youssefedweqd/working
06
12 3[](https://github.com/hiyouga/LLaMA-Factory/stargazers)4[](https://github.com/hiyouga/LLaMA-Factory/commits/main)5[](https://github.com/hiyouga/LLaMA-Factory/graphs/contributors)6[](https://github.com/hiyouga/LLaMA-Factory/actions/workflows/tests.yml)7[](https://pypi.org/project/llamafactory/)8[](https://scholar.google.com/scholar?cites=12620864006390196564)9[](https://hub.docker.com/r/hiyouga/llamafactory/tags)10 11[](https://twitter.com/llamafactory_ai)12[](https://discord.gg/rKfvV9r9FK)13[](https://gitcode.com/zhengyaowei/LLaMA-Factory)14 15[](https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing)16[](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory)17[](https://docs.alayanew.com/docs/documents/newActivities/llamafactory/?utm_source=LLaMA-Factory)18[](https://huggingface.co/spaces/hiyouga/LLaMA-Board)19[](https://modelscope.cn/studios/hiyouga/LLaMA-Board)20[](https://novita.ai/templates-library/105981?sharer=88115474-394e-4bda-968e-b88e123d0c47)21 22### Used by [Amazon](https://aws.amazon.com/cn/blogs/machine-learning/how-apoidea-group-enhances-visual-information-extraction-from-banking-documents-with-multimodal-models-using-llama-factory-on-amazon-sagemaker-hyperpod/), [NVIDIA](https://developer.nvidia.com/rtx/ai-toolkit), [Aliyun](https://help.aliyun.com/zh/pai/use-cases/fine-tune-a-llama-3-model-with-llama-factory), etc.23 24<div align="center" markdown="1">25 26### Supporters β€οΈ27 28<a href="https://warp.dev/llama-factory">29 <img alt="Warp sponsorship" width="400" src="https://github.com/user-attachments/assets/ab8dd143-b0fd-4904-bdc5-dd7ecac94eae">30</a>31 32#### [Warp, the agentic terminal for developers](https://warp.dev/llama-factory)33 34[Available for MacOS, Linux, & Windows](https://warp.dev/llama-factory)35 36----37 38### Easily fine-tune 100+ large language models with zero-code [CLI](#quickstart) and [Web UI](#fine-tuning-with-llama-board-gui-powered-by-gradio)39 4041 42</div>43 44π Join our [WeChat group](assets/wechat.jpg), [NPU user group](assets/wechat_npu.jpg) or [Alaya NeW user group](assets/wechat_alaya.png).45 46\[ English | [δΈζ](README_zh.md) \]47 48**Fine-tuning a large language model can be easy as...**49 50https://github.com/user-attachments/assets/3991a3a8-4276-4d30-9cab-4cb0c4b9b99e51 52Choose your path:53 54- **Documentation**: https://llamafactory.readthedocs.io/en/latest/55- **Colab (free)**: https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing56- **Local machine**: Please refer to [usage](#getting-started)57- **PAI-DSW (free trial)**: https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory58- **Alaya NeW (cloud GPU deal)**: https://docs.alayanew.com/docs/documents/useGuide/LLaMAFactory/mutiple/?utm_source=LLaMA-Factory59 60> [!NOTE]61> Except for the above links, all other websites are unauthorized third-party websites. Please carefully use them.62 63## Table of Contents64 65- [Features](#features)66- [Blogs](#blogs)67- [Changelog](#changelog)68- [Supported Models](#supported-models)69- [Supported Training Approaches](#supported-training-approaches)70- [Provided Datasets](#provided-datasets)71- [Requirement](#requirement)72- [Getting Started](#getting-started)73 - [Installation](#installation)74 - [Data Preparation](#data-preparation)75 - [Quickstart](#quickstart)76 - [Fine-Tuning with LLaMA Board GUI](#fine-tuning-with-llama-board-gui-powered-by-gradio)77 - [Build Docker](#build-docker)78 - [Deploy with OpenAI-style API and vLLM](#deploy-with-openai-style-api-and-vllm)79 - [Download from ModelScope Hub](#download-from-modelscope-hub)80 - [Download from Modelers Hub](#download-from-modelers-hub)81 - [Use W&B Logger](#use-wb-logger)82 - [Use SwanLab Logger](#use-swanlab-logger)83- [Projects using LLaMA Factory](#projects-using-llama-factory)84- [License](#license)85- [Citation](#citation)86- [Acknowledgement](#acknowledgement)87 88## Features89 90- **Various models**: LLaMA, LLaVA, Mistral, Mixtral-MoE, Qwen, Qwen2-VL, DeepSeek, Yi, Gemma, ChatGLM, Phi, etc.91- **Integrated methods**: (Continuous) pre-training, (multimodal) supervised fine-tuning, reward modeling, PPO, DPO, KTO, ORPO, etc.92- **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.93- **Advanced algorithms**: [GaLore](https://github.com/jiaweizzhao/GaLore), [BAdam](https://github.com/Ledzy/BAdam), [APOLLO](https://github.com/zhuhanqing/APOLLO), [Adam-mini](https://github.com/zyushun/Adam-mini), [Muon](https://github.com/KellerJordan/Muon), DoRA, LongLoRA, LLaMA Pro, Mixture-of-Depths, LoRA+, LoftQ and PiSSA.94- **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.95- **Wide tasks**: Multi-turn dialogue, tool using, image understanding, visual grounding, video recognition, audio understanding, etc.96- **Experiment monitors**: LlamaBoard, TensorBoard, Wandb, MLflow, [SwanLab](https://github.com/SwanHubX/SwanLab), etc.97- **Faster inference**: OpenAI-style API, Gradio UI and CLI with [vLLM worker](https://github.com/vllm-project/vllm) or [SGLang worker](https://github.com/sgl-project/sglang).98 99### Day-N Support for Fine-Tuning Cutting-Edge Models100 101| Support Date | Model Name |102| ------------ | ------------------------------------------------------------ |103| Day 0 | Qwen3 / Qwen2.5-VL / Gemma 3 / InternLM 3 / MiniCPM-o-2.6 |104| Day 1 | Llama 3 / GLM-4 / Mistral Small / PaliGemma2 / Llama 4 |105 106## Blogs107 108- [A One-Stop Code-Free Model Reinforcement Learning and Deployment Platform based on LLaMA-Factory and EasyR1](https://aws.amazon.com/cn/blogs/china/building-llm-model-hub-based-on-llamafactory-and-easyr1/) (Chinese)109- [Fine-tune Qwen2.5-VL for Autonomous Driving using LLaMA-Factory](https://docs.alayanew.com/docs/documents/useGuide/LLaMAFactory/mutiple/?utm_source=LLaMA-Factory) (Chinese)110- [How Apoidea Group enhances visual information extraction from banking documents with multimodal models using LLaMA-Factory on Amazon SageMaker HyperPod](https://aws.amazon.com/cn/blogs/machine-learning/how-apoidea-group-enhances-visual-information-extraction-from-banking-documents-with-multimodal-models-using-llama-factory-on-amazon-sagemaker-hyperpod/) (English)111- [Easy Dataset Γ LLaMA Factory: Enabling LLMs to Efficiently Learn Domain Knowledge](https://buaa-act.feishu.cn/wiki/GVzlwYcRFiR8OLkHbL6cQpYin7g) (English)112 113<details><summary>All Blogs</summary>114 115- [LLaMA Factory: Fine-tuning the DeepSeek-R1-Distill-Qwen-7B Model for News Classifier](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory_deepseek_r1_distill_7b) (Chinese)116- [A One-Stop Code-Free Model Fine-Tuning \& Deployment Platform based on SageMaker and LLaMA-Factory](https://aws.amazon.com/cn/blogs/china/a-one-stop-code-free-model-fine-tuning-deployment-platform-based-on-sagemaker-and-llama-factory/) (Chinese)117- [LLaMA Factory Multi-Modal Fine-Tuning Practice: Fine-Tuning Qwen2-VL for Personal Tourist Guide](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory_qwen2vl) (Chinese)118- [LLaMA Factory: Fine-tuning the LLaMA3 Model for Role-Playing](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory) (Chinese)119 120</details>121 122## Changelog123 124[25/04/28] We supported fine-tuning the **[Qwen3](https://qwenlm.github.io/blog/qwen3/)** model family.125 126[25/04/21] We supported the **[Muon](https://github.com/KellerJordan/Muon)** optimizer. See [examples](examples/README.md) for usage. Thank [@tianshijing](https://github.com/tianshijing)'s PR.127 128[25/04/16] We supported fine-tuning the **[InternVL3](https://huggingface.co/OpenGVLab/InternVL3-8B)** model. See [PR #7258](https://github.com/hiyouga/LLaMA-Factory/pull/7258) to get started.129 130[25/04/14] We supported fine-tuning the **[GLM-Z1](https://huggingface.co/THUDM/GLM-Z1-9B-0414)** and **[Kimi-VL](https://huggingface.co/moonshotai/Kimi-VL-A3B-Instruct)** models.131 132[25/04/06] We supported fine-tuning the **[Llama 4](https://ai.meta.com/blog/llama-4-multimodal-intelligence/)** model. See [PR #7611](https://github.com/hiyouga/LLaMA-Factory/pull/7611) to get started.133 134<details><summary>Full Changelog</summary>135 136[25/03/31] We supported fine-tuning the **[Qwen2.5 Omni](https://qwenlm.github.io/blog/qwen2.5-omni/)** model. See [PR #7537](https://github.com/hiyouga/LLaMA-Factory/pull/7537) to get started.137 138[25/03/15] We supported **[SGLang](https://github.com/sgl-project/sglang)** as inference backend. Try `infer_backend: sglang` to accelerate inference.139 140[25/03/12] We supported fine-tuning the **[Gemma 3](https://huggingface.co/blog/gemma3)** model.141 142[25/02/24] Announcing **[EasyR1](https://github.com/hiyouga/EasyR1)**, an efficient, scalable and multi-modality RL training framework for efficient GRPO training.143 144[25/02/11] We supported saving the **[Ollama](https://github.com/ollama/ollama)** modelfile when exporting the model checkpoints. See [examples](examples/README.md) for usage.145 146[25/02/05] We supported fine-tuning the **[Qwen2-Audio](Qwen/Qwen2-Audio-7B-Instruct)** and **[MiniCPM-o-2.6](https://huggingface.co/openbmb/MiniCPM-o-2_6)** on audio understanding tasks.147 148[25/01/31] We supported fine-tuning the **[DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1)** and **[Qwen2.5-VL](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct)** models.149 150[25/01/15] We supported **[APOLLO](https://arxiv.org/abs/2412.05270)** optimizer. See [examples](examples/README.md) for usage.151 152[25/01/14] We supported fine-tuning the **[MiniCPM-o-2.6](https://huggingface.co/openbmb/MiniCPM-o-2_6)** and **[MiniCPM-V-2.6](https://huggingface.co/openbmb/MiniCPM-V-2_6)** models. Thank [@BUAADreamer](https://github.com/BUAADreamer)'s PR.153 154[25/01/14] We supported fine-tuning the **[InternLM 3](https://huggingface.co/collections/internlm/)** models. Thank [@hhaAndroid](https://github.com/hhaAndroid)'s PR.155 156[25/01/10] We supported fine-tuning the **[Phi-4](https://huggingface.co/microsoft/phi-4)** model.157 158[24/12/21] We supported using **[SwanLab](https://github.com/SwanHubX/SwanLab)** for experiment tracking and visualization. See [this section](#use-swanlab-logger) for details.159 160[24/11/27] We supported fine-tuning the **[Skywork-o1](https://huggingface.co/Skywork/Skywork-o1-Open-Llama-3.1-8B)** model and the **[OpenO1](https://huggingface.co/datasets/O1-OPEN/OpenO1-SFT)** dataset.161 162[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.163 164[24/09/19] We supported fine-tuning the **[Qwen2.5](https://qwenlm.github.io/blog/qwen2.5/)** models.165 166[24/08/30] We supported fine-tuning the **[Qwen2-VL](https://qwenlm.github.io/blog/qwen2-vl/)** models. Thank [@simonJJJ](https://github.com/simonJJJ)'s PR.167 168[24/08/27] We supported **[Liger Kernel](https://github.com/linkedin/Liger-Kernel)**. Try `enable_liger_kernel: true` for efficient training.169 170[24/08/09] We supported **[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.171 172[24/07/04] We supported [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.173 174[24/06/16] We supported **[PiSSA](https://arxiv.org/abs/2404.02948)** algorithm. See [examples](examples/README.md) for usage.175 176[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.177 178[24/05/26] We supported **[SimPO](https://arxiv.org/abs/2405.14734)** algorithm for preference learning. See [examples](examples/README.md) for usage.179 180[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.181 182[24/05/18] We supported **[KTO](https://arxiv.org/abs/2402.01306)** algorithm for preference learning. See [examples](examples/README.md) for usage.183 184[24/05/14] We supported training and inference on the Ascend NPU devices. Check [installation](#installation) section for details.185 186[24/04/26] We supported fine-tuning the **LLaVA-1.5** multimodal LLMs. See [examples](examples/README.md) for usage.187 188[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.189 190[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.191 192[24/04/16] We supported **[BAdam](https://arxiv.org/abs/2404.02827)** optimizer. See [examples](examples/README.md) for usage.193 194[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).195 196[24/03/31] We supported **[ORPO](https://arxiv.org/abs/2403.07691)**. See [examples](examples/README.md) for usage.197 198[24/03/21] Our paper "[LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models](https://arxiv.org/abs/2403.13372)" is available at arXiv!199 200[24/03/20] We supported **FSDP+QLoRA** that fine-tunes a 70B model on 2x24GB GPUs. See [examples](examples/README.md) for usage.201 202[24/03/13] We supported **[LoRA+](https://arxiv.org/abs/2402.12354)**. See [examples](examples/README.md) for usage.203 204[24/03/07] We supported **[GaLore](https://arxiv.org/abs/2403.03507)** optimizer. See [examples](examples/README.md) for usage.205 206[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.207 208[24/02/28] We supported weight-decomposed LoRA (**[DoRA](https://arxiv.org/abs/2402.09353)**). Try `use_dora: true` to activate DoRA training.209 210[24/02/15] We supported **block expansion** proposed by [LLaMA Pro](https://github.com/TencentARC/LLaMA-Pro). See [examples](examples/README.md) for usage.211 212[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.213 214[24/01/18] We supported **agent tuning** for most models, equipping model with tool using abilities by fine-tuning with `dataset: glaive_toolcall_en`.215 216[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.217 218[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).219 220[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.221 222[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.223 224[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.225 226[23/09/23] We integrated MMLU, C-Eval and CMMLU benchmarks in this repo. See [examples](examples/README.md) for usage.227 228[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.229 230[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.231 232[23/08/11] We supported **[DPO training](https://arxiv.org/abs/2305.18290)** for instruction-tuned models. See [examples](examples/README.md) for usage.233 234[23/07/31] We supported **dataset streaming**. Try `streaming: true` and `max_steps: 10000` arguments to load your dataset in streaming mode.235 236[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.237 238[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.239 240[23/07/09] We released **[FastEdit](https://github.com/hiyouga/FastEdit)** β‘π©Ή, 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.241 242[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.243 244[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**.245 246[23/06/03] We supported quantized training and inference (aka **[QLoRA](https://github.com/artidoro/qlora)**). See [examples](examples/README.md) for usage.247 248</details>249 250> [!TIP]251> If you cannot use the latest feature, please pull the latest code and install LLaMA-Factory again.252 253## Supported Models254 255| Model | Model size | Template |256| ----------------------------------------------------------------- | -------------------------------- | ------------------- |257| [Baichuan 2](https://huggingface.co/baichuan-inc) | 7B/13B | baichuan2 |258| [BLOOM/BLOOMZ](https://huggingface.co/bigscience) | 560M/1.1B/1.7B/3B/7.1B/176B | - |259| [ChatGLM3](https://huggingface.co/THUDM) | 6B | chatglm3 |260| [Command R](https://huggingface.co/CohereForAI) | 35B/104B | cohere |261| [DeepSeek (Code/MoE)](https://huggingface.co/deepseek-ai) | 7B/16B/67B/236B | deepseek |262| [DeepSeek 2.5/3](https://huggingface.co/deepseek-ai) | 236B/671B | deepseek3 |263| [DeepSeek R1 (Distill)](https://huggingface.co/deepseek-ai) | 1.5B/7B/8B/14B/32B/70B/671B | deepseekr1 |264| [Falcon](https://huggingface.co/tiiuae) | 7B/11B/40B/180B | falcon |265| [Gemma/Gemma 2/CodeGemma](https://huggingface.co/google) | 2B/7B/9B/27B | gemma |266| [Gemma 3](https://huggingface.co/google) | 1B/4B/12B/27B | gemma3/gemma (1B) |267| [GLM-4/GLM-4-0414/GLM-Z1](https://huggingface.co/THUDM) | 9B/32B | glm4/glmz1 |268| [GPT-2](https://huggingface.co/openai-community) | 0.1B/0.4B/0.8B/1.5B | - |269| [Granite 3.0-3.3](https://huggingface.co/ibm-granite) | 1B/2B/3B/8B | granite3 |270| [Hunyuan](https://huggingface.co/tencent/) | 7B | hunyuan |271| [Index](https://huggingface.co/IndexTeam) | 1.9B | index |272| [InternLM 2-3](https://huggingface.co/internlm) | 7B/8B/20B | intern2 |273| [InternVL 2.5-3](https://huggingface.co/OpenGVLab) | 1B/2B/8B/14B/38B/78B | intern_vl |274| [Kimi-VL](https://huggingface.co/moonshotai) | 16B | kimi_vl |275| [Llama](https://github.com/facebookresearch/llama) | 7B/13B/33B/65B | - |276| [Llama 2](https://huggingface.co/meta-llama) | 7B/13B/70B | llama2 |277| [Llama 3-3.3](https://huggingface.co/meta-llama) | 1B/3B/8B/70B | llama3 |278| [Llama 4](https://huggingface.co/meta-llama) | 109B/402B | llama4 |279| [Llama 3.2 Vision](https://huggingface.co/meta-llama) | 11B/90B | mllama |280| [LLaVA-1.5](https://huggingface.co/llava-hf) | 7B/13B | llava |281| [LLaVA-NeXT](https://huggingface.co/llava-hf) | 7B/8B/13B/34B/72B/110B | llava_next |282| [LLaVA-NeXT-Video](https://huggingface.co/llava-hf) | 7B/34B | llava_next_video |283| [MiMo](https://huggingface.co/XiaomiMiMo) | 7B | mimo |284| [MiniCPM](https://huggingface.co/openbmb) | 0.5B/1B/2B/4B/8B | cpm/cpm3/cpm4 |285| [MiniCPM-o-2.6/MiniCPM-V-2.6](https://huggingface.co/openbmb) | 8B | minicpm_o/minicpm_v |286| [Ministral/Mistral-Nemo](https://huggingface.co/mistralai) | 8B/12B | ministral |287| [Mistral/Mixtral](https://huggingface.co/mistralai) | 7B/8x7B/8x22B | mistral |288| [Mistral Small](https://huggingface.co/mistralai) | 24B | mistral_small |289| [OLMo](https://huggingface.co/allenai) | 1B/7B | - |290| [PaliGemma/PaliGemma2](https://huggingface.co/google) | 3B/10B/28B | paligemma |291| [Phi-1.5/Phi-2](https://huggingface.co/microsoft) | 1.3B/2.7B | - |292| [Phi-3/Phi-3.5](https://huggingface.co/microsoft) | 4B/14B | phi |293| [Phi-3-small](https://huggingface.co/microsoft) | 7B | phi_small |294| [Phi-4](https://huggingface.co/microsoft) | 14B | phi4 |295| [Pixtral](https://huggingface.co/mistralai) | 12B | pixtral |296| [Qwen (1-2.5) (Code/Math/MoE/QwQ)](https://huggingface.co/Qwen) | 0.5B/1.5B/3B/7B/14B/32B/72B/110B | qwen |297| [Qwen3 (MoE)](https://huggingface.co/Qwen) | 0.6B/1.7B/4B/8B/14B/32B/235B | qwen3 |298| [Qwen2-Audio](https://huggingface.co/Qwen) | 7B | qwen2_audio |299| [Qwen2.5-Omni](https://huggingface.co/Qwen) | 3B/7B | qwen2_omni |300| [Qwen2-VL/Qwen2.5-VL/QVQ](https://huggingface.co/Qwen) | 2B/3B/7B/32B/72B | qwen2_vl |301| [Seed Coder](https://huggingface.co/ByteDance-Seed) | 8B | seed_coder |302| [Skywork o1](https://huggingface.co/Skywork) | 8B | skywork_o1 |303| [StarCoder 2](https://huggingface.co/bigcode) | 3B/7B/15B | - |304| [TeleChat2](https://huggingface.co/Tele-AI) | 3B/7B/35B/115B | telechat2 |305| [XVERSE](https://huggingface.co/xverse) | 7B/13B/65B | xverse |306| [Yi/Yi-1.5 (Code)](https://huggingface.co/01-ai) | 1.5B/6B/9B/34B | yi |307| [Yi-VL](https://huggingface.co/01-ai) | 6B/34B | yi_vl |308| [Yuan 2](https://huggingface.co/IEITYuan) | 2B/51B/102B | yuan |309 310> [!NOTE]311> 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.312>313> Remember to use the **SAME** template in training and inference.314>315> \*: You should install the `transformers` from main branch and use `DISABLE_VERSION_CHECK=1` to skip version check.316>317> \*\*: You need to install a specific version of `transformers` to use the corresponding model.318 319Please refer to [constants.py](src/llamafactory/extras/constants.py) for a full list of models we supported.320 321You also can add a custom chat template to [template.py](src/llamafactory/data/template.py).322 323## Supported Training Approaches324 325| Approach | Full-tuning | Freeze-tuning | LoRA | QLoRA |326| ---------------------- | ------------------ | ------------------ | ------------------ | ------------------ |327| Pre-Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |328| Supervised Fine-Tuning | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |329| Reward Modeling | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |330| PPO Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |331| DPO Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |332| KTO Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |333| ORPO Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |334| SimPO Training | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |335 336> [!TIP]337> The implementation details of PPO can be found in [this blog](https://newfacade.github.io/notes-on-reinforcement-learning/17-ppo-trl.html).338 339## Provided Datasets340 341<details><summary>Pre-training datasets</summary>342 343- [Wiki Demo (en)](data/wiki_demo.txt)344- [RefinedWeb (en)](https://huggingface.co/datasets/tiiuae/falcon-refinedweb)345- [RedPajama V2 (en)](https://huggingface.co/datasets/togethercomputer/RedPajama-Data-V2)346- [Wikipedia (en)](https://huggingface.co/datasets/olm/olm-wikipedia-20221220)347- [Wikipedia (zh)](https://huggingface.co/datasets/pleisto/wikipedia-cn-20230720-filtered)348- [Pile (en)](https://huggingface.co/datasets/EleutherAI/pile)349- [SkyPile (zh)](https://huggingface.co/datasets/Skywork/SkyPile-150B)350- [FineWeb (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb)351- [FineWeb-Edu (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu)352- [The Stack (en)](https://huggingface.co/datasets/bigcode/the-stack)353- [StarCoder (en)](https://huggingface.co/datasets/bigcode/starcoderdata)354 355</details>356 357<details><summary>Supervised fine-tuning datasets</summary>358 359- [Identity (en&zh)](data/identity.json)360- [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)361- [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca-3)362- [Alpaca GPT4 (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)363- [Glaive Function Calling V2 (en&zh)](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2)364- [LIMA (en)](https://huggingface.co/datasets/GAIR/lima)365- [Guanaco Dataset (multilingual)](https://huggingface.co/datasets/JosephusCheung/GuanacoDataset)366- [BELLE 2M (zh)](https://huggingface.co/datasets/BelleGroup/train_2M_CN)367- [BELLE 1M (zh)](https://huggingface.co/datasets/BelleGroup/train_1M_CN)368- [BELLE 0.5M (zh)](https://huggingface.co/datasets/BelleGroup/train_0.5M_CN)369- [BELLE Dialogue 0.4M (zh)](https://huggingface.co/datasets/BelleGroup/generated_chat_0.4M)370- [BELLE School Math 0.25M (zh)](https://huggingface.co/datasets/BelleGroup/school_math_0.25M)371- [BELLE Multiturn Chat 0.8M (zh)](https://huggingface.co/datasets/BelleGroup/multiturn_chat_0.8M)372- [UltraChat (en)](https://github.com/thunlp/UltraChat)373- [OpenPlatypus (en)](https://huggingface.co/datasets/garage-bAInd/Open-Platypus)374- [CodeAlpaca 20k (en)](https://huggingface.co/datasets/sahil2801/CodeAlpaca-20k)375- [Alpaca CoT (multilingual)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT)376- [OpenOrca (en)](https://huggingface.co/datasets/Open-Orca/OpenOrca)377- [SlimOrca (en)](https://huggingface.co/datasets/Open-Orca/SlimOrca)378- [MathInstruct (en)](https://huggingface.co/datasets/TIGER-Lab/MathInstruct)379- [Firefly 1.1M (zh)](https://huggingface.co/datasets/YeungNLP/firefly-train-1.1M)380- [Wiki QA (en)](https://huggingface.co/datasets/wiki_qa)381- [Web QA (zh)](https://huggingface.co/datasets/suolyer/webqa)382- [WebNovel (zh)](https://huggingface.co/datasets/zxbsmk/webnovel_cn)383- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)384- [deepctrl (en&zh)](https://www.modelscope.cn/datasets/deepctrl/deepctrl-sft-data)385- [Advertise Generating (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)386- [ShareGPT Hyperfiltered (en)](https://huggingface.co/datasets/totally-not-an-llm/sharegpt-hyperfiltered-3k)387- [ShareGPT4 (en&zh)](https://huggingface.co/datasets/shibing624/sharegpt_gpt4)388- [UltraChat 200k (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k)389- [AgentInstruct (en)](https://huggingface.co/datasets/THUDM/AgentInstruct)390- [LMSYS Chat 1M (en)](https://huggingface.co/datasets/lmsys/lmsys-chat-1m)391- [Evol Instruct V2 (en)](https://huggingface.co/datasets/WizardLM/WizardLM_evol_instruct_V2_196k)392- [Cosmopedia (en)](https://huggingface.co/datasets/HuggingFaceTB/cosmopedia)393- [STEM (zh)](https://huggingface.co/datasets/hfl/stem_zh_instruction)394- [Ruozhiba (zh)](https://huggingface.co/datasets/hfl/ruozhiba_gpt4_turbo)395- [Neo-sft (zh)](https://huggingface.co/datasets/m-a-p/neo_sft_phase2)396- [Magpie-Pro-300K-Filtered (en)](https://huggingface.co/datasets/Magpie-Align/Magpie-Pro-300K-Filtered)397- [Magpie-ultra-v0.1 (en)](https://huggingface.co/datasets/argilla/magpie-ultra-v0.1)398- [WebInstructSub (en)](https://huggingface.co/datasets/TIGER-Lab/WebInstructSub)399- [OpenO1-SFT (en&zh)](https://huggingface.co/datasets/O1-OPEN/OpenO1-SFT)400- [Open-Thoughts (en)](https://huggingface.co/datasets/open-thoughts/OpenThoughts-114k)401- [Open-R1-Math (en)](https://huggingface.co/datasets/open-r1/OpenR1-Math-220k)402- [Chinese-DeepSeek-R1-Distill (zh)](https://huggingface.co/datasets/Congliu/Chinese-DeepSeek-R1-Distill-data-110k-SFT)403- [LLaVA mixed (en&zh)](https://huggingface.co/datasets/BUAADreamer/llava-en-zh-300k)404- [Pokemon-gpt4o-captions (en&zh)](https://huggingface.co/datasets/jugg1024/pokemon-gpt4o-captions)405- [Open Assistant (de)](https://huggingface.co/datasets/mayflowergmbh/oasst_de)406- [Dolly 15k (de)](https://huggingface.co/datasets/mayflowergmbh/dolly-15k_de)407- [Alpaca GPT4 (de)](https://huggingface.co/datasets/mayflowergmbh/alpaca-gpt4_de)408- [OpenSchnabeltier (de)](https://huggingface.co/datasets/mayflowergmbh/openschnabeltier_de)409- [Evol Instruct (de)](https://huggingface.co/datasets/mayflowergmbh/evol-instruct_de)410- [Dolphin (de)](https://huggingface.co/datasets/mayflowergmbh/dolphin_de)411- [Booksum (de)](https://huggingface.co/datasets/mayflowergmbh/booksum_de)412- [Airoboros (de)](https://huggingface.co/datasets/mayflowergmbh/airoboros-3.0_de)413- [Ultrachat (de)](https://huggingface.co/datasets/mayflowergmbh/ultra-chat_de)414 415</details>416 417<details><summary>Preference datasets</summary>418 419- [DPO mixed (en&zh)](https://huggingface.co/datasets/hiyouga/DPO-En-Zh-20k)420- [UltraFeedback (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized)421- [COIG-P (zh)](https://huggingface.co/datasets/m-a-p/COIG-P)422- [RLHF-V (en)](https://huggingface.co/datasets/openbmb/RLHF-V-Dataset)423- [VLFeedback (en)](https://huggingface.co/datasets/Zhihui/VLFeedback)424- [RLAIF-V (en)](https://huggingface.co/datasets/openbmb/RLAIF-V-Dataset)425- [Orca DPO Pairs (en)](https://huggingface.co/datasets/Intel/orca_dpo_pairs)426- [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)427- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)428- [Orca DPO (de)](https://huggingface.co/datasets/mayflowergmbh/intel_orca_dpo_pairs_de)429- [KTO mixed (en)](https://huggingface.co/datasets/argilla/kto-mix-15k)430 431</details>432 433Some datasets require confirmation before using them, so we recommend logging in with your Hugging Face account using these commands.434 435```bash436pip install --upgrade huggingface_hub437huggingface-cli login438```439 440## Requirement441 442| Mandatory | Minimum | Recommend |443| ------------ | ------- | --------- |444| python | 3.9 | 3.10 |445| torch | 2.0.0 | 2.6.0 |446| torchvision | 0.15.0 | 0.21.0 |447| transformers | 4.45.0 | 4.50.0 |448| datasets | 2.16.0 | 3.2.0 |449| accelerate | 0.34.0 | 1.2.1 |450| peft | 0.14.0 | 0.15.1 |451| trl | 0.8.6 | 0.9.6 |452 453| Optional | Minimum | Recommend |454| ------------ | ------- | --------- |455| CUDA | 11.6 | 12.2 |456| deepspeed | 0.10.0 | 0.16.4 |457| bitsandbytes | 0.39.0 | 0.43.1 |458| vllm | 0.4.3 | 0.8.2 |459| flash-attn | 2.5.6 | 2.7.2 |460 461### Hardware Requirement462 463\* *estimated*464 465| Method | Bits | 7B | 14B | 30B | 70B | `x`B |466| ------------------------------- | ---- | ----- | ----- | ----- | ------ | ------- |467| Full (`bf16` or `fp16`) | 32 | 120GB | 240GB | 600GB | 1200GB | `18x`GB |468| Full (`pure_bf16`) | 16 | 60GB | 120GB | 300GB | 600GB | `8x`GB |469| Freeze/LoRA/GaLore/APOLLO/BAdam | 16 | 16GB | 32GB | 64GB | 160GB | `2x`GB |470| QLoRA | 8 | 10GB | 20GB | 40GB | 80GB | `x`GB |471| QLoRA | 4 | 6GB | 12GB | 24GB | 48GB | `x/2`GB |472| QLoRA | 2 | 4GB | 8GB | 16GB | 24GB | `x/4`GB |473 474## Getting Started475 476### Installation477 478> [!IMPORTANT]479> Installation is mandatory.480 481#### Install from Source482 483```bash484git clone --depth 1 https://github.com/hiyouga/LLaMA-Factory.git485cd LLaMA-Factory486pip install -e ".[torch,metrics]" --no-build-isolation487```488 489Extra dependencies available: torch, torch-npu, metrics, deepspeed, liger-kernel, bitsandbytes, hqq, eetq, gptq, aqlm, vllm, sglang, galore, apollo, badam, adam-mini, qwen, minicpm_v, modelscope, openmind, swanlab, dev490 491#### Install from Docker Image492 493```bash494docker run -it --rm --gpus=all --ipc=host hiyouga/llamafactory:latest495```496 497This image is built on Ubuntu 22.04 (x86\_64), CUDA 12.4, Python 3.11, PyTorch 2.6.0, and Flash-attn 2.7.4.498 499Find the pre-built images: https://hub.docker.com/r/hiyouga/llamafactory/tags500 501Please refer to [build docker](#build-docker) to build the image yourself.502 503<details><summary>Setting up a virtual environment with <b>uv</b></summary>504 505Create an isolated Python environment with [uv](https://github.com/astral-sh/uv):506 507```bash508uv sync --extra torch --extra metrics --prerelease=allow509```510 511Run LLaMA-Factory in the isolated environment:512 513```bash514uv run --prerelease=allow llamafactory-cli train examples/train_lora/llama3_lora_pretrain.yaml515```516 517</details>518 519<details><summary>For Windows users</summary>520 521#### Install PyTorch522 523You need to manually install the GPU version of PyTorch on the Windows platform. Please refer to the [official website](https://pytorch.org/get-started/locally/) and the following command to install PyTorch with CUDA support:524 525```bash526pip uninstall torch torchvision torchaudio527pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu126528python -c "import torch; print(torch.cuda.is_available())"529```530 531If you see `True` then you have successfully installed PyTorch with CUDA support.532 533Try `dataloader_num_workers: 0` if you encounter `Can't pickle local object` error.534 535#### Install BitsAndBytes536 537If 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.538 539```bash540pip install https://github.com/jllllll/bitsandbytes-windows-webui/releases/download/wheels/bitsandbytes-0.41.2.post2-py3-none-win_amd64.whl541```542 543#### Install Flash Attention-2544 545To enable FlashAttention-2 on the Windows platform, please use the script from [flash-attention-windows-wheel](https://huggingface.co/lldacing/flash-attention-windows-wheel) to compile and install it by yourself.546 547</details>548 549<details><summary>For Ascend NPU users</summary>550 551To install LLaMA Factory on Ascend NPU devices, please upgrade Python to version 3.10 or higher and 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:552 553```bash554# replace the url according to your CANN version and devices555# install CANN Toolkit556wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C20SPC702/Ascend-cann-toolkit_8.0.0.alpha002_linux-"$(uname -i)".run557bash Ascend-cann-toolkit_8.0.0.alpha002_linux-"$(uname -i)".run --install558 559# install CANN Kernels560wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C20SPC702/Ascend-cann-kernels-910b_8.0.0.alpha002_linux-"$(uname -i)".run561bash Ascend-cann-kernels-910b_8.0.0.alpha002_linux-"$(uname -i)".run --install562 563# set env variables564source /usr/local/Ascend/ascend-toolkit/set_env.sh565```566 567| Requirement | Minimum | Recommend |568| ------------ | ------- | -------------- |569| CANN | 8.0.RC1 | 8.0.0.alpha002 |570| torch | 2.1.0 | 2.4.0 |571| torch-npu | 2.1.0 | 2.4.0.post2 |572| deepspeed | 0.13.2 | 0.13.2 |573| vllm-ascend | - | 0.7.3 |574 575Remember to use `ASCEND_RT_VISIBLE_DEVICES` instead of `CUDA_VISIBLE_DEVICES` to specify the device to use.576 577If you cannot infer model on NPU devices, try setting `do_sample: false` in the configurations.578 579Download 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)580 581#### Install BitsAndBytes582 583To use QLoRA based on bitsandbytes on Ascend NPU, please follow these 3 steps:584 5851. Manually compile bitsandbytes: Refer to [the installation documentation](https://huggingface.co/docs/bitsandbytes/installation?backend=Ascend+NPU&platform=Ascend+NPU) for the NPU version of bitsandbytes to complete the compilation and installation. The compilation requires a cmake version of at least 3.22.1 and a g++ version of at least 12.x.586 587```bash588# Install bitsandbytes from source589# Clone bitsandbytes repo, Ascend NPU backend is currently enabled on multi-backend-refactor branch590git clone -b multi-backend-refactor https://github.com/bitsandbytes-foundation/bitsandbytes.git591cd bitsandbytes/592 593# Install dependencies594pip install -r requirements-dev.txt595 596# Install the dependencies for the compilation tools. Note that the commands for this step may vary depending on the operating system. The following are provided for reference597apt-get install -y build-essential cmake598 599# Compile & install 600cmake -DCOMPUTE_BACKEND=npu -S .601make602pip install .603```604 6052. Install transformers from the main branch.606 607```bash608git clone -b main https://github.com/huggingface/transformers.git609cd transformers610pip install .611```612 6133. Set `double_quantization: false` in the configuration. You can refer to the [example](examples/train_qlora/llama3_lora_sft_bnb_npu.yaml).614 615</details>616 617### Data Preparation618 619Please refer to [data/README.md](data/README.md) for checking the details about the format of dataset files. You can use datasets on HuggingFace / ModelScope / Modelers hub, load the dataset in local disk, or specify a path to s3/gcs cloud storage.620 621> [!NOTE]622> Please update `data/dataset_info.json` to use your custom dataset.623 624You can also use **[Easy Dataset](https://github.com/ConardLi/easy-dataset)** or **[GraphGen](https://github.com/open-sciencelab/GraphGen)** to create synthetic data for fine-tuning.625 626### Quickstart627 628Use the following 3 commands to run LoRA **fine-tuning**, **inference** and **merging** of the Llama3-8B-Instruct model, respectively.629 630```bash631llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml632llamafactory-cli chat examples/inference/llama3_lora_sft.yaml633llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml634```635 636See [examples/README.md](examples/README.md) for advanced usage (including distributed training).637 638> [!TIP]639> Use `llamafactory-cli help` to show help information.640>641> Read [FAQs](https://github.com/hiyouga/LLaMA-Factory/issues/4614) first if you encounter any problems.642 643### Fine-Tuning with LLaMA Board GUI (powered by [Gradio](https://github.com/gradio-app/gradio))644 645```bash646llamafactory-cli webui647```648 649### Build Docker650 651For CUDA users:652 653```bash654cd docker/docker-cuda/655docker compose up -d656docker compose exec llamafactory bash657```658 659For Ascend NPU users:660 661```bash662cd docker/docker-npu/663docker compose up -d664docker compose exec llamafactory bash665```666 667For AMD ROCm users:668 669```bash670cd docker/docker-rocm/671docker compose up -d672docker compose exec llamafactory bash673```674 675<details><summary>Build without Docker Compose</summary>676 677For CUDA users:678 679```bash680docker build -f ./docker/docker-cuda/Dockerfile \681 --build-arg PIP_INDEX=https://pypi.org/simple \682 --build-arg EXTRAS=metrics \683 -t llamafactory:latest .684 685docker run -dit --ipc=host --gpus=all \686 -p 7860:7860 \687 -p 8000:8000 \688 --name llamafactory \689 llamafactory:latest690 691docker exec -it llamafactory bash692```693 694For Ascend NPU users:695 696```bash697docker build -f ./docker/docker-npu/Dockerfile \698 --build-arg PIP_INDEX=https://pypi.org/simple \699 --build-arg EXTRAS=torch-npu,metrics \700 -t llamafactory:latest .701 702docker run -dit --ipc=host \703 -v /usr/local/dcmi:/usr/local/dcmi \704 -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \705 -v /usr/local/Ascend/driver:/usr/local/Ascend/driver \706 -v /etc/ascend_install.info:/etc/ascend_install.info \707 -p 7860:7860 \708 -p 8000:8000 \709 --device /dev/davinci0 \710 --device /dev/davinci_manager \711 --device /dev/devmm_svm \712 --device /dev/hisi_hdc \713 --name llamafactory \714 llamafactory:latest715 716docker exec -it llamafactory bash717```718 719For AMD ROCm users:720 721```bash722docker build -f ./docker/docker-rocm/Dockerfile \723 --build-arg PIP_INDEX=https://pypi.org/simple \724 --build-arg EXTRAS=metrics \725 -t llamafactory:latest .726 727docker run -dit --ipc=host \728 -p 7860:7860 \729 -p 8000:8000 \730 --device /dev/kfd \731 --device /dev/dri \732 --name llamafactory \733 llamafactory:latest734 735docker exec -it llamafactory bash736```737 738</details>739 740<details><summary>Use Docker volumes</summary>741 742You can uncomment `VOLUME [ "/root/.cache/huggingface", "/app/shared_data", "/app/output" ]` in the Dockerfile to use data volumes.743 744When building the Docker image, use `-v ./hf_cache:/root/.cache/huggingface` argument to mount the local directory to the container. The following data volumes are available.745 746- `hf_cache`: Utilize Hugging Face cache on the host machine.747- `shared_data`: The directionary to store datasets on the host machine.748- `output`: Set export dir to this location so that the merged result can be accessed directly on the host machine.749 750</details>751 752### Deploy with OpenAI-style API and vLLM753 754```bash755API_PORT=8000 llamafactory-cli api examples/inference/llama3.yaml infer_backend=vllm vllm_enforce_eager=true756```757 758> [!TIP]759> Visit [this page](https://platform.openai.com/docs/api-reference/chat/create) for API document.760>761> Examples: [Image understanding](scripts/api_example/test_image.py) | [Function calling](scripts/api_example/test_toolcall.py)762 763### Download from ModelScope Hub764 765If you have trouble with downloading models and datasets from Hugging Face, you can use ModelScope.766 767```bash768export USE_MODELSCOPE_HUB=1 # `set USE_MODELSCOPE_HUB=1` for Windows769```770 771Train 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`.772 773### Download from Modelers Hub774 775You can also use Modelers Hub to download models and datasets.776 777```bash778export USE_OPENMIND_HUB=1 # `set USE_OPENMIND_HUB=1` for Windows779```780 781Train 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`.782 783### Use W&B Logger784 785To use [Weights & Biases](https://wandb.ai) for logging experimental results, you need to add the following arguments to yaml files.786 787```yaml788report_to: wandb789run_name: test_run # optional790```791 792Set `WANDB_API_KEY` to [your key](https://wandb.ai/authorize) when launching training tasks to log in with your W&B account.793 794### Use SwanLab Logger795 796To use [SwanLab](https://github.com/SwanHubX/SwanLab) for logging experimental results, you need to add the following arguments to yaml files.797 798```yaml799use_swanlab: true800swanlab_run_name: test_run # optional801```802 803When launching training tasks, you can log in to SwanLab in three ways:804 8051. Add `swanlab_api_key=<your_api_key>` to the yaml file, and set it to your [API key](https://swanlab.cn/settings).8062. Set the environment variable `SWANLAB_API_KEY` to your [API key](https://swanlab.cn/settings).8073. Use the `swanlab login` command to complete the login.808 809## Projects using LLaMA Factory810 811If you have a project that should be incorporated, please contact via email or create a pull request.812 813<details><summary>Click to show</summary>814 8151. Wang et al. ESRL: Efficient Sampling-based Reinforcement Learning for Sequence Generation. 2023. [[arxiv]](https://arxiv.org/abs/2308.02223)8161. Yu et al. Open, Closed, or Small Language Models for Text Classification? 2023. [[arxiv]](https://arxiv.org/abs/2308.10092)8171. 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)8181. Luceri et al. Leveraging Large Language Models to Detect Influence Campaigns in Social Media. 2023. [[arxiv]](https://arxiv.org/abs/2311.07816)8191. Zhang et al. Alleviating Hallucinations of Large Language Models through Induced Hallucinations. 2023. [[arxiv]](https://arxiv.org/abs/2312.15710)8201. 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)8211. 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)8221. Choi et al. FACT-GPT: Fact-Checking Augmentation via Claim Matching with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2402.05904)8231. Zhang et al. AutoMathText: Autonomous Data Selection with Language Models for Mathematical Texts. 2024. [[arxiv]](https://arxiv.org/abs/2402.07625)8241. Lyu et al. KnowTuning: Knowledge-aware Fine-tuning for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11176)8251. Yang et al. LaCo: Large Language Model Pruning via Layer Collaps. 2024. [[arxiv]](https://arxiv.org/abs/2402.11187)8261. 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)8271. Yang et al. Enhancing Empathetic Response Generation by Augmenting LLMs with Small-scale Empathetic Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11801)8281. 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)8291. Cao et al. Head-wise Shareable Attention for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11819)8301. 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)8311. Kim et al. Efficient and Effective Vocabulary Expansion Towards Multilingual Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.14714)8321. Yu et al. KIEval: A Knowledge-grounded Interactive Evaluation Framework for Large Language Models. ACL 2024. [[arxiv]](https://arxiv.org/abs/2402.15043)8331. Huang et al. Key-Point-Driven Data Synthesis with its Enhancement on Mathematical Reasoning. 2024. [[arxiv]](https://arxiv.org/abs/2403.02333)8341. Duan et al. Negating Negatives: Alignment without Human Positive Samples via Distributional Dispreference Optimization. 2024. [[arxiv]](https://arxiv.org/abs/2403.03419)8351. Xie and Schwertfeger. Empowering Robotics with Large Language Models: osmAG Map Comprehension with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2403.08228)8361. Wu et al. Large Language Models are Parallel Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2403.09073)8371. Zhang et al. EDT: Improving Large Language Models' Generation by Entropy-based Dynamic Temperature Sampling. 2024. [[arxiv]](https://arxiv.org/abs/2403.14541)8381. Weller et al. FollowIR: Evaluating and Teaching Information Retrieval Models to Follow Instructions. 2024. [[arxiv]](https://arxiv.org/abs/2403.15246)8391. 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)8401. Zan et al. CodeS: Natural Language to Code Repository via Multi-Layer Sketch. 2024. [[arxiv]](https://arxiv.org/abs/2403.16443)8411. Liu et al. Extensive Self-Contrast Enables Feedback-Free Language Model Alignment. 2024. [[arxiv]](https://arxiv.org/abs/2404.00604)8421. Luo et al. BAdam: A Memory Efficient Full Parameter Training Method for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.02827)8431. Du et al. Chinese Tiny LLM: Pretraining a Chinese-Centric Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2404.04167)8441. Ma et al. Parameter Efficient Quasi-Orthogonal Fine-Tuning via Givens Rotation. ICML 2024. [[arxiv]](https://arxiv.org/abs/2404.04316)8451. Liu et al. Dynamic Generation of Personalities with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.07084)8461. 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)8471. Huang et al. LLMTune: Accelerate Database Knob Tuning with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.11581)8481. 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)8491. Acikgoz et al. Hippocrates: An Open-Source Framework for Advancing Large Language Models in Healthcare. 2024. [[arxiv]](https://arxiv.org/abs/2404.16621)8501. Zhang et al. Small Language Models Need Strong Verifiers to Self-Correct Reasoning. ACL 2024 Findings. [[arxiv]](https://arxiv.org/abs/2404.17140)8511. Zhou et al. FREB-TQA: A Fine-Grained Robustness Evaluation Benchmark for Table Question Answering. NAACL 2024. [[arxiv]](https://arxiv.org/abs/2404.18585)8521. Xu et al. Large Language Models for Cyber Security: A Systematic Literature Review. 2024. [[arxiv]](https://arxiv.org/abs/2405.04760)8531. Dammu et al. "They are uncultured": Unveiling Covert Harms and Social Threats in LLM Generated Conversations. 2024. [[arxiv]](https://arxiv.org/abs/2405.05378)8541. Yi et al. A safety realignment framework via subspace-oriented model fusion for large language models. 2024. [[arxiv]](https://arxiv.org/abs/2405.09055)8551. Lou et al. SPO: Multi-Dimensional Preference Sequential Alignment With Implicit Reward Modeling. 2024. [[arxiv]](https://arxiv.org/abs/2405.12739)8561. Zhang et al. Getting More from Less: Large Language Models are Good Spontaneous Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2405.13816)8571. 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)8581. Zihong Chen. Sentence Segmentation and Sentence Punctuation Based on XunziALLM. 2024. [[paper]](https://aclanthology.org/2024.lt4hala-1.30)8591. Gao et al. The Best of Both Worlds: Toward an Honest and Helpful Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2406.00380)8601. 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)8611. Hu et al. Computational Limits of Low-Rank Adaptation (LoRA) for Transformer-Based Models. 2024. [[arxiv]](https://arxiv.org/abs/2406.03136)8621. Ge et al. Time Sensitive Knowledge Editing through Efficient Finetuning. ACL 2024. [[arxiv]](https://arxiv.org/abs/2406.04496)8631. 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)8641. Song et al. Turbo Sparse: Achieving LLM SOTA Performance with Minimal Activated Parameters. 2024. [[arxiv]](https://arxiv.org/abs/2406.05955)8651. Gu et al. RWKV-CLIP: A Robust Vision-Language Representation Learner. 2024. [[arxiv]](https://arxiv.org/abs/2406.06973)8661. Chen et al. Advancing Tool-Augmented Large Language Models: Integrating Insights from Errors in Inference Trees. 2024. [[arxiv]](https://arxiv.org/abs/2406.07115)8671. Zhu et al. Are Large Language Models Good Statisticians?. 2024. [[arxiv]](https://arxiv.org/abs/2406.07815)8681. Li et al. Know the Unknown: An Uncertainty-Sensitive Method for LLM Instruction Tuning. 2024. [[arxiv]](https://arxiv.org/abs/2406.10099)8691. 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)8701. 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)8711. Lin et al. FVEL: Interactive Formal Verification Environment with Large Language Models via Theorem Proving. 2024. [[arxiv]](https://arxiv.org/abs/2406.14408)8721. 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)8731. Feng et al. SS-Bench: A Benchmark for Social Story Generation and Evaluation. 2024. [[arxiv]](https://arxiv.org/abs/2406.15695)8741. Feng et al. Self-Constructed Context Decompilation with Fined-grained Alignment Enhancement. 2024. [[arxiv]](https://arxiv.org/abs/2406.17233)8751. Liu et al. Large Language Models for Cuffless Blood Pressure Measurement From Wearable Biosignals. 2024. [[arxiv]](https://arxiv.org/abs/2406.18069)8761. 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)8771. 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)8781. Yang et al. Financial Knowledge Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2407.00365)8791. Lin et al. DogeRM: Equipping Reward Models with Domain Knowledge through Model Merging. 2024. [[arxiv]](https://arxiv.org/abs/2407.01470)8801. 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)8811. Huang et al. RoLoRA: Fine-tuning Rotated Outlier-free LLMs for Effective Weight-Activation Quantization. 2024. [[arxiv]](https://arxiv.org/abs/2407.08044)8821. Jiang et al. LLM-Collaboration on Automatic Science Journalism for the General Audience. 2024. [[arxiv]](https://arxiv.org/abs/2407.09756)8831. Inouye et al. Applied Auto-tuning on LoRA Hyperparameters. 2024. [[paper]](https://scholarcommons.scu.edu/cseng_senior/272/)8841. Qi et al. Research on Tibetan Tourism Viewpoints information generation system based on LLM. 2024. [[arxiv]](https://arxiv.org/abs/2407.13561)8851. Xu et al. Course-Correction: Safety Alignment Using Synthetic Preferences. 2024. [[arxiv]](https://arxiv.org/abs/2407.16637)8861. Sun et al. LAMBDA: A Large Model Based Data Agent. 2024. [[arxiv]](https://arxiv.org/abs/2407.17535)8871. 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)8881. Yu et al. Correcting Negative Bias in Large Language Models through Negative Attention Score Alignment. 2024. [[arxiv]](https://arxiv.org/abs/2408.00137)8891. 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)8901. 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)8911. 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)8921. Xia et al. Understanding the Performance and Estimating the Cost of LLM Fine-Tuning. 2024. [[arxiv]](https://arxiv.org/abs/2408.04693)8931. 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)8941. Xia et al. Using Pre-trained Language Model for Accurate ESG Prediction. FinNLP 2024. [[paper]](https://aclanthology.org/2024.finnlp-2.1/)8951. 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)8961. Bai et al. Aligning Large Language Model with Direct Multi-Preference Optimization for Recommendation. CIKM 2024. [[paper]](https://dl.acm.org/doi/10.1145/3627673.3679611)8971. Zhang et al. CPsyCoun: A Report-based Multi-turn Dialogue Reconstruction and Evaluation Framework for Chinese Psychological Counseling. ACL 2024. [[paper]](https://aclanthology.org/2024.findings-acl.830.pdf)8981. **[StarWhisper](https://github.com/Yu-Yang-Li/StarWhisper)**: A large language model for Astronomy, based on ChatGLM2-6B and Qwen-14B.8991. **[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.9001. **[Sunsimiao](https://github.com/X-D-Lab/Sunsimiao)**: A large language model specialized in Chinese medical domain, based on Baichuan-7B and ChatGLM-6B.9011. **[CareGPT](https://github.com/WangRongsheng/CareGPT)**: A series of large language models for Chinese medical domain, based on LLaMA2-7B and Baichuan-13B.9021. **[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.9031. **[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)9041. **[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.9051. **[AutoRE](https://github.com/THUDM/AutoRE)**: A document-level relation extraction system based on large language models.9061. **[NVIDIA RTX AI Toolkit](https://github.com/NVIDIA/RTX-AI-Toolkit)**: SDKs for fine-tuning LLMs on Windows PC for NVIDIA RTX.9071. **[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.9081. **[RAG-Retrieval](https://github.com/NLPJCL/RAG-Retrieval)**: A full pipeline for RAG retrieval model fine-tuning, inference, and distillation. [[blog]](https://zhuanlan.zhihu.com/p/987727357)9091. **[360-LLaMA-Factory](https://github.com/Qihoo360/360-LLaMA-Factory)**: A modified library that supports long sequence SFT & DPO using ring attention.9101. **[Sky-T1](https://novasky-ai.github.io/posts/sky-t1/)**: An o1-like model fine-tuned by NovaSky AI with very small cost.9111. **[WeClone](https://github.com/xming521/WeClone)**: One-stop solution for creating your digital avatar from chat logs.9121. **[EmoLLM](https://github.com/SmartFlowAI/EmoLLM)**: A project about large language models (LLMs) and mental health.913</details>914 915## License916 917This repository is licensed under the [Apache-2.0 License](LICENSE).918 919Please 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) / [GPT-2](https://github.com/openai/gpt-2/blob/master/LICENSE) / [Granite](LICENSE) / [Index](https://huggingface.co/IndexTeam/Index-1.9B/blob/main/LICENSE) / [InternLM](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/) / [Llama 4](https://github.com/meta-llama/llama-models/blob/main/models/llama4/LICENSE) / [MiniCPM](https://github.com/OpenBMB/MiniCPM/blob/main/MiniCPM%20Model%20License.md) / [Mistral/Mixtral/Pixtral](LICENSE) / [OLMo](LICENSE) / [Phi-1.5/Phi-2](https://huggingface.co/microsoft/phi-1_5/resolve/main/Research%20License.docx) / [Phi-3/Phi-4](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/blob/main/LICENSE) / [Qwen](https://github.com/QwenLM/Qwen/blob/main/Tongyi%20Qianwen%20LICENSE%20AGREEMENT) / [Skywork](https://huggingface.co/Skywork/Skywork-13B-base/blob/main/Skywork%20Community%20License.pdf) / [StarCoder 2](https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement) / [TeleChat2](https://huggingface.co/Tele-AI/telechat-7B/blob/main/TeleChat%E6%A8%A1%E5%9E%8B%E7%A4%BE%E5%8C%BA%E8%AE%B8%E5%8F%AF%E5%8D%8F%E8%AE%AE.pdf) / [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)920 921## Citation922 923If this work is helpful, please kindly cite as:924 925```bibtex926@inproceedings{zheng2024llamafactory,927 title={LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models},928 author={Yaowei Zheng and Richong Zhang and Junhao Zhang and Yanhan Ye and Zheyan Luo and Zhangchi Feng and Yongqiang Ma},929 booktitle={Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)},930 address={Bangkok, Thailand},931 publisher={Association for Computational Linguistics},932 year={2024},933 url={http://arxiv.org/abs/2403.13372}934}935```936 937## Acknowledgement938 939This 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.940 941## Star History942 943944