><summary>不使用 Docker Compose 构建</summary>
CUDA 用户：
```bash
docker build -f ./docker/docker-cuda/Dockerfile \
--build-arg INSTALL_BNB=false \
--build-arg INSTALL_VLLM=false \
--build-arg INSTALL_DEEPSPEED=false \
--build-arg INSTALL_FLASHATTN=false \
--build-arg PIP_INDEX=https://pypi.org/simple \
-t llamafactory:latest .
docker run -dit --gpus=all \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-p 7860:7860 \
-p 8000:8000 \
--shm-size 16G \
--name llamafactory \
llamafactory:latest
docker exec -it llamafactory bash
```
昇腾 NPU 用户：
```bash
# 根据您的环境选择镜像
docker build -f ./docker/docker-npu/Dockerfile \
--build-arg INSTALL_DEEPSPEED=false \
--build-arg PIP_INDEX=https://pypi.org/simple \
-t llamafactory:latest .
# 根据您的资源更改 `device`
docker run -dit \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-v /usr/local/dcmi:/usr/local/dcmi \
-v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
-v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
-v /etc/ascend_install.info:/etc/ascend_install.info \
-p 7860:7860 \
-p 8000:8000 \
--device /dev/davinci0 \
--device /dev/davinci_manager \
--device /dev/devmm_svm \
--device /dev/hisi_hdc \
--shm-size 16G \
--name llamafactory \
llamafactory:latest
docker exec -it llamafactory bash
```
AMD ROCm 用户：
```bash
docker build -f ./docker/docker-rocm/Dockerfile \
--build-arg INSTALL_BNB=false \
--build-arg INSTALL_VLLM=false \
--build-arg INSTALL_DEEPSPEED=false \
--build-arg INSTALL_FLASHATTN=false \
--build-arg PIP_INDEX=https://pypi.org/simple \
-t llamafactory:latest .
docker run -dit \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-v ./saves:/app/saves \
-p 7860:7860 \
-p 8000:8000 \
--device /dev/kfd \
--device /dev/dri \
--shm-size 16G \
--name llamafactory \
llamafactory:latest
docker exec -it llamafactory bash
```
</details>
<details><summary>数据卷详情</summary>
- `hf_cache`：使用宿主机的 Hugging Face 缓存文件夹，允许更改为新的目录。
- `ms_cache`：类似 Hugging Face 缓存文件夹，为 ModelScope 用户提供。
- `om_cache`：类似 Hugging Face 缓存文件夹，为 Modelers 用户提供。
- `data`：宿主机中存放数据集的文件夹路径。
- `output`：将导出目录设置为该路径后，即可在宿主机中访问导出后的模型。
</details>
### 利用 vLLM 部署 OpenAI API
```bash
API_PORT=8000 llamafactory-cli api examples/inference/llama3_vllm.yaml
```
> [!TIP]
> API 文档请查阅[这里](https://platform.openai.com/docs/api-reference/chat/create)。
### 从魔搭社区下载
如果您在 Hugging Face 模型和数据集的下载中遇到了问题，可以通过下述方法使用魔搭社区。
```bash
export USE_MODELSCOPE_HUB=1 # Windows 使用 `set USE_MODELSCOPE_HUB=1`
```
将 `model_name_or_path` 设置为模型 ID 来加载对应的模型。在[魔搭社区](https://modelscope.cn/models)查看所有可用的模型，例如 `LLM-Research/Meta-Llama-3-8B-Instruct`。
### 从魔乐社区下载
您也可以通过下述方法，使用魔乐社区下载数据集和模型。
```bash
export USE_OPENMIND_HUB=1 # Windows 使用 `set USE_OPENMIND_HUB=1`
```
将 `model_name_or_path` 设置为模型 ID 来加载对应的模型。在[魔乐社区](https://modelers.cn/models)查看所有可用的模型，例如 `TeleAI/TeleChat-7B-pt`。
### 使用 W&B 面板
若要使用 [Weights & Biases](https://wandb.ai) 记录实验数据，请在 yaml 文件中添加下面的参数。
```yaml
report_to: wandb
run_name: test_run # 可选
```
在启动训练任务时，将 `WANDB_API_KEY` 设置为[密钥](https://wandb.ai/authorize)来登录 W&B 账户。
## 使用了 LLaMA Factory 的项目
如果您有项目希望添加至下述列表，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)
1. Yu et al. Open, Closed, or Small Language Models for Text Classification? 2023. [[arxiv]](https://arxiv.org/abs/2308.10092)
1. Wang et al. UbiPhysio: Support Daily Functioning, Fitness, and Rehabilitation with Action Understanding and Feedback in Natural Language. 2023. [[arxiv]](https://arxiv.org/abs/2308.10526)
1. Luceri et al. Leveraging Large Language Models to Detect Influence Campaigns in Social Media. 2023. [[arxiv]](https://arxiv.org/abs/2311.07816)
1. Zhang et al. Alleviating Hallucinations of Large Language Models through Induced Hallucinations. 2023. [[arxiv]](https://arxiv.org/abs/2312.15710)
1. Wang et al. Know Your Needs Better: Towards Structured Understanding of Marketer Demands with Analogical Reasoning Augmented LLMs. KDD 2024. [[arxiv]](https://arxiv.org/abs/2401.04319)
1. Wang et al. CANDLE: Iterative Conceptualization and Instantiation Distillation from Large Language Models for Commonsense Reasoning. ACL 2024. [[arxiv]](https://arxiv.org/abs/2401.07286)
1. Choi et al. FACT-GPT: Fact-Checking Augmentation via Claim Matching with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2402.05904)
1. Zhang et al. AutoMathText: Autonomous Data Selection with Language Models for Mathematical Texts. 2024. [[arxiv]](https://arxiv.org/abs/2402.07625)
1. Lyu et al. KnowTuning: Knowledge-aware Fine-tuning for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11176)
1. Yang et al. LaCo: Large Language Model Pruning via Layer Collaps. 2024. [[arxiv]](https://arxiv.org/abs/2402.11187)
1. Bhardwaj et al. Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task Arithmetic. 2024. [[arxiv]](https://arxiv.org/abs/2402.11746)
1. Yang et al. Enhancing Empathetic Response Generation by Augmenting LLMs with Small-scale Empathetic Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11801)
1. Yi et al. Generation Meets Verification: Accelerating Large Language Model Inference with Smart Parallel Auto-Correct Decoding. ACL 2024 Findings. [[arxiv]](https://arxiv.org/abs/2402.11809)
1. Cao et al. Head-wise Shareable Attention for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11819)
1. Zhang et al. Enhancing Multilingual Capabilities of Large Language Models through Self-Distillation from Resource-Rich Languages. 2024. [[arxiv]](https://arxiv.org/abs/2402.12204)
1. Kim et al. Efficient and Effective Vocabulary Expansion Towards Multilingual Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.14714)
1. Yu et al. KIEval: A Knowledge-grounded Interactive Evaluation Framework for Large Language Models. ACL 2024. [[arxiv]](https://arxiv.org/abs/2402.15043)
1. Huang et al. Key-Point-Driven Data Synthesis with its Enhancement on Mathematical Reasoning. 2024. [[arxiv]](https://arxiv.org/abs/2403.02333)
1. Duan et al. Negating Negatives: Alignment without Human Positive Samples via Distributional Dispreference Optimization. 2024. [[arxiv]](https://arxiv.org/abs/2403.03419)
1. Xie and Schwertfeger. Empowering Robotics with Large Language Models: osmAG Map Comprehension with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2403.08228)
1. Wu et al. Large Language Models are Parallel Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2403.09073)
1. Zhang et al. EDT: Improving Large Language Models' Generation by Entropy-based Dynamic Temperature Sampling. 2024. [[arxiv]](https://arxiv.org/abs/2403.14541)
1. Weller et al. FollowIR: Evaluating and Teaching Information Retrieval Models to Follow Instructions. 2024. [[arxiv]](https://arxiv.org/abs/2403.15246)
1. Hongbin Na. CBT-LLM: A Chinese Large Language Model for Cognitive Behavioral Therapy-based Mental Health Question Answering. COLING 2024. [[arxiv]](https://arxiv.org/abs/2403.16008)
1. Zan et al. CodeS: Natural Language to Code Repository via Multi-Layer Sketch. 2024. [[arxiv]](https://arxiv.org/abs/2403.16443)
1. Liu et al. Extensive Self-Contrast Enables Feedback-Free Language Model Alignment. 2024. [[arxiv]](https://arxiv.org/abs/2404.00604)
1. Luo et al. BAdam: A Memory Efficient Full Parameter Training Method for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.02827)
1. Du et al. Chinese Tiny LLM: Pretraining a Chinese-Centric Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2404.04167)
1. Ma et al. Parameter Efficient Quasi-Orthogonal Fine-Tuning via Givens Rotation. ICML 2024. [[arxiv]](https://arxiv.org/abs/2404.04316)
1. Liu et al. Dynamic Generation of Personalities with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.07084)
1. Shang et al. How Far Have We Gone in Stripped Binary Code Understanding Using Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.09836)
1. Huang et al. LLMTune: Accelerate Database Knob Tuning with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.11581)
1. Deng et al. Text-Tuple-Table: Towards Information Integration in Text-to-Table Generation via Global Tuple Extraction. 2024. [[arxiv]](https://arxiv.org/abs/2404.14215)
1. Acikgoz et al. Hippocrates: An Open-Source Framework for Advancing Large Language Models in Healthcare. 2024. [[arxiv]](https://arxiv.org/abs/2404.16621)
1. Zhang et al. Small Language Models Need Strong Verifiers to Self-Correct Reasoning. ACL 2024 Findings. [[arxiv]](https://arxiv.org/abs/2404.17140)
1. Zhou et al. FREB-TQA: A Fine-Grained Robustness Evaluation Benchmark for Table Question Answering. NAACL 2024. [[arxiv]](https://arxiv.org/abs/2404.18585)
1. Xu et al. Large Language Models for Cyber Security: A Systematic Literature Review. 2024. [[arxiv]](https://arxiv.org/abs/2405.04760)
1. Dammu et al. "They are uncultured": Unveiling Covert Harms and Social Threats in LLM Generated Conversations. 2024. [[arxiv]](https://arxiv.org/abs/2405.05378)
1. Yi et al. A safety realignment framework via subspace-oriented model fusion for large language models. 2024. [[arxiv]](https://arxiv.org/abs/2405.09055)
1. Lou et al. SPO: Multi-Dimensional Preference Sequential Alignment With Implicit Reward Modeling. 2024. [[arxiv]](https://arxiv.org/abs/2405.12739)
1. Zhang et al. Getting More from Less: Large Language Models are Good Spontaneous Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2405.13816)
1. Zhang et al. TS-Align: A Teacher-Student Collaborative Framework for Scalable Iterative Finetuning of Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2405.20215)
1. Zihong Chen. Sentence Segmentation and Sentence Punctuation Based on XunziALLM. 2024. [[paper]](https://aclanthology.org/2024.lt4hala-1.30)
1. Gao et al. The Best of Both Worlds: Toward an Honest and Helpful Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2406.00380)
1. Wang and Song. MARS: Benchmarking the Metaphysical Reasoning Abilities of Language Models with a Multi-task Evaluation Dataset. 2024. [[arxiv]](https://arxiv.org/abs/2406.02106)
1. Hu et al. Computational Limits of Low-Rank Adaptation (LoRA) for Transformer-Based Models. 2024. [[arxiv]](https://arxiv.org/abs/2406.03136)
1. Ge et al. Time Sensitive Knowledge Editing through Efficient Finetuning. ACL 2024. [[arxiv]](https://arxiv.org/abs/2406.04496)
1. Tan et al. Peer Review as A Multi-Turn and Long-Context Dialogue with Role-Based Interactions. 2024. [[arxiv]](https://arxiv.org/abs/2406.05688)
1. Song et al. Turbo Sparse: Achieving LLM SOTA Performance with Minimal Activated Parameters. 2024. [[arxiv]](https://arxiv.org/abs/2406.05955)
1. Gu et al. RWKV-CLIP: A Robust Vision-Language Representation Learner. 2024. [[arxiv]](https://arxiv.org/abs/2406.06973)
1. Chen et al. Advancing Tool-Augmented Large Language Models: Integrating Insights from Errors in Inference Trees. 2024. [[arxiv]](https://arxiv.org/abs/2406.07115)
1. Zhu et al. Are Large Language Models Good Statisticians?. 2024. [[arxiv]](https://arxiv.org/abs/2406.07815)
1. Li et al. Know the Unknown: An Uncertainty-Sensitive Method for LLM Instruction Tuning. 2024. [[arxiv]](https://arxiv.org/abs/2406.10099)
1. Ding et al. IntentionQA: A Benchmark for Evaluating Purchase Intention Comprehension Abilities of Language Models in E-commerce. 2024. [[arxiv]](https://arxiv.org/abs/2406.10173)
1. He et al. COMMUNITY-CROSS-INSTRUCT: Unsupervised Instruction Generation for Aligning Large Language Models to Online Communities. 2024. [[arxiv]](https://arxiv.org/abs/2406.12074)
1. Lin et al. FVEL: Interactive Formal Verification Environment with Large Language Models via Theorem Proving. 2024. [[arxiv]](https://arxiv.org/abs/2406.14408)
1. Treutlein et al. Connecting the Dots: LLMs can Infer and Verbalize Latent Structure from Disparate Training Data. 2024. [[arxiv]](https://arxiv.org/abs/2406.14546)
1. Feng et al. SS-Bench: A Benchmark for Social Story Generation and Evaluation. 2024. [[arxiv]](https://arxiv.org/abs/2406.15695)
1. Feng et al. Self-Constructed Context Decompilation with Fined-grained Alignment Enhancement. 2024. [[arxiv]](https://arxiv.org/abs/2406.17233)
1. Liu et al. Large Language Models for Cuffless Blood Pressure Measurement From Wearable Biosignals. 2024. [[arxiv]](https://arxiv.org/abs/2406.18069)
1. Iyer et al. Exploring Very Low-Resource Translation with LLMs: The University of Edinburgh's Submission to AmericasNLP 2024 Translation Task. AmericasNLP 2024. [[paper]](https://aclanthology.org/2024.americasnlp-1.25)
1. Li et al. Calibrating LLMs with Preference Optimization on Thought Trees for Generating Rationale in Science Question Scoring. 2024. [[arxiv]](https://arxiv.org/abs/2406.19949)
1. Yang et al. Financial Knowledge Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2407.00365)
1. Lin et al. DogeRM: Equipping Reward Models with Domain Knowledge through Model Merging. 2024. [[arxiv]](https://arxiv.org/abs/2407.01470)
1. Bako et al. Evaluating the Semantic Profiling Abilities of LLMs for Natural Language Utterances in Data Visualization. 2024. [[arxiv]](https://arxiv.org/abs/2407.06129)
1. Huang et al. RoLoRA: Fine-tuning Rotated Outlier-free LLMs for Effective Weight-Activation Quantization. 2024. [[arxiv]](https://arxiv.org/abs/2407.08044)
1. Jiang et al. LLM-Collaboration on Automatic Science Journalism for the General Audience. 2024. [[arxiv]](https://arxiv.org/abs/2407.09756)
1. Inouye et al. Applied Auto-tuning on LoRA Hyperparameters. 2024. [[paper]](https://scholarcommons.scu.edu/cseng_senior/272/)
1. Qi et al. Research on Tibetan Tourism Viewpoints information generation system based on LLM. 2024. [[arxiv]](https://arxiv.org/abs/2407.13561)
1. Xu et al. Course-Correction: Safety Alignment Using Synthetic Preferences. 2024. [[arxiv]](https://arxiv.org/abs/2407.16637)
1. Sun et al. LAMBDA: A Large Model Based Data Agent. 2024. [[arxiv]](https://arxiv.org/abs/2407.17535)
1. Zhu et al. CollectiveSFT: Scaling Large Language Models for Chinese Medical Benchmark with Collective Instructions in Healthcare. 2024. [[arxiv]](https://arxiv.org/abs/2407.19705)
1. Yu et al. Correcting Negative Bias in Large Language Models through Negative Attention Score Alignment. 2024. [[arxiv]](https://arxiv.org/abs/2408.00137)
1. Xie et al. The Power of Personalized Datasets: Advancing Chinese Composition Writing for Elementary School through Targeted Model Fine-Tuning. IALP 2024. [[paper]](https://www.asianlp.sg/conferences/ialp2024/proceedings/papers/IALP2024_P055.pdf)
1. Liu et al. Instruct-Code-Llama: Improving Capabilities of Language Model in Competition Level Code Generation by Online Judge Feedback. ICIC 2024. [[paper]](https://link.springer.com/chapter/10.1007/978-981-97-5669-8_11)
1. Wang et al. Cybernetic Sentinels: Unveiling the Impact of Safety Data Selection on Model Security in Supervised Fine-Tuning. ICIC 2024. [[paper]](https://link.springer.com/chapter/10.1007/978-981-97-5669-8_23)
1. Xia et al. Understanding the Performance and Estimating the Cost of LLM Fine-Tuning. 2024. [[arxiv]](https://arxiv.org/abs/2408.04693)
1. Zeng et al. Perceive, Reflect, and Plan: Designing LLM Agent for Goal-Directed City Navigation without Instructions. 2024. [[arxiv]](https://arxiv.org/abs/2408.04168)
1. Xia et al. Using Pre-trained Language Model for Accurate ESG Prediction. FinNLP 2024. [[paper]](https://aclanthology.org/2024.finnlp-2.1/)
1. Liang et al. I-SHEEP: Self-Alignment of LLM from Scratch through an Iterative Self-Enhancement Paradigm. 2024. [[arxiv]](https://arxiv.org/abs/2408.08072)
1. **[StarWhisper](https://github.com/Yu-Yang-Li/StarWhisper)**: 天文大模型 StarWhisper，基于 ChatGLM2-6B 和 Qwen-14B 在天文数据上微调而得。
1. **[DISC-LawLLM](https://github.com/FudanDISC/DISC-LawLLM)**: 中文法律领域大模型 DISC-LawLLM，基于 Baichuan-13B 微调而得，具有法律推理和知识检索能力。
1. **[Sunsimiao](https://github.com/X-D-Lab/Sunsimiao)**: 孙思邈中文医疗大模型 Sumsimiao，基于 Baichuan-7B 和 ChatGLM-6B 在中文医疗数据上微调而得。
1. **[CareGPT](https://github.com/WangRongsheng/CareGPT)**: 医疗大模型项目 CareGPT，基于 LLaMA2-7B 和 Baichuan-13B 在中文医疗数据上微调而得。
1. **[MachineMindset](https://github.com/PKU-YuanGroup/Machine-Mindset/)**：MBTI性格大模型项目，根据数据集与训练方式让任意 LLM 拥有 16 个不同的性格类型。
1. **[Luminia-13B-v3](https://huggingface.co/Nekochu/Luminia-13B-v3)**：一个用于生成 Stable Diffusion 提示词的大型语言模型。[[🤗Demo]](https://huggingface.co/spaces/Nekochu/Luminia-13B_SD_Prompt)
1. **[Chinese-LLaVA-Med](https://github.com/BUAADreamer/Chinese-LLaVA-Med)**：中文多模态医学大模型，基于 LLaVA-1.5-7B 在中文多模态医疗数据上微调而得。
1. **[AutoRE](https://github.com/THUDM/AutoRE)**：基于大语言模型的文档级关系抽取系统。
1. **[NVIDIA RTX AI Toolkit](https://github.com/NVIDIA/RTX-AI-Toolkit)**：在 Windows 主机上利用英伟达 RTX 设备进行大型语言模型微调的开发包。
1. **[LazyLLM](https://github.com/LazyAGI/LazyLLM)**：一个低代码构建多 Agent 大模型应用的开发工具，支持基于 LLaMA Factory 的模型微调.
</details>
## 协议
本仓库的代码依照 [Apache-2.0](LICENSE) 协议开源。
使用模型权重时，请遵循对应的模型协议：[Baichuan 2](https://huggingface.co/baichuan-inc/Baichuan2-7B-Base/blob/main/Community%20License%20for%20Baichuan%202%20Model.pdf) / [BLOOM](https://huggingface.co/spaces/bigscience/license) / [ChatGLM3](https://github.com/THUDM/ChatGLM3/blob/main/MODEL_LICENSE) / [Command R](https://cohere.com/c4ai-cc-by-nc-license) / [DeepSeek](https://github.com/deepseek-ai/DeepSeek-LLM/blob/main/LICENSE-MODEL) / [Falcon](https://huggingface.co/tiiuae/falcon-180B/blob/main/LICENSE.txt) / [Gemma](https://ai.google.dev/gemma/terms) / [GLM-4](https://huggingface.co/THUDM/glm-4-9b/blob/main/LICENSE) / [InternLM2](https://github.com/InternLM/InternLM#license) / [Llama](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md) / [Llama 2 (LLaVA-1.5)](https://ai.meta.com/llama/license/) / [Llama 3](https://llama.meta.com/llama3/license/) / [MiniCPM](https://github.com/OpenBMB/MiniCPM/blob/main/MiniCPM%20Model%20License.md) / [Mistral](LICENSE) / [OLMo](LICENSE) / [Phi-1.5/Phi-2](https://huggingface.co/microsoft/phi-1_5/resolve/main/Research%20License.docx) / [Phi-3](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/blob/main/LICENSE) / [Qwen](https://github.com/QwenLM/Qwen/blob/main/Tongyi%20Qianwen%20LICENSE%20AGREEMENT) / [StarCoder 2](https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement) / [XVERSE](https://github.com/xverse-ai/XVERSE-13B/blob/main/MODEL_LICENSE.pdf) / [Yi](https://huggingface.co/01-ai/Yi-6B/blob/main/LICENSE) / [Yi-1.5](LICENSE) / [Yuan 2](https://github.com/IEIT-Yuan/Yuan-2.0/blob/main/LICENSE-Yuan)
## 引用
如果您觉得此项目有帮助，请考虑以下列格式引用
```bibtex
@inproceedings{zheng2024llamafactory,
title={LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models},
author={Yaowei Zheng and Richong Zhang and Junhao Zhang and Yanhan Ye and Zheyan Luo and Zhangchi Feng and Yongqiang Ma},
booktitle={Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
address={Bangkok, Thailand},
publisher={Association for Computational Linguistics},
year={2024},
url={http://arxiv.org/abs/2403.13372}
}
```
## 致谢
本项目受益于 [PEFT](https://github.com/huggingface/peft)、[TRL](https://github.com/huggingface/trl)、[QLoRA](https://github.com/artidoro/qlora) 和 [FastChat](https://github.com/lm-sys/FastChat)，感谢以上诸位作者的付出。
## Star History
![Star History Chart](https://api.star-history.com/svg?repos=hiyouga/LLaMA-Factory&type=Date)
assets/benchmark.svg 0 → 100644
View file @ 12d5cbac
<?xml version="1.0" encoding="utf-8" standalone="no"?>
<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
"http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="350.696449pt" height="268.034375pt" viewBox="0 0 350.696449 268.034375" xmlns="http://www.w3.org/2000/svg" version="1.1">
<metadata>
<rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
<cc:Work>
<dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
<dc:date>2023-11-18T11:28:03.028228</dc:date>
<dc:format>image/svg+xml</dc:format>
<dc:creator>
<cc:Agent>
<dc:title>Matplotlib v3.7.1, https://matplotlib.org/</dc:title>
</cc:Agent>
</dc:creator>
</cc:Work>
</rdf:RDF>
</metadata>
<defs>
<style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
</defs>
<g id="figure_1">
<g id="patch_1">
<path d="M 0 268.034375
L 350.696449 268.034375
L 350.696449 0
L 0 0
z
" style="fill: #ffffff"/>
</g>
<g id="axes_1">
<g id="patch_2">
<path d="M 7.2 244.078125
L 342 244.078125
L 342 22.318125
L 7.2 22.318125
z
" style="fill: #ffffff"/>
</g>
<g id="matplotlib.axis_1">
<g id="xtick_1">
<g id="line2d_1">
<defs>
<path id="md49eeea5b7" d="M 0 0
L 0 3.5
" style="stroke: #000000; stroke-width: 0.8"/>
</defs>
<g>
<use xlink:href="#md49eeea5b7" x="56.236364" y="244.078125" style="stroke: #000000; stroke-width: 0.8"/>
</g>
</g>
<g id="text_1">
<!-- Training Speed -->
<g transform="translate(14.12777 258.676562) scale(0.1 -0.1)">
<defs>
<path id="DejaVuSans-Bold-54" d="M 31 4666
L 4331 4666
L 4331 3756
L 2784 3756
L 2784 0
L 1581 0
L 1581 3756
L 31 3756
L 31 4666
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-72" d="M 3138 2547
Q 2991 2616 2845 2648
Q 2700 2681 2553 2681
Q 2122 2681 1889 2404
Q 1656 2128 1656 1613
L 1656 0
L 538 0
L 538 3500
L 1656 3500
L 1656 2925
Q 1872 3269 2151 3426
Q 2431 3584 2822 3584
Q 2878 3584 2943 3579
Q 3009 3575 3134 3559
L 3138 2547
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-61" d="M 2106 1575
Q 1756 1575 1579 1456
Q 1403 1338 1403 1106
Q 1403 894 1545 773
Q 1688 653 1941 653
Q 2256 653 2472 879
Q 2688 1106 2688 1447
L 2688 1575
L 2106 1575
z
M 3816 1997
L 3816 0
L 2688 0
L 2688 519
Q 2463 200 2181 54
Q 1900 -91 1497 -91
Q 953 -91 614 226
Q 275 544 275 1050
Q 275 1666 698 1953
Q 1122 2241 2028 2241
L 2688 2241
L 2688 2328
Q 2688 2594 2478 2717
Q 2269 2841 1825 2841
Q 1466 2841 1156 2769
Q 847 2697 581 2553
L 581 3406
Q 941 3494 1303 3539
Q 1666 3584 2028 3584
Q 2975 3584 3395 3211
Q 3816 2838 3816 1997
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-69" d="M 538 3500
L 1656 3500
L 1656 0
L 538 0
L 538 3500
z
M 538 4863
L 1656 4863
L 1656 3950
L 538 3950
L 538 4863
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-6e" d="M 4056 2131
L 4056 0
L 2931 0
L 2931 347
L 2931 1631
Q 2931 2084 2911 2256
Q 2891 2428 2841 2509
Q 2775 2619 2662 2680
Q 2550 2741 2406 2741
Q 2056 2741 1856 2470
Q 1656 2200 1656 1722
L 1656 0
L 538 0
L 538 3500
L 1656 3500
L 1656 2988
Q 1909 3294 2193 3439
Q 2478 3584 2822 3584
Q 3428 3584 3742 3212
Q 4056 2841 4056 2131
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-67" d="M 2919 594
Q 2688 288 2409 144
Q 2131 0 1766 0
Q 1125 0 706 504
Q 288 1009 288 1791
Q 288 2575 706 3076
Q 1125 3578 1766 3578
Q 2131 3578 2409 3434
Q 2688 3291 2919 2981
L 2919 3500
L 4044 3500
L 4044 353
Q 4044 -491 3511 -936
Q 2978 -1381 1966 -1381
Q 1638 -1381 1331 -1331
Q 1025 -1281 716 -1178
L 716 -306
Q 1009 -475 1290 -558
Q 1572 -641 1856 -641
Q 2406 -641 2662 -400
Q 2919 -159 2919 353
L 2919 594
z
M 2181 2772
Q 1834 2772 1640 2515
Q 1447 2259 1447 1791
Q 1447 1309 1634 1061
Q 1822 813 2181 813
Q 2531 813 2725 1069
Q 2919 1325 2919 1791
Q 2919 2259 2725 2515
Q 2531 2772 2181 2772
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-20" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-53" d="M 3834 4519
L 3834 3531
Q 3450 3703 3084 3790
Q 2719 3878 2394 3878
Q 1963 3878 1756 3759
Q 1550 3641 1550 3391
Q 1550 3203 1689 3098
Q 1828 2994 2194 2919
L 2706 2816
Q 3484 2659 3812 2340
Q 4141 2022 4141 1434
Q 4141 663 3683 286
Q 3225 -91 2284 -91
Q 1841 -91 1394 -6
Q 947 78 500 244
L 500 1259
Q 947 1022 1364 901
Q 1781 781 2169 781
Q 2563 781 2772 912
Q 2981 1044 2981 1288
Q 2981 1506 2839 1625
Q 2697 1744 2272 1838
L 1806 1941
Q 1106 2091 782 2419
Q 459 2747 459 3303
Q 459 4000 909 4375
Q 1359 4750 2203 4750
Q 2588 4750 2994 4692
Q 3400 4634 3834 4519
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-70" d="M 1656 506
L 1656 -1331
L 538 -1331
L 538 3500
L 1656 3500
L 1656 2988
Q 1888 3294 2169 3439
Q 2450 3584 2816 3584
Q 3463 3584 3878 3070
Q 4294 2556 4294 1747
Q 4294 938 3878 423
Q 3463 -91 2816 -91
Q 2450 -91 2169 54
Q 1888 200 1656 506
z
M 2400 2772
Q 2041 2772 1848 2508
Q 1656 2244 1656 1747
Q 1656 1250 1848 986
Q 2041 722 2400 722
Q 2759 722 2948 984
Q 3138 1247 3138 1747
Q 3138 2247 2948 2509
Q 2759 2772 2400 2772
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-65" d="M 4031 1759
L 4031 1441
L 1416 1441
Q 1456 1047 1700 850
Q 1944 653 2381 653
Q 2734 653 3104 758
Q 3475 863 3866 1075
L 3866 213
Q 3469 63 3072 -14
Q 2675 -91 2278 -91
Q 1328 -91 801 392
Q 275 875 275 1747
Q 275 2603 792 3093
Q 1309 3584 2216 3584
Q 3041 3584 3536 3087
Q 4031 2591 4031 1759
z
M 2881 2131
Q 2881 2450 2695 2645
Q 2509 2841 2209 2841
Q 1884 2841 1681 2658
Q 1478 2475 1428 2131
L 2881 2131
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-64" d="M 2919 2988
L 2919 4863
L 4044 4863
L 4044 0
L 2919 0
L 2919 506
Q 2688 197 2409 53
Q 2131 -91 1766 -91
Q 1119 -91 703 423
Q 288 938 288 1747
Q 288 2556 703 3070
Q 1119 3584 1766 3584
Q 2128 3584 2408 3439
Q 2688 3294 2919 2988
z
M 2181 722
Q 2541 722 2730 984
Q 2919 1247 2919 1747
Q 2919 2247 2730 2509
Q 2541 2772 2181 2772
Q 1825 2772 1636 2509
Q 1447 2247 1447 1747
Q 1447 1247 1636 984
Q 1825 722 2181 722
z
" transform="scale(0.015625)"/>
</defs>
<use xlink:href="#DejaVuSans-Bold-54"/>
<use xlink:href="#DejaVuSans-Bold-72" x="57.212891"/>
<use xlink:href="#DejaVuSans-Bold-61" x="106.529297"/>
<use xlink:href="#DejaVuSans-Bold-69" x="174.009766"/>
<use xlink:href="#DejaVuSans-Bold-6e" x="208.287109"/>
<use xlink:href="#DejaVuSans-Bold-69" x="279.478516"/>
<use xlink:href="#DejaVuSans-Bold-6e" x="313.755859"/>
<use xlink:href="#DejaVuSans-Bold-67" x="384.947266"/>
<use xlink:href="#DejaVuSans-Bold-20" x="456.529297"/>
<use xlink:href="#DejaVuSans-Bold-53" x="491.34375"/>
<use xlink:href="#DejaVuSans-Bold-70" x="563.365234"/>
<use xlink:href="#DejaVuSans-Bold-65" x="634.947266"/>
<use xlink:href="#DejaVuSans-Bold-65" x="702.769531"/>
<use xlink:href="#DejaVuSans-Bold-64" x="770.591797"/>
</g>
</g>
</g>
<g id="xtick_2">
<g id="line2d_2">
<g>
<use xlink:href="#md49eeea5b7" x="174.6" y="244.078125" style="stroke: #000000; stroke-width: 0.8"/>
</g>
</g>
<g id="text_2">
<!-- Rouge Score -->
<g transform="translate(139.1875 258.598437) scale(0.1 -0.1)">
<defs>
<path id="DejaVuSans-Bold-52" d="M 2297 2597
Q 2675 2597 2839 2737
Q 3003 2878 3003 3200
Q 3003 3519 2839 3656
Q 2675 3794 2297 3794
L 1791 3794
L 1791 2597
L 2297 2597
z
M 1791 1766
L 1791 0
L 588 0
L 588 4666
L 2425 4666
Q 3347 4666 3776 4356
Q 4206 4047 4206 3378
Q 4206 2916 3982 2619
Q 3759 2322 3309 2181
Q 3556 2125 3751 1926
Q 3947 1728 4147 1325
L 4800 0
L 3519 0
L 2950 1159
Q 2778 1509 2601 1637
Q 2425 1766 2131 1766
L 1791 1766
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-6f" d="M 2203 2784
Q 1831 2784 1636 2517
Q 1441 2250 1441 1747
Q 1441 1244 1636 976
Q 1831 709 2203 709
Q 2569 709 2762 976
Q 2956 1244 2956 1747
Q 2956 2250 2762 2517
Q 2569 2784 2203 2784
z
M 2203 3584
Q 3106 3584 3614 3096
Q 4122 2609 4122 1747
Q 4122 884 3614 396
Q 3106 -91 2203 -91
Q 1297 -91 786 396
Q 275 884 275 1747
Q 275 2609 786 3096
Q 1297 3584 2203 3584
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-75" d="M 500 1363
L 500 3500
L 1625 3500
L 1625 3150
Q 1625 2866 1622 2436
Q 1619 2006 1619 1863
Q 1619 1441 1641 1255
Q 1663 1069 1716 984
Q 1784 875 1895 815
Q 2006 756 2150 756
Q 2500 756 2700 1025
Q 2900 1294 2900 1772
L 2900 3500
L 4019 3500
L 4019 0
L 2900 0
L 2900 506
Q 2647 200 2364 54
Q 2081 -91 1741 -91
Q 1134 -91 817 281
Q 500 653 500 1363
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-63" d="M 3366 3391
L 3366 2478
Q 3138 2634 2908 2709
Q 2678 2784 2431 2784
Q 1963 2784 1702 2511
Q 1441 2238 1441 1747
Q 1441 1256 1702 982
Q 1963 709 2431 709
Q 2694 709 2930 787
Q 3166 866 3366 1019
L 3366 103
Q 3103 6 2833 -42
Q 2563 -91 2291 -91
Q 1344 -91 809 395
Q 275 881 275 1747
Q 275 2613 809 3098
Q 1344 3584 2291 3584
Q 2566 3584 2833 3536
Q 3100 3488 3366 3391
z
" transform="scale(0.015625)"/>
</defs>
<use xlink:href="#DejaVuSans-Bold-52"/>
<use xlink:href="#DejaVuSans-Bold-6f" x="77.001953"/>
<use xlink:href="#DejaVuSans-Bold-75" x="145.703125"/>
<use xlink:href="#DejaVuSans-Bold-67" x="216.894531"/>
<use xlink:href="#DejaVuSans-Bold-65" x="288.476562"/>
<use xlink:href="#DejaVuSans-Bold-20" x="356.298828"/>
<use xlink:href="#DejaVuSans-Bold-53" x="391.113281"/>
<use xlink:href="#DejaVuSans-Bold-63" x="463.134766"/>
<use xlink:href="#DejaVuSans-Bold-6f" x="522.412109"/>
<use xlink:href="#DejaVuSans-Bold-72" x="591.113281"/>
<use xlink:href="#DejaVuSans-Bold-65" x="640.429688"/>
</g>
</g>
</g>
<g id="xtick_3">
<g id="line2d_3">
<g>
<use xlink:href="#md49eeea5b7" x="292.963636" y="244.078125" style="stroke: #000000; stroke-width: 0.8"/>
</g>
</g>
<g id="text_3">
<!-- GPU Memory (GB) -->
<g transform="translate(242.430824 258.665625) scale(0.1 -0.1)">
<defs>
<path id="DejaVuSans-Bold-47" d="M 4781 347
Q 4331 128 3847 18
Q 3363 -91 2847 -91
Q 1681 -91 1000 561
Q 319 1213 319 2328
Q 319 3456 1012 4103
Q 1706 4750 2913 4750
Q 3378 4750 3804 4662
Q 4231 4575 4609 4403
L 4609 3438
Q 4219 3659 3833 3768
Q 3447 3878 3059 3878
Q 2341 3878 1952 3476
Q 1563 3075 1563 2328
Q 1563 1588 1938 1184
Q 2313 781 3003 781
Q 3191 781 3352 804
Q 3513 828 3641 878
L 3641 1784
L 2906 1784
L 2906 2591
L 4781 2591
L 4781 347
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-50" d="M 588 4666
L 2584 4666
Q 3475 4666 3951 4270
Q 4428 3875 4428 3144
Q 4428 2409 3951 2014
Q 3475 1619 2584 1619
L 1791 1619
L 1791 0
L 588 0
L 588 4666
z
M 1791 3794
L 1791 2491
L 2456 2491
Q 2806 2491 2997 2661
Q 3188 2831 3188 3144
Q 3188 3456 2997 3625
Q 2806 3794 2456 3794
L 1791 3794
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-55" d="M 588 4666
L 1791 4666
L 1791 1869
Q 1791 1291 1980 1042
Q 2169 794 2597 794
Q 3028 794 3217 1042
Q 3406 1291 3406 1869
L 3406 4666
L 4609 4666
L 4609 1869
Q 4609 878 4112 393
Q 3616 -91 2597 -91
Q 1581 -91 1084 393
Q 588 878 588 1869
L 588 4666
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-4d" d="M 588 4666
L 2119 4666
L 3181 2169
L 4250 4666
L 5778 4666
L 5778 0
L 4641 0
L 4641 3413
L 3566 897
L 2803 897
L 1728 3413
L 1728 0
L 588 0
L 588 4666
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-6d" d="M 3781 2919
Q 3994 3244 4286 3414
Q 4578 3584 4928 3584
Q 5531 3584 5847 3212
Q 6163 2841 6163 2131
L 6163 0
L 5038 0
L 5038 1825
Q 5041 1866 5042 1909
Q 5044 1953 5044 2034
Q 5044 2406 4934 2573
Q 4825 2741 4581 2741
Q 4263 2741 4089 2478
Q 3916 2216 3909 1719
L 3909 0
L 2784 0
L 2784 1825
Q 2784 2406 2684 2573
Q 2584 2741 2328 2741
Q 2006 2741 1831 2477
Q 1656 2213 1656 1722
L 1656 0
L 531 0
L 531 3500
L 1656 3500
L 1656 2988
Q 1863 3284 2130 3434
Q 2397 3584 2719 3584
Q 3081 3584 3359 3409
Q 3638 3234 3781 2919
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-79" d="M 78 3500
L 1197 3500
L 2138 1125
L 2938 3500
L 4056 3500
L 2584 -331
Q 2363 -916 2067 -1148
Q 1772 -1381 1288 -1381
L 641 -1381
L 641 -647
L 991 -647
Q 1275 -647 1404 -556
Q 1534 -466 1606 -231
L 1638 -134
L 78 3500
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-28" d="M 2413 -844
L 1484 -844
Q 1006 -72 778 623
Q 550 1319 550 2003
Q 550 2688 779 3389
Q 1009 4091 1484 4856
L 2413 4856
Q 2013 4116 1813 3408
Q 1613 2700 1613 2009
Q 1613 1319 1811 609
Q 2009 -100 2413 -844
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-42" d="M 2456 2859
Q 2741 2859 2887 2984
Q 3034 3109 3034 3353
Q 3034 3594 2887 3720
Q 2741 3847 2456 3847
L 1791 3847
L 1791 2859
L 2456 2859
z
M 2497 819
Q 2859 819 3042 972
Q 3225 1125 3225 1434
Q 3225 1738 3044 1889
Q 2863 2041 2497 2041
L 1791 2041
L 1791 819
L 2497 819
z
M 3616 2497
Q 4003 2384 4215 2081
Q 4428 1778 4428 1338
Q 4428 663 3972 331
Q 3516 0 2584 0
L 588 0
L 588 4666
L 2394 4666
Q 3366 4666 3802 4372
Q 4238 4078 4238 3431
Q 4238 3091 4078 2852
Q 3919 2613 3616 2497
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-29" d="M 513 -844
Q 913 -100 1113 609
Q 1313 1319 1313 2009
Q 1313 2700 1113 3408
Q 913 4116 513 4856
L 1441 4856
Q 1916 4091 2145 3389
Q 2375 2688 2375 2003
Q 2375 1319 2147 623
Q 1919 -72 1441 -844
L 513 -844
z
" transform="scale(0.015625)"/>
</defs>
<use xlink:href="#DejaVuSans-Bold-47"/>
<use xlink:href="#DejaVuSans-Bold-50" x="82.080078"/>
<use xlink:href="#DejaVuSans-Bold-55" x="155.371094"/>
<use xlink:href="#DejaVuSans-Bold-20" x="236.572266"/>
<use xlink:href="#DejaVuSans-Bold-4d" x="271.386719"/>
<use xlink:href="#DejaVuSans-Bold-65" x="370.898438"/>
<use xlink:href="#DejaVuSans-Bold-6d" x="438.720703"/>
<use xlink:href="#DejaVuSans-Bold-6f" x="542.919922"/>
<use xlink:href="#DejaVuSans-Bold-72" x="611.621094"/>
<use xlink:href="#DejaVuSans-Bold-79" x="660.9375"/>
<use xlink:href="#DejaVuSans-Bold-20" x="726.123047"/>
<use xlink:href="#DejaVuSans-Bold-28" x="760.9375"/>
<use xlink:href="#DejaVuSans-Bold-47" x="806.640625"/>
<use xlink:href="#DejaVuSans-Bold-42" x="888.720703"/>
<use xlink:href="#DejaVuSans-Bold-29" x="964.941406"/>
</g>
</g>
</g>
</g>
<g id="patch_3">
<path d="M 22.418182 244.078125
L 56.236364 244.078125
L 56.236364 195.339663
L 22.418182 195.339663
z
" clip-path="url(#p080f205d85)" style="fill: #6baed6"/>
</g>
<g id="patch_4">
<path d="M 140.781818 244.078125
L 174.6 244.078125
L 174.6 146.601202
L 140.781818 146.601202
z
" clip-path="url(#p080f205d85)" style="fill: #6baed6"/>
</g>
<g id="patch_5">
<path d="M 259.145455 244.078125
L 292.963636 244.078125
L 292.963636 205.087356
L 259.145455 205.087356
z
" clip-path="url(#p080f205d85)" style="fill: #6baed6"/>
</g>
<g id="patch_6">
<path d="M 56.236364 244.078125
L 90.054545 244.078125
L 90.054545 32.878125
L 56.236364 32.878125
z
" clip-path="url(#p080f205d85)" style="fill: #3182bd"/>
</g>
<g id="patch_7">
<path d="M 174.6 244.078125
L 208.418182 244.078125
L 208.418182 130.355048
L 174.6 130.355048
z
" clip-path="url(#p080f205d85)" style="fill: #3182bd"/>
</g>
<g id="patch_8">
<path d="M 292.963636 244.078125
L 326.781818 244.078125
L 326.781818 218.084279
L 292.963636 218.084279
z
" clip-path="url(#p080f205d85)" style="fill: #3182bd"/>
</g>
<g id="patch_9">
<path d="M 7.2 244.078125
L 342 244.078125
" style="fill: none; stroke: #dddddd; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
</g>
<g id="text_4">
<!-- 5.81 -->
<g transform="translate(26.991335 193.259976) scale(0.1 -0.1)">
<defs>
<path id="DejaVuSans-Bold-35" d="M 678 4666
L 3669 4666
L 3669 3781
L 1638 3781
L 1638 3059
Q 1775 3097 1914 3117
Q 2053 3138 2203 3138
Q 3056 3138 3531 2711
Q 4006 2284 4006 1522
Q 4006 766 3489 337
Q 2972 -91 2053 -91
Q 1656 -91 1267 -14
Q 878 63 494 219
L 494 1166
Q 875 947 1217 837
Q 1559 728 1863 728
Q 2300 728 2551 942
Q 2803 1156 2803 1522
Q 2803 1891 2551 2103
Q 2300 2316 1863 2316
Q 1603 2316 1309 2248
Q 1016 2181 678 2041
L 678 4666
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-2e" d="M 653 1209
L 1778 1209
L 1778 0
L 653 0
L 653 1209
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-38" d="M 2228 2088
Q 1891 2088 1709 1903
Q 1528 1719 1528 1375
Q 1528 1031 1709 848
Q 1891 666 2228 666
Q 2563 666 2741 848
Q 2919 1031 2919 1375
Q 2919 1722 2741 1905
Q 2563 2088 2228 2088
z
M 1350 2484
Q 925 2613 709 2878
Q 494 3144 494 3541
Q 494 4131 934 4440
Q 1375 4750 2228 4750
Q 3075 4750 3515 4442
Q 3956 4134 3956 3541
Q 3956 3144 3739 2878
Q 3522 2613 3097 2484
Q 3572 2353 3814 2058
Q 4056 1763 4056 1313
Q 4056 619 3595 264
Q 3134 -91 2228 -91
Q 1319 -91 855 264
Q 391 619 391 1313
Q 391 1763 633 2058
Q 875 2353 1350 2484
z
M 1631 3419
Q 1631 3141 1786 2991
Q 1941 2841 2228 2841
Q 2509 2841 2662 2991
Q 2816 3141 2816 3419
Q 2816 3697 2662 3845
Q 2509 3994 2228 3994
Q 1941 3994 1786 3844
Q 1631 3694 1631 3419
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-31" d="M 750 831
L 1813 831
L 1813 3847
L 722 3622
L 722 4441
L 1806 4666
L 2950 4666
L 2950 831
L 4013 831
L 4013 0
L 750 0
L 750 831
z
" transform="scale(0.015625)"/>
</defs>
<use xlink:href="#DejaVuSans-Bold-35"/>
<use xlink:href="#DejaVuSans-Bold-2e" x="69.580078"/>
<use xlink:href="#DejaVuSans-Bold-38" x="107.568359"/>
<use xlink:href="#DejaVuSans-Bold-31" x="177.148438"/>
</g>
</g>
<g id="text_5">
<!-- 7.20 -->
<g transform="translate(145.354972 144.521514) scale(0.1 -0.1)">
<defs>
<path id="DejaVuSans-Bold-37" d="M 428 4666
L 3944 4666
L 3944 3988
L 2125 0
L 953 0
L 2675 3781
L 428 3781
L 428 4666
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-32" d="M 1844 884
L 3897 884
L 3897 0
L 506 0
L 506 884
L 2209 2388
Q 2438 2594 2547 2791
Q 2656 2988 2656 3200
Q 2656 3528 2436 3728
Q 2216 3928 1850 3928
Q 1569 3928 1234 3808
Q 900 3688 519 3450
L 519 4475
Q 925 4609 1322 4679
Q 1719 4750 2100 4750
Q 2938 4750 3402 4381
Q 3866 4013 3866 3353
Q 3866 2972 3669 2642
Q 3472 2313 2841 1759
L 1844 884
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-30" d="M 2944 2338
Q 2944 3213 2780 3570
Q 2616 3928 2228 3928
Q 1841 3928 1675 3570
Q 1509 3213 1509 2338
Q 1509 1453 1675 1090
Q 1841 728 2228 728
Q 2613 728 2778 1090
Q 2944 1453 2944 2338
z
M 4147 2328
Q 4147 1169 3647 539
Q 3147 -91 2228 -91
Q 1306 -91 806 539
Q 306 1169 306 2328
Q 306 3491 806 4120
Q 1306 4750 2228 4750
Q 3147 4750 3647 4120
Q 4147 3491 4147 2328
z
" transform="scale(0.015625)"/>
</defs>
<use xlink:href="#DejaVuSans-Bold-37"/>
<use xlink:href="#DejaVuSans-Bold-2e" x="69.580078"/>
<use xlink:href="#DejaVuSans-Bold-32" x="107.568359"/>
<use xlink:href="#DejaVuSans-Bold-30" x="177.148438"/>
</g>
</g>
<g id="text_6">
<!-- 5.78 -->
<g transform="translate(263.718608 203.007668) scale(0.1 -0.1)">
<use xlink:href="#DejaVuSans-Bold-35"/>
<use xlink:href="#DejaVuSans-Bold-2e" x="69.580078"/>
<use xlink:href="#DejaVuSans-Bold-37" x="107.568359"/>
<use xlink:href="#DejaVuSans-Bold-38" x="177.148438"/>
</g>
</g>
<g id="text_7">
<!-- 21.67 -->
<g transform="translate(57.330611 30.798438) scale(0.1 -0.1)">
<defs>
<path id="DejaVuSans-Bold-36" d="M 2316 2303
Q 2000 2303 1842 2098
Q 1684 1894 1684 1484
Q 1684 1075 1842 870
Q 2000 666 2316 666
Q 2634 666 2792 870
Q 2950 1075 2950 1484
Q 2950 1894 2792 2098
Q 2634 2303 2316 2303
z
M 3803 4544
L 3803 3681
Q 3506 3822 3243 3889
Q 2981 3956 2731 3956
Q 2194 3956 1894 3657
Q 1594 3359 1544 2772
Q 1750 2925 1990 3001
Q 2231 3078 2516 3078
Q 3231 3078 3670 2659
Q 4109 2241 4109 1563
Q 4109 813 3618 361
Q 3128 -91 2303 -91
Q 1394 -91 895 523
Q 397 1138 397 2266
Q 397 3422 980 4083
Q 1563 4744 2578 4744
Q 2900 4744 3203 4694
Q 3506 4644 3803 4544
z
" transform="scale(0.015625)"/>
</defs>
<use xlink:href="#DejaVuSans-Bold-32"/>
<use xlink:href="#DejaVuSans-Bold-31" x="69.580078"/>
<use xlink:href="#DejaVuSans-Bold-2e" x="139.160156"/>
<use xlink:href="#DejaVuSans-Bold-36" x="177.148438"/>
<use xlink:href="#DejaVuSans-Bold-37" x="246.728516"/>
</g>
</g>
<g id="text_8">
<!-- 7.36 -->
<g transform="translate(179.173153 128.275361) scale(0.1 -0.1)">
<defs>
<path id="DejaVuSans-Bold-33" d="M 2981 2516
Q 3453 2394 3698 2092
Q 3944 1791 3944 1325
Q 3944 631 3412 270
Q 2881 -91 1863 -91
Q 1503 -91 1142 -33
Q 781 25 428 141
L 428 1069
Q 766 900 1098 814
Q 1431 728 1753 728
Q 2231 728 2486 893
Q 2741 1059 2741 1369
Q 2741 1688 2480 1852
Q 2219 2016 1709 2016
L 1228 2016
L 1228 2791
L 1734 2791
Q 2188 2791 2409 2933
Q 2631 3075 2631 3366
Q 2631 3634 2415 3781
Q 2200 3928 1806 3928
Q 1516 3928 1219 3862
Q 922 3797 628 3669
L 628 4550
Q 984 4650 1334 4700
Q 1684 4750 2022 4750
Q 2931 4750 3382 4451
Q 3834 4153 3834 3553
Q 3834 3144 3618 2883
Q 3403 2622 2981 2516
z
" transform="scale(0.015625)"/>
</defs>
<use xlink:href="#DejaVuSans-Bold-37"/>
<use xlink:href="#DejaVuSans-Bold-2e" x="69.580078"/>
<use xlink:href="#DejaVuSans-Bold-33" x="107.568359"/>
<use xlink:href="#DejaVuSans-Bold-36" x="177.148438"/>
</g>
</g>
<g id="text_9">
<!-- 5.14 -->
<g transform="translate(297.53679 216.004591) scale(0.1 -0.1)">
<defs>
<path id="DejaVuSans-Bold-34" d="M 2356 3675
L 1038 1722
L 2356 1722
L 2356 3675
z
M 2156 4666
L 3494 4666
L 3494 1722
L 4159 1722
L 4159 850
L 3494 850
L 3494 0
L 2356 0
L 2356 850
L 288 850
L 288 1881
L 2156 4666
z
" transform="scale(0.015625)"/>
</defs>
<use xlink:href="#DejaVuSans-Bold-35"/>
<use xlink:href="#DejaVuSans-Bold-2e" x="69.580078"/>
<use xlink:href="#DejaVuSans-Bold-31" x="107.568359"/>
<use xlink:href="#DejaVuSans-Bold-34" x="177.148438"/>
</g>
</g>
<g id="text_10">
<!-- ChatGLM2-6B - - 1×A100 -->
<g transform="translate(93.349688 16.318125) scale(0.12 -0.12)">
<defs>
<path id="DejaVuSans-Bold-43" d="M 4288 256
Q 3956 84 3597 -3
Q 3238 -91 2847 -91
Q 1681 -91 1000 561
Q 319 1213 319 2328
Q 319 3447 1000 4098
Q 1681 4750 2847 4750
Q 3238 4750 3597 4662
Q 3956 4575 4288 4403
L 4288 3438
Q 3953 3666 3628 3772
Q 3303 3878 2944 3878
Q 2300 3878 1931 3465
Q 1563 3053 1563 2328
Q 1563 1606 1931 1193
Q 2300 781 2944 781
Q 3303 781 3628 887
Q 3953 994 4288 1222
L 4288 256
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-68" d="M 4056 2131
L 4056 0
L 2931 0
L 2931 347
L 2931 1625
Q 2931 2084 2911 2256
Q 2891 2428 2841 2509
Q 2775 2619 2662 2680
Q 2550 2741 2406 2741
Q 2056 2741 1856 2470
Q 1656 2200 1656 1722
L 1656 0
L 538 0
L 538 4863
L 1656 4863
L 1656 2988
Q 1909 3294 2193 3439
Q 2478 3584 2822 3584
Q 3428 3584 3742 3212
Q 4056 2841 4056 2131
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-74" d="M 1759 4494
L 1759 3500
L 2913 3500
L 2913 2700
L 1759 2700
L 1759 1216
Q 1759 972 1856 886
Q 1953 800 2241 800
L 2816 800
L 2816 0
L 1856 0
Q 1194 0 917 276
Q 641 553 641 1216
L 641 2700
L 84 2700
L 84 3500
L 641 3500
L 641 4494
L 1759 4494
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-4c" d="M 588 4666
L 1791 4666
L 1791 909
L 3903 909
L 3903 0
L 588 0
L 588 4666
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-2d" d="M 347 2297
L 2309 2297
L 2309 1388
L 347 1388
L 347 2297
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-d7" d="M 4563 3359
L 3206 2003
L 4563 653
L 4038 128
L 2681 1478
L 1325 128
L 800 653
L 2156 2003
L 800 3359
L 1325 3884
L 2681 2528
L 4038 3884
L 4563 3359
z
" transform="scale(0.015625)"/>
<path id="DejaVuSans-Bold-41" d="M 3419 850
L 1538 850
L 1241 0
L 31 0
L 1759 4666
L 3194 4666
L 4922 0
L 3713 0
L 3419 850
z
M 1838 1716
L 3116 1716
L 2478 3572
L 1838 1716
z
" transform="scale(0.015625)"/>
</defs>
<use xlink:href="#DejaVuSans-Bold-43"/>
<use xlink:href="#DejaVuSans-Bold-68" x="73.388672"/>
<use xlink:href="#DejaVuSans-Bold-61" x="144.580078"/>
<use xlink:href="#DejaVuSans-Bold-74" x="212.060547"/>
<use xlink:href="#DejaVuSans-Bold-47" x="259.863281"/>
<use xlink:href="#DejaVuSans-Bold-4c" x="341.943359"/>
<use xlink:href="#DejaVuSans-Bold-4d" x="405.664062"/>
<use xlink:href="#DejaVuSans-Bold-32" x="505.175781"/>
<use xlink:href="#DejaVuSans-Bold-2d" x="574.755859"/>
<use xlink:href="#DejaVuSans-Bold-36" x="616.259766"/>
<use xlink:href="#DejaVuSans-Bold-42" x="685.839844"/>
<use xlink:href="#DejaVuSans-Bold-20" x="762.060547"/>
<use xlink:href="#DejaVuSans-Bold-2d" x="796.875"/>
<use xlink:href="#DejaVuSans-Bold-2d" x="838.378906"/>
<use xlink:href="#DejaVuSans-Bold-20" x="879.882812"/>
<use xlink:href="#DejaVuSans-Bold-31" x="914.697266"/>
<use xlink:href="#DejaVuSans-Bold-d7" x="984.277344"/>
<use xlink:href="#DejaVuSans-Bold-41" x="1068.066406"/>
<use xlink:href="#DejaVuSans-Bold-31" x="1145.458984"/>
<use xlink:href="#DejaVuSans-Bold-30" x="1215.039062"/>
<use xlink:href="#DejaVuSans-Bold-30" x="1284.619141"/>
</g>
</g>
<g id="legend_1">
<g id="patch_10">
<path d="M 201.507812 59.830625
L 335 59.830625
Q 337 59.830625 337 57.830625
L 337 29.318125
Q 337 27.318125 335 27.318125
L 201.507812 27.318125
Q 199.507812 27.318125 199.507812 29.318125
L 199.507812 57.830625
Q 199.507812 59.830625 201.507812 59.830625
L 201.507812 59.830625
z
" style="fill: none; opacity: 0"/>
</g>
<g id="patch_11">
<path d="M 203.507812 38.916562
L 223.507812 38.916562
L 223.507812 31.916562
L 203.507812 31.916562
z
" style="fill: #6baed6"/>
</g>
<g id="text_11">
<!-- ChatGLM P-Tuning -->
<g transform="translate(231.507812 38.916562) scale(0.1 -0.1)">
<use xlink:href="#DejaVuSans-Bold-43"/>
<use xlink:href="#DejaVuSans-Bold-68" x="73.388672"/>
<use xlink:href="#DejaVuSans-Bold-61" x="144.580078"/>
<use xlink:href="#DejaVuSans-Bold-74" x="212.060547"/>
<use xlink:href="#DejaVuSans-Bold-47" x="259.863281"/>
<use xlink:href="#DejaVuSans-Bold-4c" x="341.943359"/>
<use xlink:href="#DejaVuSans-Bold-4d" x="405.664062"/>
<use xlink:href="#DejaVuSans-Bold-20" x="505.175781"/>
<use xlink:href="#DejaVuSans-Bold-50" x="539.990234"/>
<use xlink:href="#DejaVuSans-Bold-2d" x="611.53125"/>
<use xlink:href="#DejaVuSans-Bold-54" x="638.285156"/>
<use xlink:href="#DejaVuSans-Bold-75" x="695.498047"/>
<use xlink:href="#DejaVuSans-Bold-6e" x="766.689453"/>
<use xlink:href="#DejaVuSans-Bold-69" x="837.880859"/>
<use xlink:href="#DejaVuSans-Bold-6e" x="872.158203"/>
<use xlink:href="#DejaVuSans-Bold-67" x="943.349609"/>
</g>
</g>
<g id="patch_12">
<path d="M 203.507812 53.672812
L 223.507812 53.672812
L 223.507812 46.672812
L 203.507812 46.672812
z
" style="fill: #3182bd"/>
</g>
<g id="text_12">
<!-- LLaMA-Factory -->
<g transform="translate(231.507812 53.672812) scale(0.1 -0.1)">
<defs>
<path id="DejaVuSans-Bold-46" d="M 588 4666
L 3834 4666
L 3834 3756
L 1791 3756
L 1791 2888
L 3713 2888
L 3713 1978
L 1791 1978
L 1791 0
L 588 0
L 588 4666
z
" transform="scale(0.015625)"/>
</defs>
<use xlink:href="#DejaVuSans-Bold-4c"/>
<use xlink:href="#DejaVuSans-Bold-4c" x="63.720703"/>
<use xlink:href="#DejaVuSans-Bold-61" x="127.441406"/>
<use xlink:href="#DejaVuSans-Bold-4d" x="194.921875"/>
<use xlink:href="#DejaVuSans-Bold-41" x="294.433594"/>
<use xlink:href="#DejaVuSans-Bold-2d" x="371.826172"/>
<use xlink:href="#DejaVuSans-Bold-46" x="413.330078"/>
<use xlink:href="#DejaVuSans-Bold-61" x="475.765625"/>
<use xlink:href="#DejaVuSans-Bold-63" x="543.246094"/>
<use xlink:href="#DejaVuSans-Bold-74" x="602.523438"/>
<use xlink:href="#DejaVuSans-Bold-6f" x="650.326172"/>
<use xlink:href="#DejaVuSans-Bold-72" x="719.027344"/>
<use xlink:href="#DejaVuSans-Bold-79" x="768.34375"/>
</g>
</g>
</g>
</g>
</g>
<defs>
<clipPath id="p080f205d85">
<rect x="7.2" y="22.318125" width="334.8" height="221.76"/>
</clipPath>
</defs>
</svg>
data/README.md 0 → 100644
View file @ 12d5cbac
The [dataset_info.json](dataset_info.json) contains all available datasets. If you are using a custom dataset, please **make sure** to add a *dataset description* in `dataset_info.json` and specify `dataset: dataset_name` before training to use it.
Currently we support datasets in **alpaca** and **sharegpt** format.
```json
"dataset_name": {
"hf_hub_url": "the name of the dataset repository on the Hugging Face hub. (if specified, ignore script_url and file_name)",
"ms_hub_url": "the name of the dataset repository on the Model Scope hub. (if specified, ignore script_url and file_name)",
"script_url": "the name of the directory containing a dataset loading script. (if specified, ignore file_name)",
"file_name": "the name of the dataset folder or dataset file in this directory. (required if above are not specified)",
"formatting": "the format of the dataset. (optional, default: alpaca, can be chosen from {alpaca, sharegpt})",
"ranking": "whether the dataset is a preference dataset or not. (default: False)",
"subset": "the name of the subset. (optional, default: None)",
"split": "the name of dataset split to be used. (optional, default: train)",
"folder": "the name of the folder of the dataset repository on the Hugging Face hub. (optional, default: None)",
"num_samples": "the number of samples in the dataset to be used. (optional, default: None)",
"columns (optional)": {
"prompt": "the column name in the dataset containing the prompts. (default: instruction)",
"query": "the column name in the dataset containing the queries. (default: input)",
"response": "the column name in the dataset containing the responses. (default: output)",
"history": "the column name in the dataset containing the histories. (default: None)",
"messages": "the column name in the dataset containing the messages. (default: conversations)",
"system": "the column name in the dataset containing the system prompts. (default: None)",
"tools": "the column name in the dataset containing the tool description. (default: None)",
"images": "the column name in the dataset containing the image inputs. (default: None)",
"videos": "the column name in the dataset containing the videos inputs. (default: None)",
"chosen": "the column name in the dataset containing the chosen answers. (default: None)",
"rejected": "the column name in the dataset containing the rejected answers. (default: None)",
"kto_tag": "the column name in the dataset containing the kto tags. (default: None)"
},
"tags (optional, used for the sharegpt format)": {
"role_tag": "the key in the message represents the identity. (default: from)",
"content_tag": "the key in the message represents the content. (default: value)",
"user_tag": "the value of the role_tag represents the user. (default: human)",
"assistant_tag": "the value of the role_tag represents the assistant. (default: gpt)",
"observation_tag": "the value of the role_tag represents the tool results. (default: observation)",
"function_tag": "the value of the role_tag represents the function call. (default: function_call)",
"system_tag": "the value of the role_tag represents the system prompt. (default: system, can override system column)"
}
}
```
## Alpaca Format
### Supervised Fine-Tuning Dataset
* [Example dataset](alpaca_en_demo.json)
In supervised fine-tuning, the `instruction` column will be concatenated with the `input` column and used as the human prompt, then the human prompt would be `instruction\ninput`. The `output` column represents the model response.
The `system` column will be used as the system prompt if specified.
The `history` column is a list consisting of string tuples representing prompt-response pairs in the history messages. Note that the responses in the history **will also be learned by the model** in supervised fine-tuning.
```json
[
{
"instruction": "human instruction (required)",
"input": "human input (optional)",
"output": "model response (required)",
"system": "system prompt (optional)",
"history": [
["human instruction in the first round (optional)", "model response in the first round (optional)"],
["human instruction in the second round (optional)", "model response in the second round (optional)"]
]
}
]
```
Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
```json
"dataset_name": {
"file_name": "data.json",
"columns": {
"prompt": "instruction",
"query": "input",
"response": "output",
"system": "system",
"history": "history"
}
}
```
### Pre-training Dataset
- [Example dataset](c4_demo.json)
In pre-training, only the `text` column will be used for model learning.
```json
[
{"text": "document"},
{"text": "document"}
]
```
Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
```json
"dataset_name": {
"file_name": "data.json",
"columns": {
"prompt": "text"
}
}
```
### Preference Dataset
Preference datasets are used for reward modeling, DPO training, ORPO and SimPO training.
It requires a better response in `chosen` column and a worse response in `rejected` column.
```json
[
{
"instruction": "human instruction (required)",
"input": "human input (optional)",
"chosen": "chosen answer (required)",
"rejected": "rejected answer (required)"
}
]
```
Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
```json
"dataset_name": {
"file_name": "data.json",
"ranking": true,
"columns": {
"prompt": "instruction",
"query": "input",
"chosen": "chosen",
"rejected": "rejected"
}
}
```
### KTO Dataset
An additional column `kto_tag` is required. Please refer to the [sharegpt](#sharegpt-format) format for details.
### Multimodal Image Dataset
An additional column `images` is required. Please refer to the [sharegpt](#sharegpt-format) format for details.
### Multimodal Video Dataset
An additional column `videos` is required. Please refer to the [sharegpt](#sharegpt-format) format for details.
## Sharegpt Format
### Supervised Fine-Tuning Dataset
- [Example dataset](glaive_toolcall_en_demo.json)
Compared to the alpaca format, the sharegpt format allows the datasets have **more roles**, such as human, gpt, observation and function. They are presented in a list of objects in the `conversations` column.
Note that the human and observation should appear in odd positions, while gpt and function should appear in even positions.
```json
[
{
"conversations": [
{
"from": "human",
"value": "human instruction"
},
{
"from": "function_call",
"value": "tool arguments"
},
{
"from": "observation",
"value": "tool result"
},
{
"from": "gpt",
"value": "model response"
}
],
"system": "system prompt (optional)",
"tools": "tool description (optional)"
}
]
```
Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
```json
"dataset_name": {
"file_name": "data.json",
"formatting": "sharegpt",
"columns": {
"messages": "conversations",
"system": "system",
"tools": "tools"
}
}
```
### Pre-training Dataset
Not yet supported, please use the [alpaca](#alpaca-format) format.
### Preference Dataset
- [Example dataset](dpo_en_demo.json)
Preference datasets in sharegpt format also require a better message in `chosen` column and a worse message in `rejected` column.
```json
[
{
"conversations": [
{
"from": "human",
"value": "human instruction"
},
{
"from": "gpt",
"value": "model response"
},
{
"from": "human",
"value": "human instruction"
}
],
"chosen": {
"from": "gpt",
"value": "chosen answer (required)"
},
"rejected": {
"from": "gpt",
"value": "rejected answer (required)"
}
}
]
```
Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
```json
"dataset_name": {
"file_name": "data.json",
"formatting": "sharegpt",
"ranking": true,
"columns": {
"messages": "conversations",
"chosen": "chosen",
"rejected": "rejected"
}
}
```
### KTO Dataset
- [Example dataset](kto_en_demo.json)
KTO datasets require a extra `kto_tag` column containing the boolean human feedback.
```json
[
{
"conversations": [
{
"from": "human",
"value": "human instruction"
},
{
"from": "gpt",
"value": "model response"
}
],
"kto_tag": "human feedback [true/false] (required)"
}
]
```
Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
```json
"dataset_name": {
"file_name": "data.json",
"formatting": "sharegpt",
"columns": {
"messages": "conversations",
"kto_tag": "kto_tag"
}
}
```
### Multimodal Image Dataset
- [Example dataset](mllm_demo.json)
Multimodal image datasets require a `images` column containing the paths to the input images.
The number of images should be identical to the `<image>` tokens in the conversations.
```json
[
{
"conversations": [
{
"from": "human",
"value": "<image>human instruction"
},
{
"from": "gpt",
"value": "model response"
}
],
"images": [
"image path (required)"
]
}
]
```
Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
```json
"dataset_name": {
"file_name": "data.json",
"formatting": "sharegpt",
"columns": {
"messages": "conversations",
"images": "images"
}
}
```
### Multimodal Video Dataset
- [Example dataset](mllm_video_demo.json)
Multimodal video datasets require a `videos` column containing the paths to the input videos.
The number of videos should be identical to the `<video>` tokens in the conversations.
```json
[
{
"conversations": [
{
"from": "human",
"value": "<video>human instruction"
},
{
"from": "gpt",
"value": "model response"
}
],
"videos": [
"video path (required)"
]
}
]
```
Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
```json
"dataset_name": {
"file_name": "data.json",
"formatting": "sharegpt",
"columns": {
"messages": "conversations",
"videos": "videos"
}
}
```
### OpenAI Format
The openai format is simply a special case of the sharegpt format, where the first message may be a system prompt.
```json
[
{
"messages": [
{
"role": "system",
"content": "system prompt (optional)"
},
{
"role": "user",
"content": "human instruction"
},
{
"role": "assistant",
"content": "model response"
}
]
}
]
```
Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
```json
"dataset_name": {
"file_name": "data.json",
"formatting": "sharegpt",
"columns": {
"messages": "messages"
},
"tags": {
"role_tag": "role",
"content_tag": "content",
"user_tag": "user",
"assistant_tag": "assistant",
"system_tag": "system"
}
}
```
data/README_zh.md 0 → 100644
View file @ 12d5cbac
[dataset_info.json](dataset_info.json) 包含了所有可用的数据集。如果您希望使用自定义数据集，请**务必**在 `dataset_info.json` 文件中添加*数据集描述*，并通过修改 `dataset: 数据集名称` 配置来使用数据集。
目前我们支持 **alpaca** 格式和 **sharegpt** 格式的数据集。
```json
"数据集名称": {
"hf_hub_url": "Hugging Face 的数据集仓库地址（若指定，则忽略 script_url 和 file_name）",
"ms_hub_url": "ModelScope 的数据集仓库地址（若指定，则忽略 script_url 和 file_name）",
"script_url": "包含数据加载脚本的本地文件夹名称（若指定，则忽略 file_name）",
"file_name": "该目录下数据集文件夹或文件的名称（若上述参数未指定，则此项必需）",
"formatting": "数据集格式（可选，默认：alpaca，可以为 alpaca 或 sharegpt）",
"ranking": "是否为偏好数据集（可选，默认：False）",
"subset": "数据集子集的名称（可选，默认：None）",
"split": "所使用的数据集切分（可选，默认：train）",
"folder": "Hugging Face 仓库的文件夹名称（可选，默认：None）",
"num_samples": "该数据集所使用的样本数量。（可选，默认：None）",
"columns（可选）": {
"prompt": "数据集代表提示词的表头名称（默认：instruction）",
"query": "数据集代表请求的表头名称（默认：input）",
"response": "数据集代表回答的表头名称（默认：output）",
"history": "数据集代表历史对话的表头名称（默认：None）",
"messages": "数据集代表消息列表的表头名称（默认：conversations）",
"system": "数据集代表系统提示的表头名称（默认：None）",
"tools": "数据集代表工具描述的表头名称（默认：None）",
"images": "数据集代表图像输入的表头名称（默认：None）",
"videos": "数据集代表视频输入的表头名称（默认：None）",
"chosen": "数据集代表更优回答的表头名称（默认：None）",
"rejected": "数据集代表更差回答的表头名称（默认：None）",
"kto_tag": "数据集代表 KTO 标签的表头名称（默认：None）"
},
"tags（可选，用于 sharegpt 格式）": {
"role_tag": "消息中代表发送者身份的键名（默认：from）",
"content_tag": "消息中代表文本内容的键名（默认：value）",
"user_tag": "消息中代表用户的 role_tag（默认：human）",
"assistant_tag": "消息中代表助手的 role_tag（默认：gpt）",
"observation_tag": "消息中代表工具返回结果的 role_tag（默认：observation）",
"function_tag": "消息中代表工具调用的 role_tag（默认：function_call）",
"system_tag": "消息中代表系统提示的 role_tag（默认：system，会覆盖 system column）"
}
}
```
## Alpaca 格式
### 指令监督微调数据集
- [样例数据集](alpaca_zh_demo.json)
在指令监督微调时，`instruction` 列对应的内容会与 `input` 列对应的内容拼接后作为人类指令，即人类指令为 `instruction\ninput`。而 `output` 列对应的内容为模型回答。
如果指定，`system` 列对应的内容将被作为系统提示词。
`history` 列是由多个字符串二元组构成的列表，分别代表历史消息中每轮对话的指令和回答。注意在指令监督微调时，历史消息中的回答内容**也会被用于模型学习**。
```json
[
{
"instruction": "人类指令（必填）",
"input": "人类输入（选填）",
"output": "模型回答（必填）",
"system": "系统提示词（选填）",
"history": [
["第一轮指令（选填）", "第一轮回答（选填）"],
["第二轮指令（选填）", "第二轮回答（选填）"]
]
}
]
```
对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
```json
"数据集名称": {
"file_name": "data.json",
"columns": {
"prompt": "instruction",
"query": "input",
"response": "output",
"system": "system",
"history": "history"
}
}
```
### 预训练数据集
- [样例数据集](c4_demo.json)
在预训练时，只有 `text` 列中的内容会用于模型学习。
```json
[
{"text": "document"},
{"text": "document"}
]
```
对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
```json
"数据集名称": {
"file_name": "data.json",
"columns": {
"prompt": "text"
}
}
```
### 偏好数据集
偏好数据集用于奖励模型训练、DPO 训练、ORPO 训练和 SimPO 训练。
它需要在 `chosen` 列中提供更优的回答，并在 `rejected` 列中提供更差的回答。
```json
[
{
"instruction": "人类指令（必填）",
"input": "人类输入（选填）",
"chosen": "优质回答（必填）",
"rejected": "劣质回答（必填）"
}
]
```
对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
```json
"数据集名称": {
"file_name": "data.json",
"ranking": true,
"columns": {
"prompt": "instruction",
"query": "input",
"chosen": "chosen",
"rejected": "rejected"
}
}
```
### KTO 数据集
KTO 数据集需要提供额外的 `kto_tag` 列。详情请参阅 [sharegpt](#sharegpt-格式)。
### 多模态图像数据集
多模态图像数据集需要提供额外的 `images` 列。详情请参阅 [sharegpt](#sharegpt-格式)。
### 多模态视频数据集
多模态视频数据集需要提供额外的 `videos` 列。详情请参阅 [sharegpt](#sharegpt-格式)。
## Sharegpt 格式
### 指令监督微调数据集
- [样例数据集](glaive_toolcall_zh_demo.json)
相比 alpaca 格式的数据集，sharegpt 格式支持**更多的角色种类**，例如 human、gpt、observation、function 等等。它们构成一个对象列表呈现在 `conversations` 列中。
注意其中 human 和 observation 必须出现在奇数位置，gpt 和 function 必须出现在偶数位置。
```json
[
{
"conversations": [
{
"from": "human",
"value": "人类指令"
},
{
"from": "function_call",
"value": "工具参数"
},
{
"from": "observation",
"value": "工具结果"
},
{
"from": "gpt",
"value": "模型回答"
}
],
"system": "系统提示词（选填）",
"tools": "工具描述（选填）"
}
]
```
对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
```json
"数据集名称": {
"file_name": "data.json",
"formatting": "sharegpt",
"columns": {
"messages": "conversations",
"system": "system",
"tools": "tools"
}
}
```
### 预训练数据集
尚不支持，请使用 [alpaca](#alpaca-格式) 格式。
### 偏好数据集
- [样例数据集](dpo_zh_demo.json)
Sharegpt 格式的偏好数据集同样需要在 `chosen` 列中提供更优的消息，并在 `rejected` 列中提供更差的消息。
```json
[
{
"conversations": [
{
"from": "human",
"value": "人类指令"
},
{
"from": "gpt",
"value": "模型回答"
},
{
"from": "human",
"value": "人类指令"
}
],
"chosen": {
"from": "gpt",
"value": "优质回答"
},
"rejected": {
"from": "gpt",
"value": "劣质回答"
}
}
]
```
对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
```json
"数据集名称": {
"file_name": "data.json",
"formatting": "sharegpt",
"ranking": true,
"columns": {
"messages": "conversations",
"chosen": "chosen",
"rejected": "rejected"
}
}
```
### KTO 数据集
- [样例数据集](kto_en_demo.json)
KTO 数据集需要额外添加一个 `kto_tag` 列，包含 bool 类型的人类反馈。
```json
[
{
"conversations": [
{
"from": "human",
"value": "人类指令"
},
{
"from": "gpt",
"value": "模型回答"
}
],
"kto_tag": "人类反馈 [true/false]（必填）"
}
]
```
对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
```json
"数据集名称": {
"file_name": "data.json",
"formatting": "sharegpt",
"columns": {
"messages": "conversations",
"kto_tag": "kto_tag"
}
}
```
### 多模态图像数据集
- [样例数据集](mllm_demo.json)
多模态图像数据集需要额外添加一个 `images` 列，包含输入图像的路径。
注意图片的数量必须与文本中所有 `<image>` 标记的数量严格一致。
```json
[
{
"conversations": [
{
"from": "human",
"value": "<image>人类指令"
},
{
"from": "gpt",
"value": "模型回答"
}
],
"images": [
"图像路径（必填）"
]
}
]
```
对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
```json
"数据集名称": {
"file_name": "data.json",
"formatting": "sharegpt",
"columns": {
"messages": "conversations",
"images": "images"
}
}
```
### 多模态视频数据集
- [样例数据集](mllm_video_demo.json)
多模态视频数据集需要额外添加一个 `videos` 列，包含输入视频的路径。
注意视频的数量必须与文本中所有 `<video>` 标记的数量严格一致。
```json
[
{
"conversations": [
{
"from": "human",
"value": "<video>人类指令"
},
{
"from": "gpt",
"value": "模型回答"
}
],
"videos": [
"视频路径（必填）"
]
}
]
```
对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
```json
"数据集名称": {
"file_name": "data.json",
"formatting": "sharegpt",
"columns": {
"messages": "conversations",
"videos": "videos"
}
}
```
### OpenAI 格式
OpenAI 格式仅仅是 sharegpt 格式的一种特殊情况，其中第一条消息可能是系统提示词。
```json
[
{
"messages": [
{
"role": "system",
"content": "系统提示词（选填）"
},
{
"role": "user",
"content": "人类指令"
},
{
"role": "assistant",
"content": "模型回答"
}
]
}
]
```
对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
```json
"数据集名称": {
"file_name": "data.json",
"formatting": "sharegpt",
"columns": {
"messages": "messages"
},
"tags": {
"role_tag": "role",
"content_tag": "content",
"user_tag": "user",
"assistant_tag": "assistant",
"system_tag": "system"
}
}
```
data/alpaca_en_demo.json 0 → 100644
View file @ 12d5cbac
This source diff could not be displayed because it is too large. You can view the blob instead.
data/alpaca_zh_demo.json 0 → 100644
View file @ 12d5cbac
This source diff could not be displayed because it is too large. You can view the blob instead.
data/belle_multiturn/belle_multiturn.py 0 → 100644
View file @ 12d5cbac
import json
import os
import datasets
_HF_ENDPOINT = os.getenv("HF_ENDPOINT", "https://huggingface.co")
_DESCRIPTION = "BELLE multiturn chat dataset."
_CITATION = """\
@article{belle2023exploring,
title={Exploring the Impact of Instruction Data Scaling on Large Language Models: An Empirical Study on Real-World Use Cases},
author={Yunjie Ji, Yong Deng, Yan Gong, Yiping Peng, Qiang Niu, Lei Zhang, Baochang Ma, Xiangang Li},
journal={arXiv preprint arXiv:2303.14742},
year={2023}
}
"""
_HOMEPAGE = "{}/datasets/BelleGroup/multiturn_chat_0.8M".format(_HF_ENDPOINT)
_LICENSE = "gpl-3.0"
_URL = "{}/datasets/BelleGroup/multiturn_chat_0.8M/resolve/main/multiturn_chat_0.8M.json".format(_HF_ENDPOINT)
class BelleMultiturn(datasets.GeneratorBasedBuilder):
VERSION = datasets.Version("0.0.0")
def _info(self):
features = datasets.Features(
{"conversations": [{"from": datasets.Value("string"), "value": datasets.Value("string")}]}
)
return datasets.DatasetInfo(
description=_DESCRIPTION, features=features, homepage=_HOMEPAGE, license=_LICENSE, citation=_CITATION
)
def _split_generators(self, dl_manager: datasets.DownloadManager):
file_path = dl_manager.download(_URL)
return [datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": file_path})]
def _generate_examples(self, filepath: str):
with open(filepath, "r", encoding="utf-8") as f:
for key, row in enumerate(f):
data = json.loads(row)
conversations = []
prompt = data["instruction"].strip()
response = data["output"].strip()
assist_idx = prompt.rfind("Assistant:")
human_idx = prompt.rfind("Human:")
query = prompt[human_idx + 6 : assist_idx].strip()
prompt = prompt[:human_idx].strip()
conversations.insert(0, {"from": "gpt", "value": response})
conversations.insert(0, {"from": "human", "value": query})
while prompt.rfind("Assistant:") != -1:
assist_idx = prompt.rfind("Assistant:")
human_idx = prompt.rfind("Human:")
if human_idx != -1:
old_query = prompt[human_idx + 6 : assist_idx].strip()
old_resp = prompt[assist_idx + 10 :].strip()
conversations.insert(0, {"from": "gpt", "value": old_resp})
conversations.insert(0, {"from": "human", "value": old_query})
else:
break
prompt = prompt[:human_idx].strip()
yield key, {"conversations": conversations}
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment