LLaMA-Factory-Mirror/README_zh.md

480 lines
31 KiB
Markdown
Raw Normal View History

2023-12-02 01:31:24 +08:00
![# LLaMA Factory](assets/logo.png)
2023-10-12 21:42:29 +08:00
[![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/llmtuner)](https://pypi.org/project/llmtuner/)
2023-09-16 17:33:01 +08:00
[![Downloads](https://static.pepy.tech/badge/llmtuner)](https://pypi.org/project/llmtuner/)
2024-04-16 18:09:16 +08:00
[![Citation](https://img.shields.io/badge/citation-34-green)](#使用了-llama-factory-的项目)
2023-10-12 21:42:29 +08:00
[![GitHub pull request](https://img.shields.io/badge/PRs-welcome-blue)](https://github.com/hiyouga/LLaMA-Factory/pulls)
2023-12-15 22:11:31 +08:00
[![Discord](https://dcbadge.vercel.app/api/server/rKfvV9r9FK?compact=true&style=flat)](https://discord.gg/rKfvV9r9FK)
2024-02-29 17:45:30 +08:00
[![Twitter](https://img.shields.io/twitter/follow/llamafactory_ai)](https://twitter.com/llamafactory_ai)
[![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)
2024-03-02 19:58:21 +08:00
[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing)
2023-07-22 14:31:16 +08:00
👋 加入我们的[微信群](assets/wechat.jpg)。
\[ [English](README.md) | 中文 \]
2024-03-03 01:41:07 +08:00
**微调大模型可以像这样轻松…**
2024-03-03 00:49:08 +08:00
https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd-d76c6d0a6594
2024-03-03 01:41:07 +08:00
选择你的打开方式:
- **Colab**https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing
- **本地机器**:请见[如何使用](#如何使用)
2023-10-15 20:28:14 +08:00
2023-11-18 11:09:52 +08:00
## 目录
2024-02-28 20:50:01 +08:00
- [项目特色](#项目特色)
2023-11-18 11:09:52 +08:00
- [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)
2023-11-29 12:05:03 +08:00
- [软硬件依赖](#软硬件依赖)
2023-11-18 11:09:52 +08:00
- [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)
2024-02-28 20:50:01 +08:00
## 项目特色
- **多种模型**LLaMA、Mistral、Mixtral-MoE、Qwen、Yi、Gemma、Baichuan、ChatGLM、Phi 等等。
2024-03-31 18:46:34 +08:00
- **集成方法**增量预训练、指令监督微调、奖励模型训练、PPO 训练、DPO 训练和 ORPO 训练。
2024-02-29 00:34:19 +08:00
- **多种精度**32 比特全参数微调、16 比特冻结微调、16 比特 LoRA 微调和基于 AQLM/AWQ/GPTQ/LLM.int8 的 2/4/8 比特 QLoRA 微调。
2024-04-21 18:11:10 +08:00
- **先进算法**GaLore、BAdam、DoRA、LongLoRA、LLaMA Pro、Mixture-of-Depths、LoRA+、LoftQ 和 Agent 微调。
2024-03-09 03:58:18 +08:00
- **实用技巧**FlashAttention-2、Unsloth、RoPE scaling、NEFTune 和 rsLoRA。
2024-02-28 23:19:25 +08:00
- **实验监控**LlamaBoard、TensorBoard、Wandb、MLflow 等等。
2024-03-07 20:26:31 +08:00
- **极速推理**:基于 vLLM 的 OpenAI 风格 API、浏览器界面和命令行接口。
2024-02-28 20:50:01 +08:00
2023-11-18 11:09:52 +08:00
## 性能指标
2024-04-02 20:07:43 +08:00
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。
2023-11-18 11:09:52 +08:00
![benchmark](assets/benchmark.svg)
2023-12-01 22:53:15 +08:00
<details><summary>变量定义</summary>
2023-11-18 11:15:56 +08:00
- **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
2023-11-18 11:30:01 +08:00
- **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
2023-11-18 11:15:56 +08:00
- **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
2024-04-02 20:07:43 +08:00
- 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`
2023-11-18 11:09:52 +08:00
2023-12-01 22:53:15 +08:00
</details>
## 更新日志
2024-04-21 18:11:10 +08:00
[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/extras/mod`
2024-04-19 02:31:24 +08:00
2024-04-19 01:13:50 +08:00
[24/04/19] 我们支持了 **Meta Llama 3** 系列模型。
2024-04-16 17:44:48 +08:00
[24/04/16] 我们支持了 **[BAdam](https://arxiv.org/abs/2404.02827)**。详细用法请参照 `examples/extras/badam`
2024-04-19 02:31:24 +08:00
[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)。
2024-04-21 18:11:10 +08:00
<details><summary>展开日志</summary>
2024-04-19 01:13:50 +08:00
[24/03/31] 我们支持了 **[ORPO](https://arxiv.org/abs/2403.07691)**。详细用法请参照 `examples/lora_single_gpu`
2024-04-16 17:44:48 +08:00
[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!
2024-04-16 02:35:36 +08:00
[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/extras/fsdp_qlora`
2024-03-28 18:16:27 +08:00
[24/03/13] 我们支持了 **[LoRA+](https://arxiv.org/abs/2402.12354)**。详细用法请参照 `examples/extras/loraplus`
2024-03-13 23:32:51 +08:00
2024-03-28 18:16:27 +08:00
[24/03/07] 我们支持了梯度低秩投影(**[GaLore](https://arxiv.org/abs/2403.03507)**)算法。详细用法请参照 `examples/extras/galore`
2024-03-07 22:41:36 +08:00
2024-03-28 22:02:32 +08:00
[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。(尚不支持 LoRA请先合并权重。
2024-03-21 00:36:06 +08:00
[24/02/28] 我们支持了 **[DoRA](https://arxiv.org/abs/2402.09353)** 微调。请使用 `--use_dora` 参数进行 DoRA 微调。
2024-02-15 02:27:36 +08:00
2024-03-21 00:36:06 +08:00
[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/extras/llama_pro`
2024-02-15 02:27:36 +08:00
2024-03-07 20:26:31 +08:00
[24/02/05] Qwen1.5Qwen2 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/)。
2024-02-28 20:50:01 +08:00
[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` 即可使模型获得工具调用能力。
2024-03-07 20:26:31 +08:00
[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` 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)。
2023-12-23 00:14:33 +08:00
2023-12-12 11:44:30 +08:00
[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)。
2023-12-12 11:39:04 +08:00
2024-01-18 14:42:52 +08:00
[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](#使用魔搭社区可跳过)。
2023-12-12 11:44:30 +08:00
[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` 参数启用 NEFTune例如 `--neftune_noise_alpha 5`
2023-09-28 14:39:16 +08:00
[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` 参数以启用该功能。
2023-09-27 21:55:50 +08:00
2023-09-23 21:10:17 +08:00
[23/09/23] 我们在项目中集成了 MMLU、C-Eval 和 CMMLU 评估集。使用方法请参阅[此示例](#模型评估)。
2023-09-10 20:43:56 +08:00
2023-12-12 22:47:06 +08:00
[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` 参数以启用 FlashAttention-2。
2023-08-12 21:00:11 +08:00
2023-09-23 00:34:17 +08:00
[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。
2023-08-11 03:02:53 +08:00
2023-09-23 00:34:17 +08:00
[23/08/11] 我们支持了指令模型的 **[DPO 训练](https://arxiv.org/abs/2305.18290)**。使用方法请参阅[此示例](#dpo-训练)。
2023-12-12 22:47:06 +08:00
[23/07/31] 我们支持了**数据流式加载**。请使用 `--streaming``--max_steps 10000` 参数来流式加载数据集。
2023-07-31 23:42:32 +08:00
2023-09-09 13:50:29 +08:00
[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))。
2023-08-01 10:08:47 +08:00
2023-10-22 16:15:08 +08:00
[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。
2023-08-12 21:29:06 +08:00
[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。
2023-09-09 13:50:29 +08:00
[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)。
2023-08-12 21:23:05 +08:00
[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。
2023-10-22 16:15:08 +08:00
[23/06/03] 我们实现了 4 比特的 LoRA 训练(也称 **[QLoRA](https://github.com/artidoro/qlora)**)。请使用 `--quantization_bit 4` 参数进行 4 比特量化微调。
2023-12-01 22:53:15 +08:00
</details>
2023-07-22 14:29:22 +08:00
## 模型
2023-08-07 15:02:02 +08:00
2023-09-06 21:43:06 +08:00
| 模型名 | 模型大小 | 默认模块 | Template |
| -------------------------------------------------------- | --------------------------- | ----------------- | --------- |
2023-12-18 22:29:45 +08:00
| [Baichuan2](https://huggingface.co/baichuan-inc) | 7B/13B | W_pack | baichuan2 |
2024-04-15 23:26:42 +08:00
| [BLOOM](https://huggingface.co/bigscience) | 560M/1.1B/1.7B/3B/7.1B/176B | query_key_value | - |
| [BLOOMZ](https://huggingface.co/bigscience) | 560M/1.1B/1.7B/3B/7.1B/176B | query_key_value | - |
| [ChatGLM3](https://huggingface.co/THUDM) | 6B | query_key_value | chatglm3 |
2024-04-16 00:11:15 +08:00
| [Command-R](https://huggingface.co/CohereForAI) | 35B/104B | q_proj,v_proj | cohere |
2024-01-15 23:34:23 +08:00
| [DeepSeek (MoE)](https://huggingface.co/deepseek-ai) | 7B/16B/67B | q_proj,v_proj | deepseek |
2023-12-18 22:29:45 +08:00
| [Falcon](https://huggingface.co/tiiuae) | 7B/40B/180B | query_key_value | falcon |
2024-04-16 00:11:15 +08:00
| [Gemma/CodeGemma](https://huggingface.co/google) | 2B/7B | q_proj,v_proj | gemma |
2024-01-18 14:30:48 +08:00
| [InternLM2](https://huggingface.co/internlm) | 7B/20B | wqkv | intern2 |
| [LLaMA](https://github.com/facebookresearch/llama) | 7B/13B/33B/65B | q_proj,v_proj | - |
| [LLaMA-2](https://huggingface.co/meta-llama) | 7B/13B/70B | q_proj,v_proj | llama2 |
2024-04-19 01:13:50 +08:00
| [LLaMA-3](https://huggingface.co/meta-llama) | 8B/70B | q_proj,v_proj | llama3 |
2024-04-17 23:35:59 +08:00
| [Mistral/Mixtral](https://huggingface.co/mistralai) | 7B/8x7B/8x22B | q_proj,v_proj | mistral |
2024-03-12 18:30:38 +08:00
| [OLMo](https://huggingface.co/allenai) | 1B/7B | att_proj | olmo |
2024-01-13 23:12:47 +08:00
| [Phi-1.5/2](https://huggingface.co/microsoft) | 1.3B/2.7B | q_proj,v_proj | - |
2023-12-18 22:29:45 +08:00
| [Qwen](https://huggingface.co/Qwen) | 1.8B/7B/14B/72B | c_attn | qwen |
2024-04-17 23:35:59 +08:00
| [Qwen1.5 (Code/MoE)](https://huggingface.co/Qwen) | 0.5B/1.8B/4B/7B/14B/32B/72B | q_proj,v_proj | qwen |
| [StarCoder2](https://huggingface.co/bigcode) | 3B/7B/15B | q_proj,v_proj | - |
2023-12-18 22:29:45 +08:00
| [XVERSE](https://huggingface.co/xverse) | 7B/13B/65B | q_proj,v_proj | xverse |
2024-03-07 23:11:57 +08:00
| [Yi](https://huggingface.co/01-ai) | 6B/9B/34B | q_proj,v_proj | yi |
2023-12-29 13:50:24 +08:00
| [Yuan](https://huggingface.co/IEITYuan) | 2B/51B/102B | q_proj,v_proj | yuan |
2023-08-07 15:02:02 +08:00
2023-09-10 21:01:20 +08:00
> [!NOTE]
2023-09-10 20:43:56 +08:00
> **默认模块**应作为 `--lora_target` 参数的默认值,可使用 `--lora_target all` 参数指定全部模块。
>
2023-10-13 13:53:43 +08:00
> 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”Chat模型请务必使用**对应的模板**。
2023-10-27 22:15:25 +08:00
2023-11-15 18:04:37 +08:00
项目所支持模型的完整列表请参阅 [constants.py](src/llmtuner/extras/constants.py)。
2023-08-11 03:02:53 +08:00
2024-03-04 19:29:26 +08:00
您也可以在 [template.py](src/llmtuner/data/template.py) 中添加自己的对话模板。
2023-08-11 03:02:53 +08:00
## 训练方法
2023-08-17 11:00:22 +08:00
| 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: |
2023-11-16 02:08:04 +08:00
| 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: |
2024-03-31 18:29:50 +08:00
| ORPO 训练 | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
2023-07-22 14:29:22 +08:00
## 数据集
2023-11-02 23:10:04 +08:00
<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)
- [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>
- [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)
- [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca)
2024-02-09 14:53:14 +08:00
- [Alpaca GPT4 (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
- [Self Cognition (zh)](data/self_cognition.json)
2023-11-02 23:10:04 +08:00
- [Open Assistant (multilingual)](https://huggingface.co/datasets/OpenAssistant/oasst1)
- [ShareGPT (zh)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT/tree/main/Chinese-instruction-collection)
- [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)
- [LIMA (en)](https://huggingface.co/datasets/GAIR/lima)
- [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)
2023-11-15 18:04:37 +08:00
- [OpenOrca (en)](https://huggingface.co/datasets/Open-Orca/OpenOrca)
2024-02-10 16:39:19 +08:00
- [SlimOrca (en)](https://huggingface.co/datasets/Open-Orca/SlimOrca)
2023-11-02 23:10:04 +08:00
- [MathInstruct (en)](https://huggingface.co/datasets/TIGER-Lab/MathInstruct)
- [Firefly 1.1M (zh)](https://huggingface.co/datasets/YeungNLP/firefly-train-1.1M)
2024-02-09 14:53:14 +08:00
- [Wiki QA (en)](https://huggingface.co/datasets/wiki_qa)
2023-11-02 23:10:04 +08:00
- [Web QA (zh)](https://huggingface.co/datasets/suolyer/webqa)
- [WebNovel (zh)](https://huggingface.co/datasets/zxbsmk/webnovel_cn)
2023-12-01 15:34:50 +08:00
- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
2023-12-25 18:29:34 +08:00
- [deepctrl (en&zh)](https://www.modelscope.cn/datasets/deepctrl/deepctrl-sft-data)
2023-11-02 23:10:04 +08:00
- [Ad Gen (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)
2023-11-02 23:42:49 +08:00
- [LMSYS Chat 1M (en)](https://huggingface.co/datasets/lmsys/lmsys-chat-1m)
2023-11-02 23:10:04 +08:00
- [Evol Instruct V2 (en)](https://huggingface.co/datasets/WizardLM/WizardLM_evol_instruct_V2_196k)
2024-01-18 14:30:48 +08:00
- [Glaive Function Calling V2 (en)](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2)
- [Cosmopedia (en)](https://huggingface.co/datasets/HuggingFaceTB/cosmopedia)
2024-02-09 14:53:14 +08:00
- [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)
2023-11-02 23:10:04 +08:00
</details>
<details><summary>偏好数据集</summary>
- [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
- [Open Assistant (multilingual)](https://huggingface.co/datasets/OpenAssistant/oasst1)
- [GPT-4 Generated Data (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
2024-03-21 00:36:06 +08:00
- [Orca DPO (en)](https://huggingface.co/datasets/Intel/orca_dpo_pairs)
2023-12-01 15:34:50 +08:00
- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
2024-04-21 18:11:10 +08:00
- [DPO mix (en&zh)](https://huggingface.co/datasets/hiyouga/DPO-En-Zh-20k)
2024-02-09 14:53:14 +08:00
- [Orca DPO (de)](https://huggingface.co/datasets/mayflowergmbh/intel_orca_dpo_pairs_de)
2023-11-02 23:10:04 +08:00
</details>
2023-07-22 14:29:22 +08:00
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
```
2023-11-29 12:05:03 +08:00
## 软硬件依赖
2024-02-28 23:19:25 +08:00
| 必需项 | 至少 | 推荐 |
| ------------ | ------- | --------- |
| python | 3.8 | 3.10 |
2024-03-09 00:09:09 +08:00
| torch | 1.13.1 | 2.2.0 |
2024-04-02 20:07:43 +08:00
| transformers | 4.37.2 | 4.39.3 |
2024-04-01 21:35:18 +08:00
| datasets | 2.14.3 | 2.18.0 |
2024-03-24 00:28:44 +08:00
| accelerate | 0.27.2 | 0.28.0 |
| peft | 0.9.0 | 0.10.0 |
| trl | 0.8.1 | 0.8.1 |
2024-02-28 23:19:25 +08:00
| 可选项 | 至少 | 推荐 |
| ------------ | ------- | --------- |
| CUDA | 11.6 | 12.2 |
2024-03-24 00:28:44 +08:00
| deepspeed | 0.10.0 | 0.14.0 |
| bitsandbytes | 0.39.0 | 0.43.0 |
| flash-attn | 2.3.0 | 2.5.6 |
2023-11-29 12:05:03 +08:00
### 硬件依赖
2024-02-28 23:19:25 +08:00
\* *估算值*
2024-04-17 23:40:49 +08:00
| 方法 | 精度 | 7B | 13B | 30B | 70B | 8x7B | 8x22B |
| ----------------- | ---- | ----- | ----- | ----- | ------ | ----- | ------ |
| Full | AMP | 120GB | 240GB | 600GB | 1200GB | 900GB | 2400GB |
| Full | 16 | 60GB | 120GB | 300GB | 600GB | 400GB | 1200GB |
| Freeze | 16 | 20GB | 40GB | 80GB | 200GB | 160GB | 400GB |
| LoRA/GaLore/BAdam | 16 | 16GB | 32GB | 64GB | 160GB | 120GB | 320GB |
| QLoRA | 8 | 10GB | 20GB | 40GB | 80GB | 60GB | 160GB |
| QLoRA | 4 | 6GB | 12GB | 24GB | 48GB | 30GB | 96GB |
| QLoRA | 2 | 4GB | 8GB | 16GB | 24GB | 18GB | 48GB |
2023-07-22 14:29:22 +08:00
## 如何使用
2024-04-02 20:07:43 +08:00
### 数据准备
2024-04-02 20:07:43 +08:00
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 上的数据集或加载本地数据集。
2023-09-10 21:01:20 +08:00
> [!NOTE]
2024-04-02 20:07:43 +08:00
> 使用自定义数据集时,请更新 `data/dataset_info.json` 文件。
2024-04-02 20:07:43 +08:00
### 安装依赖
```bash
2023-10-12 21:42:29 +08:00
git clone https://github.com/hiyouga/LLaMA-Factory.git
conda create -n llama_factory python=3.10
conda activate llama_factory
cd LLaMA-Factory
2024-04-02 20:07:43 +08:00
pip install -e .[metrics]
2023-08-11 03:02:53 +08:00
```
2024-04-18 00:22:17 +08:00
可选的额外依赖项deepspeed、metrics、unsloth、galore、badam、vllm、bitsandbytes、gptq、awq、aqlm、qwen、modelscope、quality
2024-04-02 20:07:43 +08:00
<details><summary>Windows 用户指南</summary>
2024-03-06 14:51:51 +08:00
2024-04-02 20:07:43 +08:00
如果要在 Windows 平台上开启量化 LoRAQLoRA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)。
2023-08-12 21:23:05 +08:00
```bash
2024-04-02 20:07:43 +08:00
pip install https://github.com/jllllll/bitsandbytes-windows-webui/releases/download/wheels/bitsandbytes-0.41.2.post2-py3-none-win_amd64.whl
2023-08-12 21:23:05 +08:00
```
2024-04-02 20:07:43 +08:00
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。
2023-08-12 21:23:05 +08:00
</details>
2024-04-22 00:37:32 +08:00
### 利用 LLaMA Board 可视化界面训练
2023-12-15 23:44:50 +08:00
2024-04-02 20:22:11 +08:00
> [!IMPORTANT]
2024-04-22 00:37:32 +08:00
> LLaMA Board 可视化界面目前仅支持单 GPU 训练,请使用[命令行接口](#命令行接口)来进行多 GPU 分布式训练。
2024-04-02 20:22:11 +08:00
2024-04-02 20:07:43 +08:00
#### 使用本地环境
2023-07-22 14:29:22 +08:00
```bash
2024-04-02 22:17:48 +08:00
export CUDA_VISIBLE_DEVICES=0 # Windows 使用 `set CUDA_VISIBLE_DEVICES=0`
2024-04-16 18:09:16 +08:00
export GRADIO_SERVER_PORT=7860 # Windows 使用 `set GRADIO_SERVER_PORT=7860`
2024-04-02 22:17:48 +08:00
python src/train_web.py # 或 python -m llmtuner.webui.interface
2023-08-18 01:51:55 +08:00
```
2024-04-22 00:21:01 +08:00
<details><summary>阿里云用户指南</summary>
如果您在 LLaMA Board GUI 中遇到显示问题,请尝试在启动前使用以下命令设置环境变量:
```bash
export GRADIO_ROOT_PATH=/${JUPYTER_NAME}/proxy/7860/
```
</details>
2024-03-28 22:02:32 +08:00
#### 使用 Docker
```bash
docker build -f ./Dockerfile -t llama-factory:latest .
docker run --gpus=all \
-v ./hf_cache:/root/.cache/huggingface/ \
-v ./data:/app/data \
-v ./output:/app/output \
-e CUDA_VISIBLE_DEVICES=0 \
-p 7860:7860 \
--shm-size 16G \
--name llama_factory \
-d llama-factory:latest
```
#### 使用 Docker Compose
```bash
docker compose -f ./docker-compose.yml up -d
```
2024-04-02 20:22:11 +08:00
<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。
- 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。
2024-03-28 22:02:32 +08:00
2024-04-02 20:22:11 +08:00
</details>
2024-04-02 20:07:43 +08:00
2024-04-22 00:37:32 +08:00
### 利用命令行接口训练
2024-04-02 20:07:43 +08:00
2024-04-02 20:58:45 +08:00
使用方法请参考 [examples/README_zh.md](examples/README_zh.md)。
2024-04-02 20:07:43 +08:00
2024-04-22 00:37:32 +08:00
您可以执行 `python src/train_bash.py -h` 来查看参数文档。
2024-04-02 20:37:37 +08:00
2024-04-22 00:37:32 +08:00
### 利用 vLLM 部署 OpenAI API
2024-04-02 20:37:37 +08:00
```bash
2024-04-02 22:45:20 +08:00
CUDA_VISIBLE_DEVICES=0,1 API_PORT=8000 python src/api_demo.py \
2024-04-22 00:21:01 +08:00
--model_name_or_path meta-llama/Meta-Llama-3-8B-Instruct \
--template llama3 \
2024-04-02 22:45:20 +08:00
--infer_backend vllm \
--vllm_enforce_eager
2024-04-02 20:37:37 +08:00
```
2024-04-02 20:07:43 +08:00
2024-04-22 00:37:32 +08:00
### 从魔搭社区下载
2024-04-02 20:07:43 +08:00
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`
```
2024-04-22 00:37:32 +08:00
`--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`
2024-04-02 20:07:43 +08:00
2023-10-29 22:07:13 +08:00
## 使用了 LLaMA Factory 的项目
2024-04-16 18:09:16 +08:00
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。
2024-04-02 20:37:37 +08:00
2024-04-02 20:22:11 +08:00
<details><summary>点击显示</summary>
2024-02-25 15:34:47 +08:00
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)
2024-03-24 00:28:44 +08:00
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)
2024-02-25 15:34:47 +08:00
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)
2024-02-25 15:18:58 +08:00
1. Wang et al. Know Your Needs Better: Towards Structured Understanding of Marketer Demands with Analogical Reasoning Augmented LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2401.04319)
2024-02-25 15:34:47 +08:00
1. Wang et al. CANDLE: Iterative Conceptualization and Instantiation Distillation from Large Language Models for Commonsense Reasoning. 2024. [[arxiv]](https://arxiv.org/abs/2401.07286)
2024-02-25 15:18:58 +08:00
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. 2024. [[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)
2024-03-21 17:04:10 +08:00
1. Yu et al. KIEval: A Knowledge-grounded Interactive Evaluation Framework for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.15043)
2024-03-24 00:28:44 +08:00
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)
2024-04-16 18:09:16 +08:00
1. Zhang et al. EDT: Improving Large Language Models' Generation by Entropy-based Dynamic Temperature Sampling. 2024. [[arxiv]](https://arxiv.org/abs/2403.14541)
2024-04-01 21:49:40 +08:00
1. Weller et al. FollowIR: Evaluating and Teaching Information Retrieval Models to Follow Instructions. 2024. [[arxiv]](https://arxiv.org/abs/2403.15246)
2024-03-28 20:24:27 +08:00
1. Hongbin Na. CBT-LLM: A Chinese Large Language Model for Cognitive Behavioral Therapy-based Mental Health Question Answering. 2024. [[arxiv]](https://arxiv.org/abs/2403.16008)
2024-04-16 18:09:16 +08:00
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. Liu et al. Dynamic Generation of Personalities with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.07084)
2024-02-25 15:34:47 +08:00
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/thomas-yanxin/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。
2024-01-13 23:12:47 +08:00
2024-04-02 20:22:11 +08:00
</details>
## 协议
2023-07-22 14:29:22 +08:00
本仓库的代码依照 [Apache-2.0](LICENSE) 协议开源。
2024-04-19 01:13:50 +08:00
使用模型权重时,请遵循对应的模型协议:[Baichuan2](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) / [InternLM2](https://github.com/InternLM/InternLM#license) / [LLaMA](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md) / [LLaMA-2](https://ai.meta.com/llama/license/) / [LLaMA-3](https://llama.meta.com/llama3/license/) / [Mistral](LICENSE) / [OLMo](LICENSE) / [Phi-1.5/2](https://huggingface.co/microsoft/phi-1_5/resolve/main/Research%20License.docx) / [Qwen](https://github.com/QwenLM/Qwen/blob/main/Tongyi%20Qianwen%20LICENSE%20AGREEMENT) / [StarCoder2](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) / [Yuan](https://github.com/IEIT-Yuan/Yuan-2.0/blob/main/LICENSE-Yuan)
## 引用
2023-07-22 14:29:22 +08:00
如果您觉得此项目有帮助,请考虑以下列格式引用
```bibtex
2024-03-21 13:49:17 +08:00
@article{zheng2024llamafactory,
title={LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models},
2024-03-21 17:04:10 +08:00
author={Yaowei Zheng and Richong Zhang and Junhao Zhang and Yanhan Ye and Zheyan Luo and Yongqiang Ma},
2024-03-21 13:49:17 +08:00
journal={arXiv preprint arXiv:2403.13372},
year={2024},
url={http://arxiv.org/abs/2403.13372}
}
```
## 致谢
2024-04-02 20:07:43 +08:00
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
2023-10-12 21:42:29 +08:00
![Star History Chart](https://api.star-history.com/svg?repos=hiyouga/LLaMA-Factory&type=Date)