update parser

Former-commit-id: 5262c8702382ff8bc36a172387bc4c8949f326ea

.dockerignore

@@ -4,8 +4,10 @@
 .venv
 cache
 data
-examples
+docker
+saves
+hf_cache
+output
 .dockerignore
 .gitattributes
 .gitignore
-Dockerfile

.github/ISSUE_TEMPLATE/bug-report.yml

@@ -1,18 +1,36 @@
 name: "\U0001F41B Bug / Help"
 description: Create a report to help us improve the LLaMA Factory
 body:
+  - type: markdown
+    attributes:
+      value: |
+        Issues included in **FAQs** or those with **insufficient** information may be closed without a response.
+        包含在**常见问题**内或提供信息**不完整**的 issues 可能不会被回复。
+
   - type: checkboxes
     id: reminder
     attributes:
       label: Reminder
       description: |
-        Please ensure you have read the README carefully and searched the existing issues.
+        Please ensure you have read the README carefully and searched the existing issues (including FAQs).
-        请确保您已经认真阅读了 README 并且搜索过现有的 Issue。
+        请确保您已经认真阅读了 README 并且搜索过现有的 issues（包括常见问题）。
 
       options:
         - label: I have read the README and searched the existing issues.
           required: true
 
+  - type: textarea
+    id: system-info
+    validations:
+      required: true
+    attributes:
+      label: System Info
+      description: |
+        Please share your system info with us. You can run the command **llamafactory-cli env** and copy-paste its output below.
+        请提供您的系统信息。您可以在命令行运行 **llamafactory-cli env** 并将其输出复制到该文本框中。
+
+      placeholder: llamafactory version, platform, python version, ...
+
   - type: textarea
     id: reproduction
     validations:
@@ -26,7 +44,9 @@ body:
         请合理使用 Markdown 标签来格式化您的文本。
 
       placeholder: |
-        python src/train_bash.py ...
+        ```bash
+        llamafactory-cli train ...
+        ```
 
   - type: textarea
     id: expected-behavior
@@ -38,18 +58,6 @@ body:
         Please provide a clear and concise description of what you would expect to happen.
         请提供您原本的目的，即这段代码的期望行为。
 
-  - type: textarea
-    id: system-info
-    validations:
-      required: false
-    attributes:
-      label: System Info
-      description: |
-        Please share your system info with us. You can run the command **transformers-cli env** and copy-paste its output below.
-        请提供您的系统信息。您可以在命令行运行 **transformers-cli env** 并将其输出复制到该文本框中。
-
-      placeholder: transformers version, platform, python version, ...
-
   - type: textarea
     id: others
     validations:

.github/PULL_REQUEST_TEMPLATE.md

@@ -5,3 +5,4 @@ Fixes # (issue)
 ## Before submitting
 
 - [ ] Did you read the [contributor guideline](https://github.com/hiyouga/LLaMA-Factory/blob/main/.github/CONTRIBUTING.md)?
+- [ ] Did you write any new necessary tests?

.github/workflows/label_issue.yml

@@ -0,0 +1,30 @@
+name: label_issue
+
+on:
+  issues:
+    types:
+      - opened
+
+jobs:
+  label_issue:
+    runs-on: ubuntu-latest
+
+    permissions:
+      issues: write
+
+    steps:
+      - env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          ISSUE_URL: ${{ github.event.issue.html_url }}
+          ISSUE_TITLE: ${{ github.event.issue.title }}
+        run: |
+          LABEL=pending
+          NPU_KEYWORDS=(npu huawei ascend 华为 昇腾)
+          ISSUE_TITLE_LOWER=$(echo $ISSUE_TITLE | tr '[:upper:]' '[:lower:]')
+          for KEYWORD in ${NPU_KEYWORDS[@]}; do
+            if [[ $ISSUE_TITLE_LOWER == *$KEYWORD* ]] && [[ $ISSUE_TITLE_LOWER != *input* ]]; then
+              LABEL=pending,npu
+              break
+            fi
+          done
+          gh issue edit $ISSUE_URL --add-label $LABEL

.github/workflows/publish.yml

@@ -0,0 +1,40 @@
+name: publish
+
+on:
+  release:
+    types:
+      - published
+
+jobs:
+  publish:
+    name: Upload release to PyPI
+
+    runs-on: ubuntu-latest
+
+    environment:
+      name: release
+      url: https://pypi.org/p/llamafactory
+
+    permissions:
+      id-token: write
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.8"
+      
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install build
+      
+      - name: Build package
+        run: |
+          python -m build
+      
+      - name: Publish package
+        uses: pypa/gh-action-pypi-publish@release/v1

.github/workflows/tests.yml

@@ -2,28 +2,50 @@ name: tests
 
 on:
   push:
-    branches: [ "main" ]
+    branches:
+      - main
+    paths:
+      - "**.py"
+      - "requirements.txt"
+      - ".github/workflows/*.yml"
   pull_request:
-    branches: [ "main" ]
+    branches:
+      - main
+    paths:
+      - "**.py"
+      - "requirements.txt"
+      - ".github/workflows/*.yml"
 
 jobs:
-  check_code_quality:
+  tests:
-
     runs-on: ubuntu-latest
 
+    environment:
+      name: tests
+
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
+
     steps:
-      - uses: actions/checkout@v4
+      - name: Checkout
+        uses: actions/checkout@v4
 
       - name: Set up Python
         uses: actions/setup-python@v5
         with:
           python-version: "3.8"
+          cache: "pip"
+          cache-dependency-path: "setup.py"
 
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip
-          python -m pip install ruff
+          python -m pip install ".[torch,dev]"
 
       - name: Check quality
         run: |
-            make style && make quality
+          make style && make quality
+
+      - name: Test with pytest
+        run: |
+          make test

.gitignore

@@ -160,6 +160,8 @@ cython_debug/
 .idea/
 
 # custom .gitignore
-user.config
-saves/
 cache/
+config/
+saves/
+output/
+wandb/

CITATION.cff

@@ -12,12 +12,16 @@ authors:
   given-names: "Yanhan"
 - family-names: "Luo"
   given-names: "Zheyan"
+- family-names: "Feng"
+  given-names: "Zhangchi"
 - family-names: "Ma"
   given-names: "Yongqiang"
 title: "LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models"
 url: "https://arxiv.org/abs/2403.13372"
 preferred-citation:
-  type: article
+  type: conference-paper
+  conference:
+    name: "Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)"
   authors:
     - family-names: "Zheng"
       given-names: "Yaowei"
@@ -29,9 +33,12 @@ preferred-citation:
       given-names: "Yanhan"
     - family-names: "Luo"
       given-names: "Zheyan"
+    - family-names: "Feng"
+      given-names: "Zhangchi"
     - family-names: "Ma"
       given-names: "Yongqiang"
-  journal: "arXiv preprint arXiv:2403.13372"
   title: "LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models"
   url: "https://arxiv.org/abs/2403.13372"
   year: 2024
+  publisher: "Association for Computational Linguistics"
+  address: "Bangkok, Thailand"

Dockerfile

@@ -1,14 +0,0 @@
-FROM nvcr.io/nvidia/pytorch:24.01-py3
-
-WORKDIR /app
-
-COPY requirements.txt /app/
-RUN pip install -r requirements.txt
-
-COPY . /app/
-RUN pip install -e .[deepspeed,metrics,bitsandbytes,qwen]
-
-VOLUME [ "/root/.cache/huggingface/", "/app/data", "/app/output" ]
-EXPOSE 7860
-
-CMD [ "llamafactory-cli", "webui" ]

MANIFEST.in

@@ -0,0 +1 @@
+include LICENSE requirements.txt

Makefile

@@ -1,4 +1,4 @@
-.PHONY: quality style
+.PHONY: quality style test
 
 check_dirs := scripts src tests
 
@@ -9,3 +9,6 @@ quality:
 style:
 	ruff check $(check_dirs) --fix
 	ruff format $(check_dirs)
+
+test:
+	CUDA_VISIBLE_DEVICES= pytest tests/

README.md

@@ -3,19 +3,19 @@
 [![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/)
+[![PyPI](https://img.shields.io/pypi/v/llamafactory)](https://pypi.org/project/llamafactory/)
-[![Downloads](https://static.pepy.tech/badge/llmtuner)](https://pypi.org/project/llmtuner/)
+[![Citation](https://img.shields.io/badge/citation-72-green)](#projects-using-llama-factory)
-[![Citation](https://img.shields.io/badge/citation-44-green)](#projects-using-llama-factory)
 [![GitHub pull request](https://img.shields.io/badge/PRs-welcome-blue)](https://github.com/hiyouga/LLaMA-Factory/pulls)
 [![Discord](https://dcbadge.vercel.app/api/server/rKfvV9r9FK?compact=true&style=flat)](https://discord.gg/rKfvV9r9FK)
 [![Twitter](https://img.shields.io/twitter/follow/llamafactory_ai)](https://twitter.com/llamafactory_ai)
+[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing)
+[![Open in DSW](https://gallery.pai-ml.com/assets/open-in-dsw.svg)](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory)
 [![Spaces](https://img.shields.io/badge/🤗-Open%20in%20Spaces-blue)](https://huggingface.co/spaces/hiyouga/LLaMA-Board)
 [![Studios](https://img.shields.io/badge/ModelScope-Open%20in%20Studios-blue)](https://modelscope.cn/studios/hiyouga/LLaMA-Board)
-[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing)
 
 [![GitHub Tread](https://trendshift.io/api/badge/repositories/4535)](https://trendshift.io/repositories/4535)
 
-👋 Join our [WeChat](assets/wechat.jpg).
+👋 Join our [WeChat](assets/wechat.jpg) or [NPU user group](assets/wechat_npu.jpg).
 
 \[ English | [中文](README_zh.md) \]
 
@@ -26,6 +26,7 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/9840a653-7e9c-41c8-ae89
 Choose your path:
 
 - **Colab**: https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing
+- **PAI-DSW**: https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory
 - **Local machine**: Please refer to [usage](#getting-started)
 
 ## Table of Contents
@@ -46,9 +47,9 @@ Choose your path:
 ## Features
 
 - **Various models**: LLaMA, LLaVA, Mistral, Mixtral-MoE, Qwen, Yi, Gemma, Baichuan, ChatGLM, Phi, etc.
-- **Integrated methods**: (Continuous) pre-training, (multimodal) supervised fine-tuning, reward modeling, PPO, DPO and ORPO.
+- **Integrated methods**: (Continuous) pre-training, (multimodal) supervised fine-tuning, reward modeling, PPO, DPO, KTO, ORPO, etc.
-- **Scalable resources**: 32-bit full-tuning, 16-bit freeze-tuning, 16-bit LoRA and 2/4/8-bit QLoRA via AQLM/AWQ/GPTQ/LLM.int8.
+- **Scalable resources**: 16-bit full-tuning, freeze-tuning, LoRA and 2/3/4/5/6/8-bit QLoRA via AQLM/AWQ/GPTQ/LLM.int8/HQQ/EETQ.
-- **Advanced algorithms**: GaLore, BAdam, DoRA, LongLoRA, LLaMA Pro, Mixture-of-Depths, LoRA+, LoftQ and Agent tuning.
+- **Advanced algorithms**: GaLore, BAdam, DoRA, LongLoRA, LLaMA Pro, Mixture-of-Depths, LoRA+, LoftQ, PiSSA and Agent tuning.
 - **Practical tricks**: FlashAttention-2, Unsloth, RoPE scaling, NEFTune and rsLoRA.
 - **Experiment monitors**: LlamaBoard, TensorBoard, Wandb, MLflow, etc.
 - **Faster inference**: OpenAI-style API, Gradio UI and CLI with vLLM worker.
@@ -70,14 +71,22 @@ Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/
 
 ## Changelog
 
-[24/05/14] We supported training and inference on the Ascend NPU devices. Check [installation](#installation) section for details.
+[24/06/16] We support **[PiSSA](https://arxiv.org/abs/2404.02948)** algorithm. See [examples](examples/README.md) for usage.
 
-[24/05/13] We supported fine-tuning the **Yi-1.5** series models.
+[24/06/07] We supported fine-tuning the **[Qwen2](https://qwenlm.github.io/blog/qwen2/)** and **[GLM-4](https://github.com/THUDM/GLM-4)** models.
 
-[24/04/26] We supported fine-tuning the **LLaVA-1.5** multimodal LLMs. See [examples](examples/README.md) for usage.
+[24/05/26] We supported **[SimPO](https://arxiv.org/abs/2405.14734)** algorithm for preference learning. See [examples](examples/README.md) for usage.
 
 <details><summary>Full Changelog</summary>
 
+[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 `gemma` template for chat completion.
+
+[24/05/18] We supported **[KTO](https://arxiv.org/abs/2402.01306)** algorithm for preference learning. See [examples](examples/README.md) for usage.
+
+[24/05/14] We supported training and inference on the Ascend NPU devices. Check [installation](#installation) section for details.
+
+[24/04/26] We supported fine-tuning the **LLaVA-1.5** multimodal LLMs. See [examples](examples/README.md) for usage.
+
 [24/04/22] We provided a **[Colab notebook](https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing)** for fine-tuning the Llama-3 model on a free T4 GPU. Two Llama-3-derived models fine-tuned using LLaMA Factory are available at Hugging Face, check [Llama3-8B-Chinese-Chat](https://huggingface.co/shenzhi-wang/Llama3-8B-Chinese-Chat) and [Llama3-Chinese](https://huggingface.co/zhichen/Llama3-Chinese) for details.
 
 [24/04/21] We supported **[Mixture-of-Depths](https://arxiv.org/abs/2404.02258)** according to [AstraMindAI's implementation](https://github.com/astramind-ai/Mixture-of-depths). See [examples](examples/README.md) for usage.
@@ -104,7 +113,7 @@ Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/
 
 [24/02/05] Qwen1.5 (Qwen2 beta version) series models are supported in LLaMA-Factory. Check this [blog post](https://qwenlm.github.io/blog/qwen1.5/) for details.
 
-[24/01/18] We supported **agent tuning** for most models, equipping model with tool using abilities by fine-tuning with `dataset: glaive_toolcall`.
+[24/01/18] We supported **agent tuning** for most models, equipping model with tool using abilities by fine-tuning with `dataset: glaive_toolcall_en`.
 
 [23/12/23] We supported **[unsloth](https://github.com/unslothai/unsloth)**'s implementation to boost LoRA tuning for the LLaMA, Mistral and Yi models. Try `use_unsloth: true` argument to activate unsloth patch. It achieves **170%** speed in our benchmark, check [this page](https://github.com/hiyouga/LLaMA-Factory/wiki/Performance-comparison) for details.
 
@@ -142,43 +151,41 @@ Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/
 
 ## Supported Models
 
-| Model                                                    | Model size                       | Default module    | Template  |
+| Model                                                        | Model size                       | Template  |
-| -------------------------------------------------------- | -------------------------------- | ----------------- | --------- |
+| ------------------------------------------------------------ | -------------------------------- | --------- |
-| [Baichuan2](https://huggingface.co/baichuan-inc)         | 7B/13B                           | W_pack            | baichuan2 |
+| [Baichuan 2](https://huggingface.co/baichuan-inc)            | 7B/13B                           | baichuan2 |
-| [BLOOM](https://huggingface.co/bigscience)               | 560M/1.1B/1.7B/3B/7.1B/176B      | query_key_value   | -         |
+| [BLOOM/BLOOMZ](https://huggingface.co/bigscience)            | 560M/1.1B/1.7B/3B/7.1B/176B      | -         |
-| [BLOOMZ](https://huggingface.co/bigscience)              | 560M/1.1B/1.7B/3B/7.1B/176B      | query_key_value   | -         |
+| [ChatGLM3](https://huggingface.co/THUDM)                     | 6B                               | chatglm3  |
-| [ChatGLM3](https://huggingface.co/THUDM)                 | 6B                               | query_key_value   | chatglm3  |
+| [Command R](https://huggingface.co/CohereForAI)              | 35B/104B                         | cohere    |
-| [Command-R](https://huggingface.co/CohereForAI)          | 35B/104B                         | q_proj,v_proj     | cohere    |
+| [DeepSeek (Code/MoE)](https://huggingface.co/deepseek-ai)    | 7B/16B/67B/236B                  | deepseek  |
-| [DeepSeek (MoE)](https://huggingface.co/deepseek-ai)     | 7B/16B/67B/236B                  | q_proj,v_proj     | deepseek  |
+| [Falcon](https://huggingface.co/tiiuae)                      | 7B/11B/40B/180B                  | falcon    |
-| [Falcon](https://huggingface.co/tiiuae)                  | 7B/40B/180B                      | query_key_value   | falcon    |
+| [Gemma/Gemma 2/CodeGemma](https://huggingface.co/google)     | 2B/7B/9B/27B                     | gemma     |
-| [Gemma/CodeGemma](https://huggingface.co/google)         | 2B/7B                            | q_proj,v_proj     | gemma     |
+| [GLM-4](https://huggingface.co/THUDM)                        | 9B                               | glm4      |
-| [InternLM2](https://huggingface.co/internlm)             | 7B/20B                           | wqkv              | intern2   |
+| [InternLM2](https://huggingface.co/internlm)                 | 7B/20B                           | intern2   |
-| [LLaMA](https://github.com/facebookresearch/llama)       | 7B/13B/33B/65B                   | q_proj,v_proj     | -         |
+| [Llama](https://github.com/facebookresearch/llama)           | 7B/13B/33B/65B                   | -         |
-| [LLaMA-2](https://huggingface.co/meta-llama)             | 7B/13B/70B                       | q_proj,v_proj     | llama2    |
+| [Llama 2](https://huggingface.co/meta-llama)                 | 7B/13B/70B                       | llama2    |
-| [LLaMA-3](https://huggingface.co/meta-llama)             | 8B/70B                           | q_proj,v_proj     | llama3    |
+| [Llama 3](https://huggingface.co/meta-llama)                 | 8B/70B                           | llama3    |
-| [LLaVA-1.5](https://huggingface.co/llava-hf)             | 7B/13B                           | q_proj,v_proj     | vicuna    |
+| [LLaVA-1.5](https://huggingface.co/llava-hf)                 | 7B/13B                           | vicuna    |
-| [Mistral/Mixtral](https://huggingface.co/mistralai)      | 7B/8x7B/8x22B                    | q_proj,v_proj     | mistral   |
+| [Mistral/Mixtral](https://huggingface.co/mistralai)          | 7B/8x7B/8x22B                    | mistral   |
-| [OLMo](https://huggingface.co/allenai)                   | 1B/7B                            | q_proj,v_proj     | -         |
+| [OLMo](https://huggingface.co/allenai)                       | 1B/7B                            | -         |
-| [Phi-1.5/2](https://huggingface.co/microsoft)            | 1.3B/2.7B                        | q_proj,v_proj     | -         |
+| [PaliGemma](https://huggingface.co/google)                   | 3B                               | gemma     |
-| [Phi-3](https://huggingface.co/microsoft)                | 3.8B                             | qkv_proj          | phi       |
+| [Phi-1.5/Phi-2](https://huggingface.co/microsoft)            | 1.3B/2.7B                        | -         |
-| [Qwen](https://huggingface.co/Qwen)                      | 1.8B/7B/14B/72B                  | c_attn            | qwen      |
+| [Phi-3](https://huggingface.co/microsoft)                    | 4B/7B/14B                        | phi       |
-| [Qwen1.5 (Code/MoE)](https://huggingface.co/Qwen)        | 0.5B/1.8B/4B/7B/14B/32B/72B/110B | q_proj,v_proj     | qwen      |
+| [Qwen/Qwen1.5/Qwen2 (Code/MoE)](https://huggingface.co/Qwen) | 0.5B/1.5B/4B/7B/14B/32B/72B/110B | qwen      |
-| [StarCoder2](https://huggingface.co/bigcode)             | 3B/7B/15B                        | q_proj,v_proj     | -         |
+| [StarCoder 2](https://huggingface.co/bigcode)                | 3B/7B/15B                        | -         |
-| [XVERSE](https://huggingface.co/xverse)                  | 7B/13B/65B                       | q_proj,v_proj     | xverse    |
+| [XVERSE](https://huggingface.co/xverse)                      | 7B/13B/65B                       | xverse    |
-| [Yi (1/1.5)](https://huggingface.co/01-ai)               | 6B/9B/34B                        | q_proj,v_proj     | yi        |
+| [Yi/Yi-1.5](https://huggingface.co/01-ai)                    | 6B/9B/34B                        | yi        |
-| [Yi-VL](https://huggingface.co/01-ai)                    | 6B/34B                           | q_proj,v_proj     | yi_vl     |
+| [Yi-VL](https://huggingface.co/01-ai)                        | 6B/34B                           | yi_vl     |
-| [Yuan](https://huggingface.co/IEITYuan)                  | 2B/51B/102B                      | q_proj,v_proj     | yuan      |
+| [Yuan 2](https://huggingface.co/IEITYuan)                    | 2B/51B/102B                      | yuan      |
 
 > [!NOTE]
-> **Default module** is used for the `--lora_target` argument, you can use `--lora_target all` to specify all the available modules for better convergence.
+> 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.
->
-> For the "base" models, the `--template` argument can be chosen from `default`, `alpaca`, `vicuna` etc. But make sure to use the **corresponding template** for the "instruct/chat" models.
 >
 > Remember to use the **SAME** template in training and inference.
 
-Please refer to [constants.py](src/llmtuner/extras/constants.py) for a full list of models we supported.
+Please refer to [constants.py](src/llamafactory/extras/constants.py) for a full list of models we supported.
 
-You also can add a custom chat template to [template.py](src/llmtuner/data/template.py).
+You also can add a custom chat template to [template.py](src/llamafactory/data/template.py).
 
 ## Supported Training Approaches
 
@@ -189,7 +196,9 @@ You also can add a custom chat template to [template.py](src/llmtuner/data/templ
 | Reward Modeling        | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
 | PPO Training           | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
 | DPO Training           | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+| KTO Training           | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
 | ORPO Training          | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+| SimPO Training         | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
 
 ## Provided Datasets
 
@@ -202,6 +211,8 @@ You also can add a custom chat template to [template.py](src/llmtuner/data/templ
 - [Wikipedia (zh)](https://huggingface.co/datasets/pleisto/wikipedia-cn-20230720-filtered)
 - [Pile (en)](https://huggingface.co/datasets/EleutherAI/pile)
 - [SkyPile (zh)](https://huggingface.co/datasets/Skywork/SkyPile-150B)
+- [FineWeb (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb)
+- [FineWeb-Edu (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu)
 - [The Stack (en)](https://huggingface.co/datasets/bigcode/the-stack)
 - [StarCoder (en)](https://huggingface.co/datasets/bigcode/starcoderdata)
 
@@ -209,12 +220,12 @@ You also can add a custom chat template to [template.py](src/llmtuner/data/templ
 
 <details><summary>Supervised fine-tuning datasets</summary>
 
-- [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)
-- [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca)
-- [Alpaca GPT4 (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
 - [Identity (en&zh)](data/identity.json)
-- [Open Assistant (zh)](https://huggingface.co/datasets/OpenAssistant/oasst1)
+- [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)
-- [ShareGPT (zh)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT/tree/main/Chinese-instruction-collection)
+- [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca-3)
+- [Alpaca GPT4 (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
+- [Glaive Function Calling V2 (en&zh)](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2)
+- [LIMA (en)](https://huggingface.co/datasets/GAIR/lima)
 - [Guanaco Dataset (multilingual)](https://huggingface.co/datasets/JosephusCheung/GuanacoDataset)
 - [BELLE 2M (zh)](https://huggingface.co/datasets/BelleGroup/train_2M_CN)
 - [BELLE 1M (zh)](https://huggingface.co/datasets/BelleGroup/train_1M_CN)
@@ -223,7 +234,6 @@ You also can add a custom chat template to [template.py](src/llmtuner/data/templ
 - [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)
@@ -236,15 +246,19 @@ You also can add a custom chat template to [template.py](src/llmtuner/data/templ
 - [WebNovel (zh)](https://huggingface.co/datasets/zxbsmk/webnovel_cn)
 - [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
 - [deepctrl (en&zh)](https://www.modelscope.cn/datasets/deepctrl/deepctrl-sft-data)
-- [Ad Gen (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)
+- [Advertise Generating (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)
 - [ShareGPT Hyperfiltered (en)](https://huggingface.co/datasets/totally-not-an-llm/sharegpt-hyperfiltered-3k)
 - [ShareGPT4 (en&zh)](https://huggingface.co/datasets/shibing624/sharegpt_gpt4)
 - [UltraChat 200k (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k)
 - [AgentInstruct (en)](https://huggingface.co/datasets/THUDM/AgentInstruct)
 - [LMSYS Chat 1M (en)](https://huggingface.co/datasets/lmsys/lmsys-chat-1m)
 - [Evol Instruct V2 (en)](https://huggingface.co/datasets/WizardLM/WizardLM_evol_instruct_V2_196k)
-- [Glaive Function Calling V2 (en)](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2)
 - [Cosmopedia (en)](https://huggingface.co/datasets/HuggingFaceTB/cosmopedia)
+- [STEM (zh)](https://huggingface.co/datasets/hfl/stem_zh_instruction)
+- [Ruozhiba (zh)](https://huggingface.co/datasets/hfl/ruozhiba_gpt4_turbo)
+- [Neo-sft (zh)](https://huggingface.co/datasets/m-a-p/neo_sft_phase2)
+- [WebInstructSub (en)](https://huggingface.co/datasets/TIGER-Lab/WebInstructSub)
+- [Magpie-Pro-300K-Filtered (en)](https://huggingface.co/datasets/Magpie-Align/Magpie-Pro-300K-Filtered)
 - [LLaVA mixed (en&zh)](https://huggingface.co/datasets/BUAADreamer/llava-en-zh-300k)
 - [Open Assistant (de)](https://huggingface.co/datasets/mayflowergmbh/oasst_de)
 - [Dolly 15k (de)](https://huggingface.co/datasets/mayflowergmbh/dolly-15k_de)
@@ -260,13 +274,13 @@ You also can add a custom chat template to [template.py](src/llmtuner/data/templ
 
 <details><summary>Preference datasets</summary>
 
-- [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
-- [GPT-4 Generated Data (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
-- [Orca DPO (en)](https://huggingface.co/datasets/Intel/orca_dpo_pairs)
-- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
 - [DPO mixed (en&zh)](https://huggingface.co/datasets/hiyouga/DPO-En-Zh-20k)
-- [Open Assistant (zh)](https://huggingface.co/datasets/OpenAssistant/oasst1)
+- [UltraFeedback (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized)
+- [Orca DPO Pairs (en)](https://huggingface.co/datasets/Intel/orca_dpo_pairs)
+- [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
+- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
 - [Orca DPO (de)](https://huggingface.co/datasets/mayflowergmbh/intel_orca_dpo_pairs_de)
+- [KTO mixed (en)](https://huggingface.co/datasets/argilla/kto-mix-15k)
 
 </details>
 
@@ -281,21 +295,21 @@ huggingface-cli login
 
 | Mandatory    | Minimum | Recommend |
 | ------------ | ------- | --------- |
-| python       | 3.8     | 3.10      |
+| python       | 3.8     | 3.11      |
-| torch        | 1.13.1  | 2.2.0     |
+| torch        | 1.13.1  | 2.3.0     |
-| transformers | 4.37.2  | 4.40.1    |
+| transformers | 4.41.2  | 4.41.2    |
-| datasets     | 2.14.3  | 2.19.1    |
+| datasets     | 2.16.0  | 2.19.2    |
-| accelerate   | 0.27.2  | 0.30.0    |
+| accelerate   | 0.30.1  | 0.30.1    |
-| peft         | 0.9.0   | 0.10.0    |
+| peft         | 0.11.1  | 0.11.1    |
-| trl          | 0.8.1   | 0.8.6     |
+| trl          | 0.8.6   | 0.9.4     |
 
 | Optional     | Minimum | Recommend |
 | ------------ | ------- | --------- |
 | CUDA         | 11.6    | 12.2      |
 | deepspeed    | 0.10.0  | 0.14.0    |
 | bitsandbytes | 0.39.0  | 0.43.1    |
-| vllm         | 0.4.0   | 0.4.2     |
+| vllm         | 0.4.3   | 0.4.3     |
-| flash-attn   | 2.3.0   | 2.5.8     |
+| flash-attn   | 2.3.0   | 2.5.9     |
 
 ### Hardware Requirement
 
@@ -319,12 +333,12 @@ huggingface-cli login
 > Installation is mandatory.
 
 ```bash
-git clone https://github.com/hiyouga/LLaMA-Factory.git
+git clone --depth 1 https://github.com/hiyouga/LLaMA-Factory.git
 cd LLaMA-Factory
-pip install -e .[torch,metrics]
+pip install -e ".[torch,metrics]"
 ```
 
-Extra dependencies available: torch, metrics, deepspeed, bitsandbytes, vllm, galore, badam, gptq, awq, aqlm, qwen, modelscope, quality
+Extra dependencies available: torch, torch-npu, metrics, deepspeed, bitsandbytes, hqq, eetq, gptq, awq, aqlm, vllm, galore, badam, qwen, modelscope, quality
 
 > [!TIP]
 > Use `pip install --no-deps -e .` to resolve package conflicts.
@@ -343,24 +357,35 @@ To enable FlashAttention-2 on the Windows platform, you need to install the prec
 
 <details><summary>For Ascend NPU users</summary>
 
-To utilize Ascend NPU devices for (distributed) training and inference, you need to install the **[torch-npu](https://gitee.com/ascend/pytorch)** library and the **[Ascend CANN Kernels](https://www.hiascend.com/developer/download/community/result?module=cann)**.
+To install LLaMA Factory on Ascend NPU devices, please specify extra dependencies: `pip install -e ".[torch-npu,metrics]"`. Additionally, you need to install the **[Ascend CANN Toolkit and Kernels](https://www.hiascend.com/developer/download/community/result?module=cann)**. Please follow the [installation tutorial](https://www.hiascend.com/document/detail/en/CANNCommunityEdition/600alphaX/softwareinstall/instg/atlasdeploy_03_0031.html) or use the following commands:
 
-| Requirement  | Minimum | Recommend |
+```bash
-| ------------ | ------- | --------- |
+# replace the url according to your CANN version and devices
-| CANN         | 8.0.RC1 | 8.0.RC1   |
+# install CANN Toolkit
-| torch        | 2.2.0   | 2.2.0     |
+wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C17SPC701/Ascend-cann-toolkit_8.0.RC1.alpha001_linux-"$(uname -i)".run
-| torch-npu    | 2.2.0   | 2.2.0     |
+bash Ascend-cann-toolkit_8.0.RC1.alpha001_linux-"$(uname -i)".run --install
-| deepspeed    | 0.13.2  | 0.13.2    |
 
-Docker image:
+# install CANN Kernels
+wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C17SPC701/Ascend-cann-kernels-910b_8.0.RC1.alpha001_linux.run
+bash Ascend-cann-kernels-910b_8.0.RC1.alpha001_linux.run --install
 
-- 32GB: [Download page](http://mirrors.cn-central-221.ovaijisuan.com/detail/130.html)
+# set env variables
-- 64GB: Coming soon
+source /usr/local/Ascend/ascend-toolkit/set_env.sh
+```
+
+| Requirement  | Minimum | Recommend   |
+| ------------ | ------- | ----------- |
+| CANN         | 8.0.RC1 | 8.0.RC1     |
+| torch        | 2.1.0   | 2.1.0       |
+| torch-npu    | 2.1.0   | 2.1.0.post3 |
+| deepspeed    | 0.13.2  | 0.13.2      |
 
 Remember to use `ASCEND_RT_VISIBLE_DEVICES` instead of `CUDA_VISIBLE_DEVICES` to specify the device to use.
 
 If you cannot infer model on NPU devices, try setting `do_sample: false` in the configurations.
 
+Download the pre-built Docker images: [32GB](http://mirrors.cn-central-221.ovaijisuan.com/detail/130.html) | [64GB](http://mirrors.cn-central-221.ovaijisuan.com/detail/131.html)
+
 </details>
 
 ### Data Preparation
@@ -375,9 +400,9 @@ Please refer to [data/README.md](data/README.md) for checking the details about
 Use the following 3 commands to run LoRA **fine-tuning**, **inference** and **merging** of the Llama3-8B-Instruct model, respectively.
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_sft.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli chat examples/inference/llama3_lora_sft.yaml
+llamafactory-cli chat examples/inference/llama3_lora_sft.yaml
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
+llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
 ```
 
 See [examples/README.md](examples/README.md) for advanced usage (including distributed training).
@@ -387,52 +412,89 @@ See [examples/README.md](examples/README.md) for advanced usage (including distr
 
 ### Fine-Tuning with LLaMA Board GUI (powered by [Gradio](https://github.com/gradio-app/gradio))
 
-> [!IMPORTANT]
-> LLaMA Board GUI only supports training on a single GPU.
-
-#### Use local environment
-
 ```bash
-CUDA_VISIBLE_DEVICES=0 GRADIO_SHARE=1 llamafactory-cli webui
+llamafactory-cli webui
 ```
 
-<details><summary>For Alibaba Cloud PAI or AutoDL users</summary>
+### Build Docker
 
-If you encountered display problems in LLaMA Board on Alibaba Cloud PAI, try using the following command to set environment variables before starting LLaMA Board:
+For CUDA users:
 
 ```bash
-export GRADIO_SERVER_PORT=7860 GRADIO_ROOT_PATH=/${JUPYTER_NAME}/proxy/7860/
+cd docker/docker-cuda/
+docker-compose up -d
+docker-compose exec llamafactory bash
 ```
 
-If you are using AutoDL, please install a specific version of Gradio:
+For Ascend NPU users:
 
 ```bash
-pip install gradio==4.10.0
+cd docker/docker-npu/
+docker-compose up -d
+docker-compose exec llamafactory bash
+```
+
+<details><summary>Build without Docker Compose</summary>
+
+For CUDA users:
+
+```bash
+docker build -f ./docker/docker-cuda/Dockerfile \
+    --build-arg INSTALL_BNB=false \
+    --build-arg INSTALL_VLLM=false \
+    --build-arg INSTALL_DEEPSPEED=false \
+    --build-arg INSTALL_FLASHATTN=false \
+    --build-arg PIP_INDEX=https://pypi.org/simple \
+    -t llamafactory:latest .
+
+docker run -dit --gpus=all \
+    -v ./hf_cache:/root/.cache/huggingface \
+    -v ./ms_cache:/root/.cache/modelscope \
+    -v ./data:/app/data \
+    -v ./output:/app/output \
+    -p 7860:7860 \
+    -p 8000:8000 \
+    --shm-size 16G \
+    --name llamafactory \
+    llamafactory:latest
+
+docker exec -it llamafactory bash
+```
+
+For Ascend NPU users:
+
+```bash
+# Choose docker image upon your environment
+docker build -f ./docker/docker-npu/Dockerfile \
+    --build-arg INSTALL_DEEPSPEED=false \
+    --build-arg PIP_INDEX=https://pypi.org/simple \
+    -t llamafactory:latest .
+
+# Change `device` upon your resources
+docker run -dit \
+    -v ./hf_cache:/root/.cache/huggingface \
+    -v ./ms_cache:/root/.cache/modelscope \
+    -v ./data:/app/data \
+    -v ./output:/app/output \
+    -v /usr/local/dcmi:/usr/local/dcmi \
+    -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
+    -v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
+    -v /etc/ascend_install.info:/etc/ascend_install.info \
+    -p 7860:7860 \
+    -p 8000:8000 \
+    --device /dev/davinci0 \
+    --device /dev/davinci_manager \
+    --device /dev/devmm_svm \
+    --device /dev/hisi_hdc \
+    --shm-size 16G \
+    --name llamafactory \
+    llamafactory:latest
+
+docker exec -it llamafactory bash
 ```
 
 </details>
 
-#### Use 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
-```
-
-#### Use Docker Compose
-
-```bash
-docker compose -f ./docker-compose.yml up -d
-```
-
 <details><summary>Details about volume</summary>
 
 - hf_cache: Utilize Hugging Face cache on the host machine. Reassignable if a cache already exists in a different directory.
@@ -444,9 +506,12 @@ docker compose -f ./docker-compose.yml up -d
 ### Deploy with OpenAI-style API and vLLM
 
 ```bash
-CUDA_VISIBLE_DEVICES=0,1 API_PORT=8000 llamafactory-cli api examples/inference/llama3_vllm.yaml
+API_PORT=8000 llamafactory-cli api examples/inference/llama3_vllm.yaml
 ```
 
+> [!TIP]
+> Visit https://platform.openai.com/docs/api-reference/chat/create for API document.
+
 ### Download from ModelScope Hub
 
 If you have trouble with downloading models and datasets from Hugging Face, you can use ModelScope.
@@ -455,7 +520,18 @@ If you have trouble with downloading models and datasets from Hugging Face, you
 export USE_MODELSCOPE_HUB=1 # `set USE_MODELSCOPE_HUB=1` for Windows
 ```
 
-Train the model by specifying a model ID of the ModelScope Hub as the `--model_name_or_path`. You can find a full list of model IDs at [ModelScope Hub](https://modelscope.cn/models), e.g., `LLM-Research/Meta-Llama-3-8B-Instruct`.
+Train the model by specifying a model ID of the ModelScope Hub as the `model_name_or_path`. You can find a full list of model IDs at [ModelScope Hub](https://modelscope.cn/models), e.g., `LLM-Research/Meta-Llama-3-8B-Instruct`.
+
+### Use W&B Logger
+
+To use [Weights & Biases](https://wandb.ai) for logging experimental results, you need to add the following arguments to yaml files.
+
+```yaml
+report_to: wandb
+run_name: test_run # optional
+```
+
+Set `WANDB_API_KEY` to [your key](https://wandb.ai/authorize) when launching training tasks to log in with your W&B account.
 
 ## Projects using LLaMA Factory
 
@@ -468,45 +544,73 @@ If you have a project that should be incorporated, please contact via email or c
 1. Wang et al. UbiPhysio: Support Daily Functioning, Fitness, and Rehabilitation with Action Understanding and Feedback in Natural Language. 2023. [[arxiv]](https://arxiv.org/abs/2308.10526)
 1. Luceri et al. Leveraging Large Language Models to Detect Influence Campaigns in Social Media. 2023. [[arxiv]](https://arxiv.org/abs/2311.07816)
 1. Zhang et al. Alleviating Hallucinations of Large Language Models through Induced Hallucinations. 2023. [[arxiv]](https://arxiv.org/abs/2312.15710)
-1. Wang et al. Know Your Needs Better: Towards Structured Understanding of Marketer Demands with Analogical Reasoning Augmented LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2401.04319)
+1. Wang et al. Know Your Needs Better: Towards Structured Understanding of Marketer Demands with Analogical Reasoning Augmented LLMs. KDD 2024. [[arxiv]](https://arxiv.org/abs/2401.04319)
-1. Wang et al. CANDLE: Iterative Conceptualization and Instantiation Distillation from Large Language Models for Commonsense Reasoning. 2024. [[arxiv]](https://arxiv.org/abs/2401.07286)
+1. Wang et al. CANDLE: Iterative Conceptualization and Instantiation Distillation from Large Language Models for Commonsense Reasoning. ACL 2024. [[arxiv]](https://arxiv.org/abs/2401.07286)
 1. Choi et al. FACT-GPT: Fact-Checking Augmentation via Claim Matching with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2402.05904)
 1. Zhang et al. AutoMathText: Autonomous Data Selection with Language Models for Mathematical Texts. 2024. [[arxiv]](https://arxiv.org/abs/2402.07625)
 1. Lyu et al. KnowTuning: Knowledge-aware Fine-tuning for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11176)
 1. Yang et al. LaCo: Large Language Model Pruning via Layer Collaps. 2024. [[arxiv]](https://arxiv.org/abs/2402.11187)
 1. Bhardwaj et al. Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task Arithmetic. 2024. [[arxiv]](https://arxiv.org/abs/2402.11746)
 1. Yang et al. Enhancing Empathetic Response Generation by Augmenting LLMs with Small-scale Empathetic Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11801)
-1. Yi et al. Generation Meets Verification: Accelerating Large Language Model Inference with Smart Parallel Auto-Correct Decoding. 2024. [[arxiv]](https://arxiv.org/abs/2402.11809)
+1. Yi et al. Generation Meets Verification: Accelerating Large Language Model Inference with Smart Parallel Auto-Correct Decoding. ACL 2024 Findings. [[arxiv]](https://arxiv.org/abs/2402.11809)
 1. Cao et al. Head-wise Shareable Attention for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11819)
 1. Zhang et al. Enhancing Multilingual Capabilities of Large Language Models through Self-Distillation from Resource-Rich Languages. 2024. [[arxiv]](https://arxiv.org/abs/2402.12204)
 1. Kim et al. Efficient and Effective Vocabulary Expansion Towards Multilingual Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.14714)
-1. Yu et al. KIEval: A Knowledge-grounded Interactive Evaluation Framework for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.15043)
+1. Yu et al. KIEval: A Knowledge-grounded Interactive Evaluation Framework for Large Language Models. ACL 2024. [[arxiv]](https://arxiv.org/abs/2402.15043)
 1. Huang et al. Key-Point-Driven Data Synthesis with its Enhancement on Mathematical Reasoning. 2024. [[arxiv]](https://arxiv.org/abs/2403.02333)
 1. Duan et al. Negating Negatives: Alignment without Human Positive Samples via Distributional Dispreference Optimization. 2024. [[arxiv]](https://arxiv.org/abs/2403.03419)
 1. Xie and Schwertfeger. Empowering Robotics with Large Language Models: osmAG Map Comprehension with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2403.08228)
 1. Wu et al. Large Language Models are Parallel Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2403.09073)
 1. Zhang et al. EDT: Improving Large Language Models' Generation by Entropy-based Dynamic Temperature Sampling. 2024. [[arxiv]](https://arxiv.org/abs/2403.14541)
 1. Weller et al. FollowIR: Evaluating and Teaching Information Retrieval Models to Follow Instructions. 2024. [[arxiv]](https://arxiv.org/abs/2403.15246)
-1. Hongbin Na. CBT-LLM: A Chinese Large Language Model for Cognitive Behavioral Therapy-based Mental Health Question Answering. 2024. [[arxiv]](https://arxiv.org/abs/2403.16008)
+1. Hongbin Na. CBT-LLM: A Chinese Large Language Model for Cognitive Behavioral Therapy-based Mental Health Question Answering. COLING 2024. [[arxiv]](https://arxiv.org/abs/2403.16008)
 1. Zan et al. CodeS: Natural Language to Code Repository via Multi-Layer Sketch. 2024. [[arxiv]](https://arxiv.org/abs/2403.16443)
 1. Liu et al. Extensive Self-Contrast Enables Feedback-Free Language Model Alignment. 2024. [[arxiv]](https://arxiv.org/abs/2404.00604)
 1. Luo et al. BAdam: A Memory Efficient Full Parameter Training Method for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.02827)
 1. Du et al. Chinese Tiny LLM: Pretraining a Chinese-Centric Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2404.04167)
-1. Ma et al. Parameter Efficient Quasi-Orthogonal Fine-Tuning via Givens Rotation. 2024. [[arxiv]](https://arxiv.org/abs/2404.04316)
+1. Ma et al. Parameter Efficient Quasi-Orthogonal Fine-Tuning via Givens Rotation. ICML 2024. [[arxiv]](https://arxiv.org/abs/2404.04316)
 1. Liu et al. Dynamic Generation of Personalities with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.07084)
 1. Shang et al. How Far Have We Gone in Stripped Binary Code Understanding Using Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.09836)
 1. Huang et al. LLMTune: Accelerate Database Knob Tuning with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.11581)
 1. Deng et al. Text-Tuple-Table: Towards Information Integration in Text-to-Table Generation via Global Tuple Extraction. 2024. [[arxiv]](https://arxiv.org/abs/2404.14215)
 1. Acikgoz et al. Hippocrates: An Open-Source Framework for Advancing Large Language Models in Healthcare. 2024. [[arxiv]](https://arxiv.org/abs/2404.16621)
-1. Zhang et al. Small Language Models Need Strong Verifiers to Self-Correct Reasoning. 2024. [[arxiv]](https://arxiv.org/abs/2404.17140)
+1. Zhang et al. Small Language Models Need Strong Verifiers to Self-Correct Reasoning. ACL 2024 Findings. [[arxiv]](https://arxiv.org/abs/2404.17140)
-1. Zhou et al. FREB-TQA: A Fine-Grained Robustness Evaluation Benchmark for Table Question Answering. 2024. [[arxiv]](https://arxiv.org/abs/2404.18585)
+1. Zhou et al. FREB-TQA: A Fine-Grained Robustness Evaluation Benchmark for Table Question Answering. NAACL 2024. [[arxiv]](https://arxiv.org/abs/2404.18585)
+1. Xu et al. Large Language Models for Cyber Security: A Systematic Literature Review. 2024. [[arxiv]](https://arxiv.org/abs/2405.04760)
+1. Dammu et al. "They are uncultured": Unveiling Covert Harms and Social Threats in LLM Generated Conversations. 2024. [[arxiv]](https://arxiv.org/abs/2405.05378)
+1. Yi et al. A safety realignment framework via subspace-oriented model fusion for large language models. 2024. [[arxiv]](https://arxiv.org/abs/2405.09055)
+1. Lou et al. SPO: Multi-Dimensional Preference Sequential Alignment With Implicit Reward Modeling. 2024. [[arxiv]](https://arxiv.org/abs/2405.12739)
+1. Zhang et al. Getting More from Less: Large Language Models are Good Spontaneous Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2405.13816)
+1. Zhang et al. TS-Align: A Teacher-Student Collaborative Framework for Scalable Iterative Finetuning of Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2405.20215)
+1. Zihong Chen. Sentence Segmentation and Sentence Punctuation Based on XunziALLM. 2024. [[paper]](https://aclanthology.org/2024.lt4hala-1.30)
+1. Gao et al. The Best of Both Worlds: Toward an Honest and Helpful Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2406.00380)
+1. Wang and Song. MARS: Benchmarking the Metaphysical Reasoning Abilities of Language Models with a Multi-task Evaluation Dataset. 2024. [[arxiv]](https://arxiv.org/abs/2406.02106)
+1. Hu et al. Computational Limits of Low-Rank Adaptation (LoRA) for Transformer-Based Models. 2024. [[arxiv]](https://arxiv.org/abs/2406.03136)
+1. Ge et al. Time Sensitive Knowledge Editing through Efficient Finetuning. ACL 2024. [[arxiv]](https://arxiv.org/abs/2406.04496)
+1. Tan et al. Peer Review as A Multi-Turn and Long-Context Dialogue with Role-Based Interactions. 2024. [[arxiv]](https://arxiv.org/abs/2406.05688)
+1. Song et al. Turbo Sparse: Achieving LLM SOTA Performance with Minimal Activated Parameters. 2024. [[arxiv]](https://arxiv.org/abs/2406.05955)
+1. Gu et al. RWKV-CLIP: A Robust Vision-Language Representation Learner. 2024. [[arxiv]](https://arxiv.org/abs/2406.06973)
+1. Chen et al. Advancing Tool-Augmented Large Language Models: Integrating Insights from Errors in Inference Trees. 2024. [[arxiv]](https://arxiv.org/abs/2406.07115)
+1. Zhu et al. Are Large Language Models Good Statisticians?. 2024. [[arxiv]](https://arxiv.org/abs/2406.07815)
+1. Li et al. Know the Unknown: An Uncertainty-Sensitive Method for LLM Instruction Tuning. 2024. [[arxiv]](https://arxiv.org/abs/2406.10099)
+1. Ding et al. IntentionQA: A Benchmark for Evaluating Purchase Intention Comprehension Abilities of Language Models in E-commerce. 2024. [[arxiv]](https://arxiv.org/abs/2406.10173)
+1. He et al. COMMUNITY-CROSS-INSTRUCT: Unsupervised Instruction Generation for Aligning Large Language Models to Online Communities. 2024. [[arxiv]](https://arxiv.org/abs/2406.12074)
+1. Lin et al. FVEL: Interactive Formal Verification Environment with Large Language Models via Theorem Proving. 2024. [[arxiv]](https://arxiv.org/abs/2406.14408)
+1. Treutlein et al. Connecting the Dots: LLMs can Infer and Verbalize Latent Structure from Disparate Training Data. 2024. [[arxiv]](https://arxiv.org/abs/2406.14546)
+1. Feng et al. SS-Bench: A Benchmark for Social Story Generation and Evaluation. 2024. [[arxiv]](https://arxiv.org/abs/2406.15695)
+1. Feng et al. Self-Constructed Context Decompilation with Fined-grained Alignment Enhancement. 2024. [[arxiv]](https://arxiv.org/abs/2406.17233)
+1. Liu et al. Large Language Models for Cuffless Blood Pressure Measurement From Wearable Biosignals. 2024. [[arxiv]](https://arxiv.org/abs/2406.18069)
+1. Iyer et al. Exploring Very Low-Resource Translation with LLMs: The University of Edinburgh’s Submission to AmericasNLP 2024 Translation Task. AmericasNLP 2024. [[paper]](https://aclanthology.org/2024.americasnlp-1.25)
 1. **[StarWhisper](https://github.com/Yu-Yang-Li/StarWhisper)**: A large language model for Astronomy, based on ChatGLM2-6B and Qwen-14B.
 1. **[DISC-LawLLM](https://github.com/FudanDISC/DISC-LawLLM)**: A large language model specialized in Chinese legal domain, based on Baichuan-13B, is capable of retrieving and reasoning on legal knowledge.
-1. **[Sunsimiao](https://github.com/thomas-yanxin/Sunsimiao)**: A large language model specialized in Chinese medical domain, based on Baichuan-7B and ChatGLM-6B.
+1. **[Sunsimiao](https://github.com/X-D-Lab/Sunsimiao)**: A large language model specialized in Chinese medical domain, based on Baichuan-7B and ChatGLM-6B.
 1. **[CareGPT](https://github.com/WangRongsheng/CareGPT)**: A series of large language models for Chinese medical domain, based on LLaMA2-7B and Baichuan-13B.
 1. **[MachineMindset](https://github.com/PKU-YuanGroup/Machine-Mindset/)**: A series of MBTI Personality large language models, capable of giving any LLM 16 different personality types based on different datasets and training methods.
 1. **[Luminia-13B-v3](https://huggingface.co/Nekochu/Luminia-13B-v3)**: A large language model specialized in generate metadata for stable diffusion. [[🤗Demo]](https://huggingface.co/spaces/Nekochu/Luminia-13B_SD_Prompt)
 1. **[Chinese-LLaVA-Med](https://github.com/BUAADreamer/Chinese-LLaVA-Med)**: A multimodal large language model specialized in Chinese medical domain, based on LLaVA-1.5-7B.
+1. **[AutoRE](https://github.com/THUDM/AutoRE)**: A document-level relation extraction system based on large language models.
+1. **[NVIDIA RTX AI Toolkit](https://github.com/NVIDIA/RTX-AI-Toolkit)**: SDKs for fine-tuning LLMs on Windows PC for NVIDIA RTX.
+1. **[LazyLLM](https://github.com/LazyAGI/LazyLLM)**: An easy and lazy way for building multi-agent LLMs applications and supports model fine-tuning via LLaMA Factory.
 
 </details>
 
@@ -514,17 +618,19 @@ If you have a project that should be incorporated, please contact via email or c
 
 This repository is licensed under the [Apache-2.0 License](LICENSE).
 
-Please follow the model licenses to use the corresponding model weights: [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 (LLaVA-1.5)](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) / [Phi-3](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/blob/main/LICENSE) / [Qwen](https://github.com/QwenLM/Qwen/blob/main/Tongyi%20Qianwen%20LICENSE%20AGREEMENT) / [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) / [Yi-1.5](LICENSE) / [Yuan](https://github.com/IEIT-Yuan/Yuan-2.0/blob/main/LICENSE-Yuan)
+Please follow the model licenses to use the corresponding model weights: [Baichuan 2](https://huggingface.co/baichuan-inc/Baichuan2-7B-Base/blob/main/Community%20License%20for%20Baichuan%202%20Model.pdf) / [BLOOM](https://huggingface.co/spaces/bigscience/license) / [ChatGLM3](https://github.com/THUDM/ChatGLM3/blob/main/MODEL_LICENSE) / [Command R](https://cohere.com/c4ai-cc-by-nc-license) / [DeepSeek](https://github.com/deepseek-ai/DeepSeek-LLM/blob/main/LICENSE-MODEL) / [Falcon](https://huggingface.co/tiiuae/falcon-180B/blob/main/LICENSE.txt) / [Gemma](https://ai.google.dev/gemma/terms) / [GLM-4](https://huggingface.co/THUDM/glm-4-9b/blob/main/LICENSE) / [InternLM2](https://github.com/InternLM/InternLM#license) / [Llama](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md) / [Llama 2 (LLaVA-1.5)](https://ai.meta.com/llama/license/) / [Llama 3](https://llama.meta.com/llama3/license/) / [Mistral](LICENSE) / [OLMo](LICENSE) / [Phi-1.5/Phi-2](https://huggingface.co/microsoft/phi-1_5/resolve/main/Research%20License.docx) / [Phi-3](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/blob/main/LICENSE) / [Qwen](https://github.com/QwenLM/Qwen/blob/main/Tongyi%20Qianwen%20LICENSE%20AGREEMENT) / [StarCoder 2](https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement) / [XVERSE](https://github.com/xverse-ai/XVERSE-13B/blob/main/MODEL_LICENSE.pdf) / [Yi](https://huggingface.co/01-ai/Yi-6B/blob/main/LICENSE) / [Yi-1.5](LICENSE) / [Yuan 2](https://github.com/IEIT-Yuan/Yuan-2.0/blob/main/LICENSE-Yuan)
 
 ## Citation
 
 If this work is helpful, please kindly cite as:
 
 ```bibtex
-@article{zheng2024llamafactory,
+@inproceedings{zheng2024llamafactory,
   title={LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models},
-  author={Yaowei Zheng and Richong Zhang and Junhao Zhang and Yanhan Ye and Zheyan Luo and Yongqiang Ma},
+  author={Yaowei Zheng and Richong Zhang and Junhao Zhang and Yanhan Ye and Zheyan Luo and Zhangchi Feng and Yongqiang Ma},
-  journal={arXiv preprint arXiv:2403.13372},
+  booktitle={Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
+  address={Bangkok, Thailand},
+  publisher={Association for Computational Linguistics},
   year={2024},
   url={http://arxiv.org/abs/2403.13372}
 }

README_zh.md

@@ -3,19 +3,19 @@
 [![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/)
+[![PyPI](https://img.shields.io/pypi/v/llamafactory)](https://pypi.org/project/llamafactory/)
-[![Downloads](https://static.pepy.tech/badge/llmtuner)](https://pypi.org/project/llmtuner/)
+[![Citation](https://img.shields.io/badge/citation-72-green)](#使用了-llama-factory-的项目)
-[![Citation](https://img.shields.io/badge/citation-44-green)](#使用了-llama-factory-的项目)
 [![GitHub pull request](https://img.shields.io/badge/PRs-welcome-blue)](https://github.com/hiyouga/LLaMA-Factory/pulls)
 [![Discord](https://dcbadge.vercel.app/api/server/rKfvV9r9FK?compact=true&style=flat)](https://discord.gg/rKfvV9r9FK)
 [![Twitter](https://img.shields.io/twitter/follow/llamafactory_ai)](https://twitter.com/llamafactory_ai)
+[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1d5KQtbemerlSDSxZIfAaWXhKr30QypiK?usp=sharing)
+[![Open in DSW](https://gallery.pai-ml.com/assets/open-in-dsw.svg)](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory)
 [![Spaces](https://img.shields.io/badge/🤗-Open%20in%20Spaces-blue)](https://huggingface.co/spaces/hiyouga/LLaMA-Board)
 [![Studios](https://img.shields.io/badge/ModelScope-Open%20in%20Studios-blue)](https://modelscope.cn/studios/hiyouga/LLaMA-Board)
-[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1d5KQtbemerlSDSxZIfAaWXhKr30QypiK?usp=sharing)
 
 [![GitHub Tread](https://trendshift.io/api/badge/repositories/4535)](https://trendshift.io/repositories/4535)
 
-👋 加入我们的[微信群](assets/wechat.jpg)。
+👋 加入我们的[微信群](assets/wechat.jpg)或 [NPU 用户群](assets/wechat_npu.jpg)。
 
 \[ [English](README.md) | 中文 \]
 
@@ -26,6 +26,7 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 选择你的打开方式：
 
 - **Colab**：https://colab.research.google.com/drive/1d5KQtbemerlSDSxZIfAaWXhKr30QypiK?usp=sharing
+- **PAI-DSW**: https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory
 - **本地机器**：请见[如何使用](#如何使用)
 
 ## 目录
@@ -46,9 +47,9 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 ## 项目特色
 
 - **多种模型**：LLaMA、LLaVA、Mistral、Mixtral-MoE、Qwen、Yi、Gemma、Baichuan、ChatGLM、Phi 等等。
-- **集成方法**：（增量）预训练、（多模态）指令监督微调、奖励模型训练、PPO 训练、DPO 训练和 ORPO 训练。
+- **集成方法**：（增量）预训练、（多模态）指令监督微调、奖励模型训练、PPO 训练、DPO 训练、KTO 训练、ORPO 训练等等。
-- **多种精度**：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。
+- **多种精度**：16 比特全参数微调、冻结微调、LoRA 微调和基于 AQLM/AWQ/GPTQ/LLM.int8/HQQ/EETQ 的 2/3/4/5/6/8 比特 QLoRA 微调。
-- **先进算法**：GaLore、BAdam、DoRA、LongLoRA、LLaMA Pro、Mixture-of-Depths、LoRA+、LoftQ 和 Agent 微调。
+- **先进算法**：GaLore、BAdam、DoRA、LongLoRA、LLaMA Pro、Mixture-of-Depths、LoRA+、LoftQ、PiSSA 和 Agent 微调。
 - **实用技巧**：FlashAttention-2、Unsloth、RoPE scaling、NEFTune 和 rsLoRA。
 - **实验监控**：LlamaBoard、TensorBoard、Wandb、MLflow 等等。
 - **极速推理**：基于 vLLM 的 OpenAI 风格 API、浏览器界面和命令行接口。
@@ -70,14 +71,22 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 
 ## 更新日志
 
-[24/05/14] 我们支持了昇腾 NPU 设备的训练和推理。详情请查阅[安装](#安装-llama-factory)部分。
+[24/06/16] 我们支持了 **[PiSSA](https://arxiv.org/abs/2404.02948)** 算法。详细用法请参照 [examples](examples/README_zh.md)。
 
-[24/05/13] 我们支持了 Yi-1.5 系列模型的微调。
+[24/06/07] 我们支持了 **[Qwen2](https://qwenlm.github.io/blog/qwen2/)** 和 **[GLM-4](https://github.com/THUDM/GLM-4)** 模型的微调。
 
-[24/04/26] 我们支持了多模态模型 **LLaVA-1.5** 的微调。详细用法请参照 [examples](examples/README_zh.md)。
+[24/05/26] 我们支持了 **[SimPO](https://arxiv.org/abs/2405.14734)** 偏好对齐算法。详细用法请参照 [examples](examples/README_zh.md)。
 
 <details><summary>展开日志</summary>
 
+[24/05/20] 我们支持了 **PaliGemma** 系列模型的微调。注意 PaliGemma 是预训练模型，你需要使用 `gemma` 模板进行微调使其获得对话能力。
+
+[24/05/18] 我们支持了 **[KTO](https://arxiv.org/abs/2402.01306)** 偏好对齐算法。详细用法请参照 [examples](examples/README_zh.md)。
+
+[24/05/14] 我们支持了昇腾 NPU 设备的训练和推理。详情请查阅[安装](#安装-llama-factory)部分。
+
+[24/04/26] 我们支持了多模态模型 **LLaVA-1.5** 的微调。详细用法请参照 [examples](examples/README_zh.md)。
+
 [24/04/22] 我们提供了在免费 T4 GPU 上微调 Llama-3 模型的 **[Colab 笔记本](https://colab.research.google.com/drive/1d5KQtbemerlSDSxZIfAaWXhKr30QypiK?usp=sharing)**。Hugging Face 社区公开了两个利用 LLaMA Factory 微调的 Llama-3 模型，详情请见 [Llama3-8B-Chinese-Chat](https://huggingface.co/shenzhi-wang/Llama3-8B-Chinese-Chat) 和 [Llama3-Chinese](https://huggingface.co/zhichen/Llama3-Chinese)。
 
 [24/04/21] 我们基于 [AstraMindAI 的仓库](https://github.com/astramind-ai/Mixture-of-depths)支持了 **[混合深度训练](https://arxiv.org/abs/2404.02258)**。详细用法请参照 [examples](examples/README_zh.md)。
@@ -104,7 +113,7 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 
 [24/02/05] Qwen1.5（Qwen2 测试版）系列模型已在 LLaMA-Factory 中实现微调支持。详情请查阅该[博客页面](https://qwenlm.github.io/zh/blog/qwen1.5/)。
 
-[24/01/18] 我们针对绝大多数模型实现了 **Agent 微调**，微调时指定 `dataset: glaive_toolcall` 即可使模型获得工具调用能力。
+[24/01/18] 我们针对绝大多数模型实现了 **Agent 微调**，微调时指定 `dataset: glaive_toolcall_zh` 即可使模型获得工具调用能力。
 
 [23/12/23] 我们针对 LLaMA, Mistral 和 Yi 模型支持了 **[unsloth](https://github.com/unslothai/unsloth)** 的 LoRA 训练加速。请使用 `use_unsloth: true` 参数启用 unsloth 优化。该方法可提供 **170%** 的训练速度，详情请查阅[此页面](https://github.com/hiyouga/LLaMA-Factory/wiki/Performance-comparison)。
 
@@ -142,43 +151,41 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 
 ## 模型
 
-| 模型名                                                   | 模型大小                          | 默认模块           | Template  |
+| 模型名                                                       | 模型大小                          | Template  |
-| -------------------------------------------------------- | -------------------------------- | ----------------- | --------- |
+| ------------------------------------------------------------ | -------------------------------- | --------- |
-| [Baichuan2](https://huggingface.co/baichuan-inc)         | 7B/13B                           | W_pack            | baichuan2 |
+| [Baichuan 2](https://huggingface.co/baichuan-inc)            | 7B/13B                           | baichuan2 |
-| [BLOOM](https://huggingface.co/bigscience)               | 560M/1.1B/1.7B/3B/7.1B/176B      | query_key_value   | -         |
+| [BLOOM/BLOOMZ](https://huggingface.co/bigscience)            | 560M/1.1B/1.7B/3B/7.1B/176B      | -         |
-| [BLOOMZ](https://huggingface.co/bigscience)              | 560M/1.1B/1.7B/3B/7.1B/176B      | query_key_value   | -         |
+| [ChatGLM3](https://huggingface.co/THUDM)                     | 6B                               | chatglm3  |
-| [ChatGLM3](https://huggingface.co/THUDM)                 | 6B                               | query_key_value   | chatglm3  |
+| [Command R](https://huggingface.co/CohereForAI)              | 35B/104B                         | cohere    |
-| [Command-R](https://huggingface.co/CohereForAI)          | 35B/104B                         | q_proj,v_proj     | cohere    |
+| [DeepSeek (Code/MoE)](https://huggingface.co/deepseek-ai)    | 7B/16B/67B/236B                  | deepseek  |
-| [DeepSeek (MoE)](https://huggingface.co/deepseek-ai)     | 7B/16B/67B/236B                  | q_proj,v_proj     | deepseek  |
+| [Falcon](https://huggingface.co/tiiuae)                      | 7B/11B/40B/180B                  | falcon    |
-| [Falcon](https://huggingface.co/tiiuae)                  | 7B/40B/180B                      | query_key_value   | falcon    |
+| [Gemma/Gemma 2/CodeGemma](https://huggingface.co/google)     | 2B/7B/9B/27B                     | gemma     |
-| [Gemma/CodeGemma](https://huggingface.co/google)         | 2B/7B                            | q_proj,v_proj     | gemma     |
+| [GLM-4](https://huggingface.co/THUDM)                        | 9B                               | glm4      |
-| [InternLM2](https://huggingface.co/internlm)             | 7B/20B                           | wqkv              | intern2   |
+| [InternLM2](https://huggingface.co/internlm)                 | 7B/20B                           | intern2   |
-| [LLaMA](https://github.com/facebookresearch/llama)       | 7B/13B/33B/65B                   | q_proj,v_proj     | -         |
+| [Llama](https://github.com/facebookresearch/llama)           | 7B/13B/33B/65B                   | -         |
-| [LLaMA-2](https://huggingface.co/meta-llama)             | 7B/13B/70B                       | q_proj,v_proj     | llama2    |
+| [Llama 2](https://huggingface.co/meta-llama)                 | 7B/13B/70B                       | llama2    |
-| [LLaMA-3](https://huggingface.co/meta-llama)             | 8B/70B                           | q_proj,v_proj     | llama3    |
+| [Llama 3](https://huggingface.co/meta-llama)                 | 8B/70B                           | llama3    |
-| [LLaVA-1.5](https://huggingface.co/llava-hf)             | 7B/13B                           | q_proj,v_proj     | vicuna    |
+| [LLaVA-1.5](https://huggingface.co/llava-hf)                 | 7B/13B                           | vicuna    |
-| [Mistral/Mixtral](https://huggingface.co/mistralai)      | 7B/8x7B/8x22B                    | q_proj,v_proj     | mistral   |
+| [Mistral/Mixtral](https://huggingface.co/mistralai)          | 7B/8x7B/8x22B                    | mistral   |
-| [OLMo](https://huggingface.co/allenai)                   | 1B/7B                            | q_proj,v_proj     | -         |
+| [OLMo](https://huggingface.co/allenai)                       | 1B/7B                            | -         |
-| [Phi-1.5/2](https://huggingface.co/microsoft)            | 1.3B/2.7B                        | q_proj,v_proj     | -         |
+| [PaliGemma](https://huggingface.co/google)                   | 3B                               | gemma     |
-| [Phi-3](https://huggingface.co/microsoft)                | 3.8B                             | qkv_proj          | phi       |
+| [Phi-1.5/Phi-2](https://huggingface.co/microsoft)            | 1.3B/2.7B                        | -         |
-| [Qwen](https://huggingface.co/Qwen)                      | 1.8B/7B/14B/72B                  | c_attn            | qwen      |
+| [Phi-3](https://huggingface.co/microsoft)                    | 4B/7B/14B                        | phi       |
-| [Qwen1.5 (Code/MoE)](https://huggingface.co/Qwen)        | 0.5B/1.8B/4B/7B/14B/32B/72B/110B | q_proj,v_proj     | qwen      |
+| [Qwen/Qwen1.5/Qwen2 (Code/MoE)](https://huggingface.co/Qwen) | 0.5B/1.5B/4B/7B/14B/32B/72B/110B | qwen      |
-| [StarCoder2](https://huggingface.co/bigcode)             | 3B/7B/15B                        | q_proj,v_proj     | -         |
+| [StarCoder 2](https://huggingface.co/bigcode)                | 3B/7B/15B                        | -         |
-| [XVERSE](https://huggingface.co/xverse)                  | 7B/13B/65B                       | q_proj,v_proj     | xverse    |
+| [XVERSE](https://huggingface.co/xverse)                      | 7B/13B/65B                       | xverse    |
-| [Yi (1/1.5)](https://huggingface.co/01-ai)               | 6B/9B/34B                        | q_proj,v_proj     | yi        |
+| [Yi/Yi-1.5](https://huggingface.co/01-ai)                    | 6B/9B/34B                        | yi        |
-| [Yi-VL](https://huggingface.co/01-ai)                    | 6B/34B                           | q_proj,v_proj     | yi_vl     |
+| [Yi-VL](https://huggingface.co/01-ai)                        | 6B/34B                           | yi_vl     |
-| [Yuan](https://huggingface.co/IEITYuan)                  | 2B/51B/102B                      | q_proj,v_proj     | yuan      |
+| [Yuan 2](https://huggingface.co/IEITYuan)                    | 2B/51B/102B                      | yuan      |
 
 > [!NOTE]
-> **默认模块**应作为 `--lora_target` 参数的默认值，可使用 `--lora_target all` 参数指定全部模块以取得更好的效果。
+> 对于所有“基座”（Base）模型，`template` 参数可以是 `default`, `alpaca`, `vicuna` 等任意值。但“对话”（Instruct/Chat）模型请务必使用**对应的模板**。
 >
-> 对于所有“基座”（Base）模型，`--template` 参数可以是 `default`, `alpaca`, `vicuna` 等任意值。但“对话”（Instruct/Chat）模型请务必使用**对应的模板**。
+> 请务必在训练和推理时采用**完全一致**的模板。
->
-> 请务必在训练和推理时使用**完全一致**的模板。
 
-项目所支持模型的完整列表请参阅 [constants.py](src/llmtuner/extras/constants.py)。
+项目所支持模型的完整列表请参阅 [constants.py](src/llamafactory/extras/constants.py)。
 
-您也可以在 [template.py](src/llmtuner/data/template.py) 中添加自己的对话模板。
+您也可以在 [template.py](src/llamafactory/data/template.py) 中添加自己的对话模板。
 
 ## 训练方法
 
@@ -189,7 +196,9 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 | 奖励模型训练            | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
 | PPO 训练               | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
 | DPO 训练               | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+| KTO 训练               | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
 | ORPO 训练              | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+| SimPO 训练             | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
 
 ## 数据集
 
@@ -202,6 +211,8 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 - [Wikipedia (zh)](https://huggingface.co/datasets/pleisto/wikipedia-cn-20230720-filtered)
 - [Pile (en)](https://huggingface.co/datasets/EleutherAI/pile)
 - [SkyPile (zh)](https://huggingface.co/datasets/Skywork/SkyPile-150B)
+- [FineWeb (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb)
+- [FineWeb-Edu (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu)
 - [The Stack (en)](https://huggingface.co/datasets/bigcode/the-stack)
 - [StarCoder (en)](https://huggingface.co/datasets/bigcode/starcoderdata)
 
@@ -209,12 +220,12 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 
 <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)
-- [Alpaca GPT4 (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
 - [Identity (en&zh)](data/identity.json)
-- [Open Assistant (zh)](https://huggingface.co/datasets/OpenAssistant/oasst1)
+- [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)
-- [ShareGPT (zh)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT/tree/main/Chinese-instruction-collection)
+- [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca-3)
+- [Alpaca GPT4 (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
+- [Glaive Function Calling V2 (en&zh)](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2)
+- [LIMA (en)](https://huggingface.co/datasets/GAIR/lima)
 - [Guanaco Dataset (multilingual)](https://huggingface.co/datasets/JosephusCheung/GuanacoDataset)
 - [BELLE 2M (zh)](https://huggingface.co/datasets/BelleGroup/train_2M_CN)
 - [BELLE 1M (zh)](https://huggingface.co/datasets/BelleGroup/train_1M_CN)
@@ -223,7 +234,6 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 - [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)
@@ -236,15 +246,19 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 - [WebNovel (zh)](https://huggingface.co/datasets/zxbsmk/webnovel_cn)
 - [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
 - [deepctrl (en&zh)](https://www.modelscope.cn/datasets/deepctrl/deepctrl-sft-data)
-- [Ad Gen (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)
+- [Advertise Generating (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)
 - [ShareGPT Hyperfiltered (en)](https://huggingface.co/datasets/totally-not-an-llm/sharegpt-hyperfiltered-3k)
 - [ShareGPT4 (en&zh)](https://huggingface.co/datasets/shibing624/sharegpt_gpt4)
 - [UltraChat 200k (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k)
 - [AgentInstruct (en)](https://huggingface.co/datasets/THUDM/AgentInstruct)
 - [LMSYS Chat 1M (en)](https://huggingface.co/datasets/lmsys/lmsys-chat-1m)
 - [Evol Instruct V2 (en)](https://huggingface.co/datasets/WizardLM/WizardLM_evol_instruct_V2_196k)
-- [Glaive Function Calling V2 (en)](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2)
 - [Cosmopedia (en)](https://huggingface.co/datasets/HuggingFaceTB/cosmopedia)
+- [STEM (zh)](https://huggingface.co/datasets/hfl/stem_zh_instruction)
+- [Ruozhiba (zh)](https://huggingface.co/datasets/hfl/ruozhiba_gpt4_turbo)
+- [Neo-sft (zh)](https://huggingface.co/datasets/m-a-p/neo_sft_phase2)
+- [WebInstructSub (en)](https://huggingface.co/datasets/TIGER-Lab/WebInstructSub)
+- [Magpie-Pro-300K-Filtered (en)](https://huggingface.co/datasets/Magpie-Align/Magpie-Pro-300K-Filtered)
 - [LLaVA mixed (en&zh)](https://huggingface.co/datasets/BUAADreamer/llava-en-zh-300k)
 - [Open Assistant (de)](https://huggingface.co/datasets/mayflowergmbh/oasst_de)
 - [Dolly 15k (de)](https://huggingface.co/datasets/mayflowergmbh/dolly-15k_de)
@@ -260,13 +274,13 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/ec36a9dd-37f4-4f72-81bd
 
 <details><summary>偏好数据集</summary>
 
-- [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
-- [GPT-4 Generated Data (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
-- [Orca DPO (en)](https://huggingface.co/datasets/Intel/orca_dpo_pairs)
-- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
 - [DPO mixed (en&zh)](https://huggingface.co/datasets/hiyouga/DPO-En-Zh-20k)
-- [Open Assistant (zh)](https://huggingface.co/datasets/OpenAssistant/oasst1)
+- [UltraFeedback (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized)
+- [Orca DPO Pairs (en)](https://huggingface.co/datasets/Intel/orca_dpo_pairs)
+- [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
+- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
 - [Orca DPO (de)](https://huggingface.co/datasets/mayflowergmbh/intel_orca_dpo_pairs_de)
+- [KTO mixed (en)](https://huggingface.co/datasets/argilla/kto-mix-15k)
 
 </details>
 
@@ -281,21 +295,21 @@ huggingface-cli login
 
 | 必需项       | 至少     | 推荐      |
 | ------------ | ------- | --------- |
-| python       | 3.8     | 3.10      |
+| python       | 3.8     | 3.11      |
-| torch        | 1.13.1  | 2.2.0     |
+| torch        | 1.13.1  | 2.3.0     |
-| transformers | 4.37.2  | 4.40.1    |
+| transformers | 4.41.2  | 4.41.2    |
-| datasets     | 2.14.3  | 2.19.1    |
+| datasets     | 2.16.0  | 2.19.2    |
-| accelerate   | 0.27.2  | 0.30.0    |
+| accelerate   | 0.30.1  | 0.30.1    |
-| peft         | 0.9.0   | 0.10.0    |
+| peft         | 0.11.1  | 0.11.1    |
-| trl          | 0.8.1   | 0.8.6     |
+| trl          | 0.8.6   | 0.9.4     |
 
 | 可选项       | 至少     | 推荐      |
 | ------------ | ------- | --------- |
 | CUDA         | 11.6    | 12.2      |
 | deepspeed    | 0.10.0  | 0.14.0    |
 | bitsandbytes | 0.39.0  | 0.43.1    |
-| vllm         | 0.4.0   | 0.4.2     |
+| vllm         | 0.4.3   | 0.4.3     |
-| flash-attn   | 2.3.0   | 2.5.8     |
+| flash-attn   | 2.3.0   | 2.5.9     |
 
 ### 硬件依赖
 
@@ -319,12 +333,12 @@ huggingface-cli login
 > 此步骤为必需。
 
 ```bash
-git clone https://github.com/hiyouga/LLaMA-Factory.git
+git clone --depth 1 https://github.com/hiyouga/LLaMA-Factory.git
 cd LLaMA-Factory
-pip install -e .[torch,metrics]
+pip install -e ".[torch,metrics]"
 ```
 
-可选的额外依赖项：torch、metrics、deepspeed、bitsandbytes、vllm、galore、badam、gptq、awq、aqlm、qwen、modelscope、quality
+可选的额外依赖项：torch、torch-npu、metrics、deepspeed、bitsandbytes、hqq、eetq、gptq、awq、aqlm、vllm、galore、badam、qwen、modelscope、quality
 
 > [!TIP]
 > 遇到包冲突时，可使用 `pip install --no-deps -e .` 解决。
@@ -343,24 +357,35 @@ pip install https://github.com/jllllll/bitsandbytes-windows-webui/releases/downl
 
 <details><summary>昇腾 NPU 用户指南</summary>
 
-如果使用昇腾 NPU 设备进行（分布式）训练或推理，需要安装 **[torch-npu](https://gitee.com/ascend/pytorch)** 库和 **[Ascend CANN Kernels](https://www.hiascend.com/developer/download/community/result?module=cann)**。
+在昇腾 NPU 设备上安装 LLaMA Factory 时，需要指定额外依赖项，使用 `pip install -e ".[torch-npu,metrics]"` 命令安装。此外，还需要安装 **[Ascend CANN Toolkit 与 Kernels](https://www.hiascend.com/developer/download/community/result?module=cann)**，安装方法请参考[安装教程](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC2alpha002/quickstart/quickstart/quickstart_18_0004.html)或使用以下命令：
 
-| 依赖项       | 至少     | 推荐      |
+```bash
-| ------------ | ------- | --------- |
+# 请替换 URL 为 CANN 版本和设备型号对应的 URL
-| CANN         | 8.0.RC1 | 8.0.RC1   |
+# 安装 CANN Toolkit
-| torch        | 2.2.0   | 2.2.0     |
+wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C17SPC701/Ascend-cann-toolkit_8.0.RC1.alpha001_linux-"$(uname -i)".run
-| torch-npu    | 2.2.0   | 2.2.0     |
+bash Ascend-cann-toolkit_8.0.RC1.alpha001_linux-"$(uname -i)".run --install
-| deepspeed    | 0.13.2  | 0.13.2    |
 
-Docker 镜像：
+# 安装 CANN Kernels
+wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C17SPC701/Ascend-cann-kernels-910b_8.0.RC1.alpha001_linux.run
+bash Ascend-cann-kernels-910b_8.0.RC1.alpha001_linux.run --install
 
-- 32GB：[下载地址](http://mirrors.cn-central-221.ovaijisuan.com/detail/130.html)
+# 设置环境变量
-- 64GB：敬请期待
+source /usr/local/Ascend/ascend-toolkit/set_env.sh
+```
 
-请记得使用 `ASCEND_RT_VISIBLE_DEVICES` 而非 `CUDA_VISIBLE_DEVICES` 来指定您使用的设备。
+| 依赖项       | 至少     | 推荐        |
+| ------------ | ------- | ----------- |
+| CANN         | 8.0.RC1 | 8.0.RC1     |
+| torch        | 2.1.0   | 2.1.0       |
+| torch-npu    | 2.1.0   | 2.1.0.post3 |
+| deepspeed    | 0.13.2  | 0.13.2      |
+
+请使用 `ASCEND_RT_VISIBLE_DEVICES` 而非 `CUDA_VISIBLE_DEVICES` 来指定运算设备。
 
 如果遇到无法正常推理的情况，请尝试设置 `do_sample: false`。
 
+下载预构建 Docker 镜像：[32GB](http://mirrors.cn-central-221.ovaijisuan.com/detail/130.html) | [64GB](http://mirrors.cn-central-221.ovaijisuan.com/detail/131.html)
+
 </details>
 
 ### 数据准备
@@ -375,9 +400,9 @@ Docker 镜像：
 下面三行命令分别对 Llama3-8B-Instruct 模型进行 LoRA **微调**、**推理**和**合并**。
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_sft.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli chat examples/inference/llama3_lora_sft.yaml
+llamafactory-cli chat examples/inference/llama3_lora_sft.yaml
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
+llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
 ```
 
 高级用法请参考 [examples/README_zh.md](examples/README_zh.md)（包括多 GPU 微调）。
@@ -387,52 +412,89 @@ CUDA_VISIBLE_DEVICES=0 llamafactory-cli export examples/merge_lora/llama3_lora_s
 
 ### LLaMA Board 可视化微调（由 [Gradio](https://github.com/gradio-app/gradio) 驱动）
 
-> [!IMPORTANT]
-> LLaMA Board 可视化界面目前仅支持单 GPU 训练。
-
-#### 使用本地环境
-
 ```bash
-CUDA_VISIBLE_DEVICES=0 GRADIO_SHARE=1 llamafactory-cli webui
+llamafactory-cli webui
 ```
 
-<details><summary>阿里云 PAI 和 AutoDL 用户指南</summary>
+### 构建 Docker
 
-如果您在阿里云 PAI 上使用 LLaMA Board 时遇到显示问题，请尝试在启动前使用以下命令设置环境变量：
+CUDA 用户：
 
 ```bash
-export GRADIO_SERVER_PORT=7860 GRADIO_ROOT_PATH=/${JUPYTER_NAME}/proxy/7860/
+cd docker/docker-cuda/
+docker-compose up -d
+docker-compose exec llamafactory bash
 ```
 
-如果您正在使用 AutoDL，请安装下述 Gradio 版本：
+昇腾 NPU 用户：
 
 ```bash
-pip install gradio==4.10.0
+cd docker/docker-npu/
+docker-compose up -d
+docker-compose exec llamafactory bash
+```
+
+<details><summary>不使用 Docker Compose 构建</summary>
+
+CUDA 用户：
+
+```bash
+docker build -f ./docker/docker-cuda/Dockerfile \
+    --build-arg INSTALL_BNB=false \
+    --build-arg INSTALL_VLLM=false \
+    --build-arg INSTALL_DEEPSPEED=false \
+    --build-arg INSTALL_FLASHATTN=false \
+    --build-arg PIP_INDEX=https://pypi.org/simple \
+    -t llamafactory:latest .
+
+docker run -dit --gpus=all \
+    -v ./hf_cache:/root/.cache/huggingface \
+    -v ./ms_cache:/root/.cache/modelscope \
+    -v ./data:/app/data \
+    -v ./output:/app/output \
+    -p 7860:7860 \
+    -p 8000:8000 \
+    --shm-size 16G \
+    --name llamafactory \
+    llamafactory:latest
+
+docker exec -it llamafactory bash
+```
+
+昇腾 NPU 用户：
+
+```bash
+# 根据您的环境选择镜像
+docker build -f ./docker/docker-npu/Dockerfile \
+    --build-arg INSTALL_DEEPSPEED=false \
+    --build-arg PIP_INDEX=https://pypi.org/simple \
+    -t llamafactory:latest .
+
+# 根据您的资源更改 `device`
+docker run -dit \
+    -v ./hf_cache:/root/.cache/huggingface \
+    -v ./ms_cache:/root/.cache/modelscope \
+    -v ./data:/app/data \
+    -v ./output:/app/output \
+    -v /usr/local/dcmi:/usr/local/dcmi \
+    -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
+    -v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
+    -v /etc/ascend_install.info:/etc/ascend_install.info \
+    -p 7860:7860 \
+    -p 8000:8000 \
+    --device /dev/davinci0 \
+    --device /dev/davinci_manager \
+    --device /dev/devmm_svm \
+    --device /dev/hisi_hdc \
+    --shm-size 16G \
+    --name llamafactory \
+    llamafactory:latest
+
+docker exec -it llamafactory bash
 ```
 
 </details>
 
-#### 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
-```
-
 <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。
@@ -444,9 +506,12 @@ docker compose -f ./docker-compose.yml up -d
 ### 利用 vLLM 部署 OpenAI API
 
 ```bash
-CUDA_VISIBLE_DEVICES=0,1 API_PORT=8000 llamafactory-cli api examples/inference/llama3_vllm.yaml
+API_PORT=8000 llamafactory-cli api examples/inference/llama3_vllm.yaml
 ```
 
+> [!TIP]
+> API 文档请查阅 https://platform.openai.com/docs/api-reference/chat/create。
+
 ### 从魔搭社区下载
 
 如果您在 Hugging Face 模型和数据集的下载中遇到了问题，可以通过下述方法使用魔搭社区。
@@ -455,7 +520,18 @@ CUDA_VISIBLE_DEVICES=0,1 API_PORT=8000 llamafactory-cli api examples/inference/l
 export USE_MODELSCOPE_HUB=1 # Windows 使用 `set USE_MODELSCOPE_HUB=1`
 ```
 
-将 `--model_name_or_path` 设置为模型 ID 来加载对应的模型。在[魔搭社区](https://modelscope.cn/models)查看所有可用的模型，例如 `LLM-Research/Meta-Llama-3-8B-Instruct`。
+将 `model_name_or_path` 设置为模型 ID 来加载对应的模型。在[魔搭社区](https://modelscope.cn/models)查看所有可用的模型，例如 `LLM-Research/Meta-Llama-3-8B-Instruct`。
+
+### 使用 W&B 面板
+
+若要使用 [Weights & Biases](https://wandb.ai) 记录实验数据，请在 yaml 文件中添加下面的参数。
+
+```yaml
+report_to: wandb
+run_name: test_run # 可选
+```
+
+在启动训练任务时，将 `WANDB_API_KEY` 设置为[密钥](https://wandb.ai/authorize)来登录 W&B 账户。
 
 ## 使用了 LLaMA Factory 的项目
 
@@ -468,45 +544,73 @@ export USE_MODELSCOPE_HUB=1 # Windows 使用 `set USE_MODELSCOPE_HUB=1`
 1. Wang et al. UbiPhysio: Support Daily Functioning, Fitness, and Rehabilitation with Action Understanding and Feedback in Natural Language. 2023. [[arxiv]](https://arxiv.org/abs/2308.10526)
 1. Luceri et al. Leveraging Large Language Models to Detect Influence Campaigns in Social Media. 2023. [[arxiv]](https://arxiv.org/abs/2311.07816)
 1. Zhang et al. Alleviating Hallucinations of Large Language Models through Induced Hallucinations. 2023. [[arxiv]](https://arxiv.org/abs/2312.15710)
-1. Wang et al. Know Your Needs Better: Towards Structured Understanding of Marketer Demands with Analogical Reasoning Augmented LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2401.04319)
+1. Wang et al. Know Your Needs Better: Towards Structured Understanding of Marketer Demands with Analogical Reasoning Augmented LLMs. KDD 2024. [[arxiv]](https://arxiv.org/abs/2401.04319)
-1. Wang et al. CANDLE: Iterative Conceptualization and Instantiation Distillation from Large Language Models for Commonsense Reasoning. 2024. [[arxiv]](https://arxiv.org/abs/2401.07286)
+1. Wang et al. CANDLE: Iterative Conceptualization and Instantiation Distillation from Large Language Models for Commonsense Reasoning. ACL 2024. [[arxiv]](https://arxiv.org/abs/2401.07286)
 1. Choi et al. FACT-GPT: Fact-Checking Augmentation via Claim Matching with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2402.05904)
 1. Zhang et al. AutoMathText: Autonomous Data Selection with Language Models for Mathematical Texts. 2024. [[arxiv]](https://arxiv.org/abs/2402.07625)
 1. Lyu et al. KnowTuning: Knowledge-aware Fine-tuning for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11176)
 1. Yang et al. LaCo: Large Language Model Pruning via Layer Collaps. 2024. [[arxiv]](https://arxiv.org/abs/2402.11187)
 1. Bhardwaj et al. Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task Arithmetic. 2024. [[arxiv]](https://arxiv.org/abs/2402.11746)
 1. Yang et al. Enhancing Empathetic Response Generation by Augmenting LLMs with Small-scale Empathetic Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11801)
-1. Yi et al. Generation Meets Verification: Accelerating Large Language Model Inference with Smart Parallel Auto-Correct Decoding. 2024. [[arxiv]](https://arxiv.org/abs/2402.11809)
+1. Yi et al. Generation Meets Verification: Accelerating Large Language Model Inference with Smart Parallel Auto-Correct Decoding. ACL 2024 Findings. [[arxiv]](https://arxiv.org/abs/2402.11809)
 1. Cao et al. Head-wise Shareable Attention for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11819)
 1. Zhang et al. Enhancing Multilingual Capabilities of Large Language Models through Self-Distillation from Resource-Rich Languages. 2024. [[arxiv]](https://arxiv.org/abs/2402.12204)
 1. Kim et al. Efficient and Effective Vocabulary Expansion Towards Multilingual Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.14714)
-1. Yu et al. KIEval: A Knowledge-grounded Interactive Evaluation Framework for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.15043)
+1. Yu et al. KIEval: A Knowledge-grounded Interactive Evaluation Framework for Large Language Models. ACL 2024. [[arxiv]](https://arxiv.org/abs/2402.15043)
 1. Huang et al. Key-Point-Driven Data Synthesis with its Enhancement on Mathematical Reasoning. 2024. [[arxiv]](https://arxiv.org/abs/2403.02333)
 1. Duan et al. Negating Negatives: Alignment without Human Positive Samples via Distributional Dispreference Optimization. 2024. [[arxiv]](https://arxiv.org/abs/2403.03419)
 1. Xie and Schwertfeger. Empowering Robotics with Large Language Models: osmAG Map Comprehension with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2403.08228)
 1. Wu et al. Large Language Models are Parallel Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2403.09073)
 1. Zhang et al. EDT: Improving Large Language Models' Generation by Entropy-based Dynamic Temperature Sampling. 2024. [[arxiv]](https://arxiv.org/abs/2403.14541)
 1. Weller et al. FollowIR: Evaluating and Teaching Information Retrieval Models to Follow Instructions. 2024. [[arxiv]](https://arxiv.org/abs/2403.15246)
-1. Hongbin Na. CBT-LLM: A Chinese Large Language Model for Cognitive Behavioral Therapy-based Mental Health Question Answering. 2024. [[arxiv]](https://arxiv.org/abs/2403.16008)
+1. Hongbin Na. CBT-LLM: A Chinese Large Language Model for Cognitive Behavioral Therapy-based Mental Health Question Answering. COLING 2024. [[arxiv]](https://arxiv.org/abs/2403.16008)
 1. Zan et al. CodeS: Natural Language to Code Repository via Multi-Layer Sketch. 2024. [[arxiv]](https://arxiv.org/abs/2403.16443)
 1. Liu et al. Extensive Self-Contrast Enables Feedback-Free Language Model Alignment. 2024. [[arxiv]](https://arxiv.org/abs/2404.00604)
 1. Luo et al. BAdam: A Memory Efficient Full Parameter Training Method for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.02827)
 1. Du et al. Chinese Tiny LLM: Pretraining a Chinese-Centric Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2404.04167)
-1. Ma et al. Parameter Efficient Quasi-Orthogonal Fine-Tuning via Givens Rotation. 2024. [[arxiv]](https://arxiv.org/abs/2404.04316)
+1. Ma et al. Parameter Efficient Quasi-Orthogonal Fine-Tuning via Givens Rotation. ICML 2024. [[arxiv]](https://arxiv.org/abs/2404.04316)
 1. Liu et al. Dynamic Generation of Personalities with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.07084)
 1. Shang et al. How Far Have We Gone in Stripped Binary Code Understanding Using Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.09836)
 1. Huang et al. LLMTune: Accelerate Database Knob Tuning with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.11581)
 1. Deng et al. Text-Tuple-Table: Towards Information Integration in Text-to-Table Generation via Global Tuple Extraction. 2024. [[arxiv]](https://arxiv.org/abs/2404.14215)
 1. Acikgoz et al. Hippocrates: An Open-Source Framework for Advancing Large Language Models in Healthcare. 2024. [[arxiv]](https://arxiv.org/abs/2404.16621)
-1. Zhang et al. Small Language Models Need Strong Verifiers to Self-Correct Reasoning. 2024. [[arxiv]](https://arxiv.org/abs/2404.17140)
+1. Zhang et al. Small Language Models Need Strong Verifiers to Self-Correct Reasoning. ACL 2024 Findings. [[arxiv]](https://arxiv.org/abs/2404.17140)
-1. Zhou et al. FREB-TQA: A Fine-Grained Robustness Evaluation Benchmark for Table Question Answering. 2024. [[arxiv]](https://arxiv.org/abs/2404.18585)
+1. Zhou et al. FREB-TQA: A Fine-Grained Robustness Evaluation Benchmark for Table Question Answering. NAACL 2024. [[arxiv]](https://arxiv.org/abs/2404.18585)
+1. Xu et al. Large Language Models for Cyber Security: A Systematic Literature Review. 2024. [[arxiv]](https://arxiv.org/abs/2405.04760)
+1. Dammu et al. "They are uncultured": Unveiling Covert Harms and Social Threats in LLM Generated Conversations. 2024. [[arxiv]](https://arxiv.org/abs/2405.05378)
+1. Yi et al. A safety realignment framework via subspace-oriented model fusion for large language models. 2024. [[arxiv]](https://arxiv.org/abs/2405.09055)
+1. Lou et al. SPO: Multi-Dimensional Preference Sequential Alignment With Implicit Reward Modeling. 2024. [[arxiv]](https://arxiv.org/abs/2405.12739)
+1. Zhang et al. Getting More from Less: Large Language Models are Good Spontaneous Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2405.13816)
+1. Zhang et al. TS-Align: A Teacher-Student Collaborative Framework for Scalable Iterative Finetuning of Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2405.20215)
+1. Zihong Chen. Sentence Segmentation and Sentence Punctuation Based on XunziALLM. 2024. [[paper]](https://aclanthology.org/2024.lt4hala-1.30)
+1. Gao et al. The Best of Both Worlds: Toward an Honest and Helpful Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2406.00380)
+1. Wang and Song. MARS: Benchmarking the Metaphysical Reasoning Abilities of Language Models with a Multi-task Evaluation Dataset. 2024. [[arxiv]](https://arxiv.org/abs/2406.02106)
+1. Hu et al. Computational Limits of Low-Rank Adaptation (LoRA) for Transformer-Based Models. 2024. [[arxiv]](https://arxiv.org/abs/2406.03136)
+1. Ge et al. Time Sensitive Knowledge Editing through Efficient Finetuning. ACL 2024. [[arxiv]](https://arxiv.org/abs/2406.04496)
+1. Tan et al. Peer Review as A Multi-Turn and Long-Context Dialogue with Role-Based Interactions. 2024. [[arxiv]](https://arxiv.org/abs/2406.05688)
+1. Song et al. Turbo Sparse: Achieving LLM SOTA Performance with Minimal Activated Parameters. 2024. [[arxiv]](https://arxiv.org/abs/2406.05955)
+1. Gu et al. RWKV-CLIP: A Robust Vision-Language Representation Learner. 2024. [[arxiv]](https://arxiv.org/abs/2406.06973)
+1. Chen et al. Advancing Tool-Augmented Large Language Models: Integrating Insights from Errors in Inference Trees. 2024. [[arxiv]](https://arxiv.org/abs/2406.07115)
+1. Zhu et al. Are Large Language Models Good Statisticians?. 2024. [[arxiv]](https://arxiv.org/abs/2406.07815)
+1. Li et al. Know the Unknown: An Uncertainty-Sensitive Method for LLM Instruction Tuning. 2024. [[arxiv]](https://arxiv.org/abs/2406.10099)
+1. Ding et al. IntentionQA: A Benchmark for Evaluating Purchase Intention Comprehension Abilities of Language Models in E-commerce. 2024. [[arxiv]](https://arxiv.org/abs/2406.10173)
+1. He et al. COMMUNITY-CROSS-INSTRUCT: Unsupervised Instruction Generation for Aligning Large Language Models to Online Communities. 2024. [[arxiv]](https://arxiv.org/abs/2406.12074)
+1. Lin et al. FVEL: Interactive Formal Verification Environment with Large Language Models via Theorem Proving. 2024. [[arxiv]](https://arxiv.org/abs/2406.14408)
+1. Treutlein et al. Connecting the Dots: LLMs can Infer and Verbalize Latent Structure from Disparate Training Data. 2024. [[arxiv]](https://arxiv.org/abs/2406.14546)
+1. Feng et al. SS-Bench: A Benchmark for Social Story Generation and Evaluation. 2024. [[arxiv]](https://arxiv.org/abs/2406.15695)
+1. Feng et al. Self-Constructed Context Decompilation with Fined-grained Alignment Enhancement. 2024. [[arxiv]](https://arxiv.org/abs/2406.17233)
+1. Liu et al. Large Language Models for Cuffless Blood Pressure Measurement From Wearable Biosignals. 2024. [[arxiv]](https://arxiv.org/abs/2406.18069)
+1. Iyer et al. Exploring Very Low-Resource Translation with LLMs: The University of Edinburgh’s Submission to AmericasNLP 2024 Translation Task. AmericasNLP 2024. [[paper]](https://aclanthology.org/2024.americasnlp-1.25)
 1. **[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. **[Sunsimiao](https://github.com/X-D-Lab/Sunsimiao)**: 孙思邈中文医疗大模型 Sumsimiao，基于 Baichuan-7B 和 ChatGLM-6B 在中文医疗数据上微调而得。
 1. **[CareGPT](https://github.com/WangRongsheng/CareGPT)**: 医疗大模型项目 CareGPT，基于 LLaMA2-7B 和 Baichuan-13B 在中文医疗数据上微调而得。
 1. **[MachineMindset](https://github.com/PKU-YuanGroup/Machine-Mindset/)**：MBTI性格大模型项目，根据数据集与训练方式让任意 LLM 拥有 16 个不同的性格类型。
 1. **[Luminia-13B-v3](https://huggingface.co/Nekochu/Luminia-13B-v3)**：一个用于生成 Stable Diffusion 提示词的大型语言模型。[[🤗Demo]](https://huggingface.co/spaces/Nekochu/Luminia-13B_SD_Prompt)
 1. **[Chinese-LLaVA-Med](https://github.com/BUAADreamer/Chinese-LLaVA-Med)**：中文多模态医学大模型，基于 LLaVA-1.5-7B 在中文多模态医疗数据上微调而得。
+1. **[AutoRE](https://github.com/THUDM/AutoRE)**：基于大语言模型的文档级关系抽取系统。
+1. **[NVIDIA RTX AI Toolkit](https://github.com/NVIDIA/RTX-AI-Toolkit)**：在 Windows 主机上利用英伟达 RTX 设备进行大型语言模型微调的开发包。
+1. **[LazyLLM](https://github.com/LazyAGI/LazyLLM)**：一个低代码构建多 Agent 大模型应用的开发工具，支持基于 LLaMA Factory 的模型微调.
 
 </details>
 
@@ -514,17 +618,19 @@ export USE_MODELSCOPE_HUB=1 # Windows 使用 `set USE_MODELSCOPE_HUB=1`
 
 本仓库的代码依照 [Apache-2.0](LICENSE) 协议开源。
 
-使用模型权重时，请遵循对应的模型协议：[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 (LLaVA-1.5)](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) / [Phi-3](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/blob/main/LICENSE) / [Qwen](https://github.com/QwenLM/Qwen/blob/main/Tongyi%20Qianwen%20LICENSE%20AGREEMENT) / [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) / [Yi-1.5](LICENSE) / [Yuan](https://github.com/IEIT-Yuan/Yuan-2.0/blob/main/LICENSE-Yuan)
+使用模型权重时，请遵循对应的模型协议：[Baichuan 2](https://huggingface.co/baichuan-inc/Baichuan2-7B-Base/blob/main/Community%20License%20for%20Baichuan%202%20Model.pdf) / [BLOOM](https://huggingface.co/spaces/bigscience/license) / [ChatGLM3](https://github.com/THUDM/ChatGLM3/blob/main/MODEL_LICENSE) / [Command R](https://cohere.com/c4ai-cc-by-nc-license) / [DeepSeek](https://github.com/deepseek-ai/DeepSeek-LLM/blob/main/LICENSE-MODEL) / [Falcon](https://huggingface.co/tiiuae/falcon-180B/blob/main/LICENSE.txt) / [Gemma](https://ai.google.dev/gemma/terms) / [GLM-4](https://huggingface.co/THUDM/glm-4-9b/blob/main/LICENSE) / [InternLM2](https://github.com/InternLM/InternLM#license) / [Llama](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md) / [Llama 2 (LLaVA-1.5)](https://ai.meta.com/llama/license/) / [Llama 3](https://llama.meta.com/llama3/license/) / [Mistral](LICENSE) / [OLMo](LICENSE) / [Phi-1.5/Phi-2](https://huggingface.co/microsoft/phi-1_5/resolve/main/Research%20License.docx) / [Phi-3](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/blob/main/LICENSE) / [Qwen](https://github.com/QwenLM/Qwen/blob/main/Tongyi%20Qianwen%20LICENSE%20AGREEMENT) / [StarCoder 2](https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement) / [XVERSE](https://github.com/xverse-ai/XVERSE-13B/blob/main/MODEL_LICENSE.pdf) / [Yi](https://huggingface.co/01-ai/Yi-6B/blob/main/LICENSE) / [Yi-1.5](LICENSE) / [Yuan 2](https://github.com/IEIT-Yuan/Yuan-2.0/blob/main/LICENSE-Yuan)
 
 ## 引用
 
 如果您觉得此项目有帮助，请考虑以下列格式引用
 
 ```bibtex
-@article{zheng2024llamafactory,
+@inproceedings{zheng2024llamafactory,
-  title={LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models}, 
+  title={LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models},
-  author={Yaowei Zheng and Richong Zhang and Junhao Zhang and Yanhan Ye and Zheyan Luo and Yongqiang Ma},
+  author={Yaowei Zheng and Richong Zhang and Junhao Zhang and Yanhan Ye and Zheyan Luo and Zhangchi Feng and Yongqiang Ma},
-  journal={arXiv preprint arXiv:2403.13372},
+  booktitle={Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
+  address={Bangkok, Thailand},
+  publisher={Association for Computational Linguistics},
   year={2024},
   url={http://arxiv.org/abs/2403.13372}
 }

data/README.md

@@ -1,16 +1,19 @@
-If you are using a custom dataset, please add your **dataset description** to `dataset_info.json` according to the following format. We also provide several examples in the next section.
+The [dataset_info.json](dataset_info.json) contains all available datasets. If you are using a custom dataset, please **make sure** to add a *dataset description* in `dataset_info.json` and specify `dataset: dataset_name` before training to use it.
+
+Currently we support datasets in **alpaca** and **sharegpt** format.
 
 ```json
 "dataset_name": {
   "hf_hub_url": "the name of the dataset repository on the Hugging Face hub. (if specified, ignore script_url and file_name)",
-  "ms_hub_url": "the name of the dataset repository on the ModelScope hub. (if specified, ignore script_url and file_name)",
+  "ms_hub_url": "the name of the dataset repository on the Model Scope hub. (if specified, ignore script_url and file_name)",
   "script_url": "the name of the directory containing a dataset loading script. (if specified, ignore file_name)",
-  "file_name": "the name of the dataset file in this directory. (required if above are not specified)",
+  "file_name": "the name of the dataset folder or dataset file in this directory. (required if above are not specified)",
-  "file_sha1": "the SHA-1 hash value of the dataset file. (optional, does not affect training)",
-  "subset": "the name of the subset. (optional, default: None)",
-  "folder": "the name of the folder of the dataset repository on the Hugging Face hub. (optional, default: None)",
-  "ranking": "whether the dataset is a preference dataset or not. (default: false)",
   "formatting": "the format of the dataset. (optional, default: alpaca, can be chosen from {alpaca, sharegpt})",
+  "ranking": "whether the dataset is a preference dataset or not. (default: False)",
+  "subset": "the name of the subset. (optional, default: None)",
+  "split": "the name of dataset split to be used. (optional, default: train)",
+  "folder": "the name of the folder of the dataset repository on the Hugging Face hub. (optional, default: None)",
+  "num_samples": "the number of samples in the dataset to be used. (optional, default: None)",
   "columns (optional)": {
     "prompt": "the column name in the dataset containing the prompts. (default: instruction)",
     "query": "the column name in the dataset containing the queries. (default: input)",
@@ -19,7 +22,10 @@ If you are using a custom dataset, please add your **dataset description** to `d
     "messages": "the column name in the dataset containing the messages. (default: conversations)",
     "system": "the column name in the dataset containing the system prompts. (default: None)",
     "tools": "the column name in the dataset containing the tool description. (default: None)",
-    "images": "the column name in the dataset containing the image inputs. (default: None)"
+    "images": "the column name in the dataset containing the image inputs. (default: None)",
+    "chosen": "the column name in the dataset containing the chosen answers. (default: None)",
+    "rejected": "the column name in the dataset containing the rejected answers. (default: None)",
+    "kto_tag": "the column name in the dataset containing the kto tags. (default: None)"
   },
   "tags (optional, used for the sharegpt format)": {
     "role_tag": "the key in the message represents the identity. (default: from)",
@@ -33,28 +39,34 @@ If you are using a custom dataset, please add your **dataset description** to `d
 }
 ```
 
-After that, you can load the custom dataset by specifying `--dataset dataset_name`.
+## Alpaca Format
 
-----
+### Supervised Fine-Tuning Dataset
 
-Currently we support dataset in **alpaca** or **sharegpt** format, the dataset in alpaca format should follow the below format:
+* [Example dataset](alpaca_en_demo.json)
+
+In supervised fine-tuning, the `instruction` column will be concatenated with the `input` column and used as the human prompt, then the human prompt would be `instruction\ninput`. The `output` column represents the model response.
+
+The `system` column will be used as the system prompt if specified.
+
+The `history` column is a list consisting of string tuples representing prompt-response pairs in the history messages. Note that the responses in the history **will also be learned by the model** in supervised fine-tuning.
 
 ```json
 [
   {
-    "instruction": "user instruction (required)",
+    "instruction": "human instruction (required)",
-    "input": "user input (optional)",
+    "input": "human input (optional)",
     "output": "model response (required)",
     "system": "system prompt (optional)",
     "history": [
-      ["user instruction in the first round (optional)", "model response in the first round (optional)"],
+      ["human instruction in the first round (optional)", "model response in the first round (optional)"],
-      ["user instruction in the second round (optional)", "model response in the second round (optional)"]
+      ["human instruction in the second round (optional)", "model response in the second round (optional)"]
     ]
   }
 ]
 ```
 
-Regarding the above dataset, the description in `dataset_info.json` should be:
+Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
 
 ```json
 "dataset_name": {
@@ -69,11 +81,11 @@ Regarding the above dataset, the description in `dataset_info.json` should be:
 }
 ```
 
-The `query` column will be concatenated with the `prompt` column and used as the user prompt, then the user prompt would be `prompt\nquery`. The `response` column represents the model response.
+### Pre-training Dataset
 
-The `system` column will be used as the system prompt. The `history` column is a list consisting string tuples representing prompt-response pairs in the history. Note that the responses in the history **will also be used for training** in supervised fine-tuning.
+- [Example dataset](c4_demo.json)
 
-For the **pre-training datasets**, only the `prompt` column will be used for training, for example:
+In pre-training, only the `text` column will be used for model learning.
 
 ```json
 [
@@ -82,7 +94,7 @@ For the **pre-training datasets**, only the `prompt` column will be used for tra
 ]
 ```
 
-Regarding the above dataset, the description in `dataset_info.json` should be:
+Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
 
 ```json
 "dataset_name": {
@@ -93,22 +105,24 @@ Regarding the above dataset, the description in `dataset_info.json` should be:
 }
 ```
 
-For the **preference datasets**, the `response` column should be a string list whose length is 2, with the preferred answers appearing first, for example:
+### Preference Dataset
+
+Preference datasets are used for reward modeling, DPO training and ORPO training.
+
+It requires a better response in `chosen` column and a worse response in `rejected` column.
 
 ```json
 [
   {
-    "instruction": "user instruction",
+    "instruction": "human instruction (required)",
-    "input": "user input",
+    "input": "human input (optional)",
-    "output": [
+    "chosen": "chosen answer (required)",
-      "chosen answer",
+    "rejected": "rejected answer (required)"
-      "rejected answer"
-    ]
   }
 ]
 ```
 
-Regarding the above dataset, the description in `dataset_info.json` should be:
+Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
 
 ```json
 "dataset_name": {
@@ -117,14 +131,85 @@ Regarding the above dataset, the description in `dataset_info.json` should be:
   "columns": {
     "prompt": "instruction",
     "query": "input",
-    "response": "output",
+    "chosen": "chosen",
+    "rejected": "rejected"
   }
 }
 ```
 
-----
+### KTO Dataset
 
-The dataset in **sharegpt** format should follow the below format:
+- [Example dataset](kto_en_demo.json)
+
+KTO datasets require a extra `kto_tag` column containing the boolean human feedback.
+
+```json
+[
+  {
+    "instruction": "human instruction (required)",
+    "input": "human input (optional)",
+    "output": "model response (required)",
+    "kto_tag": "human feedback [true/false] (required)"
+  }
+]
+```
+
+Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
+
+```json
+"dataset_name": {
+  "file_name": "data.json",
+  "columns": {
+    "prompt": "instruction",
+    "query": "input",
+    "response": "output",
+    "kto_tag": "kto_tag"
+  }
+}
+```
+
+### Multimodal Dataset
+
+- [Example dataset](mllm_demo.json)
+
+Multimodal datasets require a `images` column containing the paths to the input images. Currently we only support one image.
+
+```json
+[
+  {
+    "instruction": "human instruction (required)",
+    "input": "human input (optional)",
+    "output": "model response (required)",
+    "images": [
+      "image path (required)"
+    ]
+  }
+]
+```
+
+Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
+
+```json
+"dataset_name": {
+  "file_name": "data.json",
+  "columns": {
+    "prompt": "instruction",
+    "query": "input",
+    "response": "output",
+    "images": "images"
+  }
+}
+```
+
+## Sharegpt Format
+
+### Supervised Fine-Tuning Dataset
+
+- [Example dataset](glaive_toolcall_en_demo.json)
+
+Compared to the alpaca format, the sharegpt format allows the datasets have **more roles**, such as human, gpt, observation and function. They are presented in a list of objects in the `conversations` column.
+
+Note that the human and observation should appear in odd positions, while gpt and function should appear in even positions.
 
 ```json
 [
@@ -132,7 +217,15 @@ The dataset in **sharegpt** format should follow the below format:
     "conversations": [
       {
         "from": "human",
-        "value": "user instruction"
+        "value": "human instruction"
+      },
+      {
+        "from": "function_call",
+        "value": "tool arguments"
+      },
+      {
+        "from": "observation",
+        "value": "tool result"
       },
       {
         "from": "gpt",
@@ -145,7 +238,7 @@ The dataset in **sharegpt** format should follow the below format:
 ]
 ```
 
-Regarding the above dataset, the description in `dataset_info.json` should be:
+Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
 
 ```json
 "dataset_name": {
@@ -155,19 +248,63 @@ Regarding the above dataset, the description in `dataset_info.json` should be:
     "messages": "conversations",
     "system": "system",
     "tools": "tools"
-  },
-  "tags": {
-    "role_tag": "from",
-    "content_tag": "value",
-    "user_tag": "human",
-    "assistant_tag": "gpt"
   }
 }
 ```
 
-where the `messages` column should be a list following the `u/a/u/a/u/a` order.
+### Preference Dataset
 
-We also supports the dataset in the **openai** format:
+- [Example dataset](dpo_en_demo.json)
+
+Preference datasets in sharegpt format also require a better message in `chosen` column and a worse message in `rejected` column.
+
+```json
+[
+  {
+    "conversations": [
+      {
+        "from": "human",
+        "value": "human instruction"
+      },
+      {
+        "from": "gpt",
+        "value": "model response"
+      },
+      {
+        "from": "human",
+        "value": "human instruction"
+      }
+    ],
+    "chosen": {
+      "from": "gpt",
+      "value": "chosen answer (required)"
+    },
+    "rejected": {
+      "from": "gpt",
+      "value": "rejected answer (required)"
+    }
+  }
+]
+```
+
+Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
+
+```json
+"dataset_name": {
+  "file_name": "data.json",
+  "formatting": "sharegpt",
+  "ranking": true,
+  "columns": {
+    "messages": "conversations",
+    "chosen": "chosen",
+    "rejected": "rejected"
+  }
+}
+```
+
+### OpenAI Format
+
+The openai format is simply a special case of the sharegpt format, where the first message may be a system prompt.
 
 ```json
 [
@@ -179,7 +316,7 @@ We also supports the dataset in the **openai** format:
       },
       {
         "role": "user",
-        "content": "user instruction"
+        "content": "human instruction"
       },
       {
         "role": "assistant",
@@ -190,7 +327,7 @@ We also supports the dataset in the **openai** format:
 ]
 ```
 
-Regarding the above dataset, the description in `dataset_info.json` should be:
+Regarding the above dataset, the *dataset description* in `dataset_info.json` should be:
 
 ```json
 "dataset_name": {
@@ -209,4 +346,6 @@ Regarding the above dataset, the description in `dataset_info.json` should be:
 }
 ```
 
-Pre-training datasets and preference datasets are **incompatible** with the sharegpt format yet.
+The KTO datasets and multimodal datasets in sharegpt format are similar to the alpaca format.
+
+Pre-training datasets are **incompatible** with the sharegpt format.

data/README_zh.md

@@ -1,16 +1,19 @@
-如果您使用自定义数据集，请务必按照以下格式在 `dataset_info.json` 文件中添加**数据集描述**。我们在下面也提供了一些例子。
+[dataset_info.json](dataset_info.json) 包含了所有可用的数据集。如果您希望使用自定义数据集，请**务必**在 `dataset_info.json` 文件中添加*数据集描述*，并通过修改 `dataset: 数据集名称` 配置来使用数据集。
+
+目前我们支持 **alpaca** 格式和 **sharegpt** 格式的数据集。
 
 ```json
 "数据集名称": {
   "hf_hub_url": "Hugging Face 的数据集仓库地址（若指定，则忽略 script_url 和 file_name）",
   "ms_hub_url": "ModelScope 的数据集仓库地址（若指定，则忽略 script_url 和 file_name）",
   "script_url": "包含数据加载脚本的本地文件夹名称（若指定，则忽略 file_name）",
-  "file_name": "该目录下数据集文件的名称（若上述参数未指定，则此项必需）",
+  "file_name": "该目录下数据集文件夹或文件的名称（若上述参数未指定，则此项必需）",
-  "file_sha1": "数据集文件的 SHA-1 哈希值（可选，留空不影响训练）",
-  "subset": "数据集子集的名称（可选，默认：None）",
-  "folder": "Hugging Face 仓库的文件夹名称（可选，默认：None）",
-  "ranking": "是否为偏好数据集（可选，默认：False）",
   "formatting": "数据集格式（可选，默认：alpaca，可以为 alpaca 或 sharegpt）",
+  "ranking": "是否为偏好数据集（可选，默认：False）",
+  "subset": "数据集子集的名称（可选，默认：None）",
+  "split": "所使用的数据集切分（可选，默认：train）",
+  "folder": "Hugging Face 仓库的文件夹名称（可选，默认：None）",
+  "num_samples": "该数据集所使用的样本数量。（可选，默认：None）",
   "columns（可选）": {
     "prompt": "数据集代表提示词的表头名称（默认：instruction）",
     "query": "数据集代表请求的表头名称（默认：input）",
@@ -19,7 +22,10 @@
     "messages": "数据集代表消息列表的表头名称（默认：conversations）",
     "system": "数据集代表系统提示的表头名称（默认：None）",
     "tools": "数据集代表工具描述的表头名称（默认：None）",
-    "images": "数据集代表图像输入的表头名称（默认：None）"
+    "images": "数据集代表图像输入的表头名称（默认：None）",
+    "chosen": "数据集代表更优回答的表头名称（默认：None）",
+    "rejected": "数据集代表更差回答的表头名称（默认：None）",
+    "kto_tag": "数据集代表 KTO 标签的表头名称（默认：None）"
   },
   "tags（可选，用于 sharegpt 格式）": {
     "role_tag": "消息中代表发送者身份的键名（默认：from）",
@@ -28,22 +34,28 @@
     "assistant_tag": "消息中代表助手的 role_tag（默认：gpt）",
     "observation_tag": "消息中代表工具返回结果的 role_tag（默认：observation）",
     "function_tag": "消息中代表工具调用的 role_tag（默认：function_call）",
-    "system_tag": "消息中代表系统提示的 role_tag（默认：system，会覆盖 system 列）"
+    "system_tag": "消息中代表系统提示的 role_tag（默认：system，会覆盖 system column）"
   }
 }
 ```
 
-然后，可通过使用 `--dataset 数据集名称` 参数加载自定义数据集。
+## Alpaca 格式
 
-----
+### 指令监督微调数据集
 
-该项目目前支持两种格式的数据集：**alpaca** 和 **sharegpt**，其中 alpaca 格式的数据集按照以下方式组织：
+- [样例数据集](alpaca_zh_demo.json)
+
+在指令监督微调时，`instruction` 列对应的内容会与 `input` 列对应的内容拼接后作为人类指令，即人类指令为 `instruction\ninput`。而 `output` 列对应的内容为模型回答。
+
+如果指定，`system` 列对应的内容将被作为系统提示词。
+
+`history` 列是由多个字符串二元组构成的列表，分别代表历史消息中每轮对话的指令和回答。注意在指令监督微调时，历史消息中的回答内容**也会被用于模型学习**。
 
 ```json
 [
   {
-    "instruction": "用户指令（必填）",
+    "instruction": "人类指令（必填）",
-    "input": "用户输入（选填）",
+    "input": "人类输入（选填）",
     "output": "模型回答（必填）",
     "system": "系统提示词（选填）",
     "history": [
@@ -54,7 +66,7 @@
 ]
 ```
 
-对于上述格式的数据，`dataset_info.json` 中的描述应为：
+对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
 
 ```json
 "数据集名称": {
@@ -69,11 +81,11 @@
 }
 ```
 
-其中 `query` 列对应的内容会与 `prompt` 列对应的内容拼接后作为用户指令，即用户指令为 `prompt\nquery`。`response` 列对应的内容为模型回答。
+### 预训练数据集
 
-`system` 列对应的内容将被作为系统提示词。`history` 列是由多个字符串二元组构成的列表，分别代表历史消息中每轮的指令和回答。注意在指令监督学习时，历史消息中的回答**也会被用于训练**。
+- [样例数据集](c4_demo.json)
 
-对于**预训练数据集**，仅 `prompt` 列中的内容会用于模型训练，例如：
+在预训练时，只有 `text` 列中的内容会用于模型学习。
 
 ```json
 [
@@ -82,7 +94,7 @@
 ]
 ```
 
-对于上述格式的数据，`dataset_info.json` 中的描述应为：
+对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
 
 ```json
 "数据集名称": {
@@ -93,22 +105,24 @@
 }
 ```
 
-对于**偏好数据集**，`response` 列应当是一个长度为 2 的字符串列表，排在前面的代表更优的回答，例如：
+### 偏好数据集
+
+偏好数据集用于奖励模型训练、DPO 训练和 ORPO 训练。
+
+它需要在 `chosen` 列中提供更优的回答，并在 `rejected` 列中提供更差的回答。
 
 ```json
 [
   {
-    "instruction": "用户指令",
+    "instruction": "人类指令（必填）",
-    "input": "用户输入",
+    "input": "人类输入（选填）",
-    "output": [
+    "chosen": "优质回答（必填）",
-      "优质回答",
+    "rejected": "劣质回答（必填）"
-      "劣质回答"
-    ]
   }
 ]
 ```
 
-对于上述格式的数据，`dataset_info.json` 中的描述应为：
+对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
 
 ```json
 "数据集名称": {
@@ -117,14 +131,85 @@
   "columns": {
     "prompt": "instruction",
     "query": "input",
-    "response": "output",
+    "chosen": "chosen",
+    "rejected": "rejected"
   }
 }
 ```
 
-----
+### KTO 数据集
 
-而 **sharegpt** 格式的数据集按照以下方式组织：
+- [样例数据集](kto_en_demo.json)
+
+KTO 数据集需要额外添加一个 `kto_tag` 列，包含 bool 类型的人类反馈。
+
+```json
+[
+  {
+    "instruction": "人类指令（必填）",
+    "input": "人类输入（选填）",
+    "output": "模型回答（必填）",
+    "kto_tag": "人类反馈 [true/false]（必填）"
+  }
+]
+```
+
+对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
+
+```json
+"数据集名称": {
+  "file_name": "data.json",
+  "columns": {
+    "prompt": "instruction",
+    "query": "input",
+    "response": "output",
+    "kto_tag": "kto_tag"
+  }
+}
+```
+
+### 多模态数据集
+
+- [样例数据集](mllm_demo.json)
+
+多模态数据集需要额外添加一个 `images` 列，包含输入图像的路径。目前我们仅支持单张图像输入。
+
+```json
+[
+  {
+    "instruction": "人类指令（必填）",
+    "input": "人类输入（选填）",
+    "output": "模型回答（必填）",
+    "images": [
+      "图像路径（必填）"
+    ]
+  }
+]
+```
+
+对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
+
+```json
+"数据集名称": {
+  "file_name": "data.json",
+  "columns": {
+    "prompt": "instruction",
+    "query": "input",
+    "response": "output",
+    "images": "images"
+  }
+}
+```
+
+## Sharegpt 格式
+
+### 指令监督微调数据集
+
+- [样例数据集](glaive_toolcall_zh_demo.json)
+
+相比 alpaca 格式的数据集，sharegpt 格式支持**更多的角色种类**，例如 human、gpt、observation、function 等等。它们构成一个对象列表呈现在 `conversations` 列中。
+
+注意其中 human 和 observation 必须出现在奇数位置，gpt 和 function 必须出现在偶数位置。
 
 ```json
 [
@@ -132,7 +217,15 @@
     "conversations": [
       {
         "from": "human",
-        "value": "用户指令"
+        "value": "人类指令"
+      },
+      {
+        "from": "function_call",
+        "value": "工具参数"
+      },
+      {
+        "from": "observation",
+        "value": "工具结果"
       },
       {
         "from": "gpt",
@@ -145,7 +238,7 @@
 ]
 ```
 
-对于上述格式的数据，`dataset_info.json` 中的描述应为：
+对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
 
 ```json
 "数据集名称": {
@@ -155,19 +248,63 @@
     "messages": "conversations",
     "system": "system",
     "tools": "tools"
-  },
-  "tags": {
-    "role_tag": "from",
-    "content_tag": "value",
-    "user_tag": "human",
-    "assistant_tag": "gpt"
   }
 }
 ```
 
-其中 `messages` 列应当是一个列表，且符合 `用户/模型/用户/模型/用户/模型` 的顺序。
+### 偏好数据集
 
-我们同样支持 **openai** 格式的数据集：
+- [样例数据集](dpo_zh_demo.json)
+
+Sharegpt 格式的偏好数据集同样需要在 `chosen` 列中提供更优的消息，并在 `rejected` 列中提供更差的消息。
+
+```json
+[
+  {
+    "conversations": [
+      {
+        "from": "human",
+        "value": "人类指令"
+      },
+      {
+        "from": "gpt",
+        "value": "模型回答"
+      },
+      {
+        "from": "human",
+        "value": "人类指令"
+      }
+    ],
+    "chosen": {
+      "from": "gpt",
+      "value": "优质回答"
+    },
+    "rejected": {
+      "from": "gpt",
+      "value": "劣质回答"
+    }
+  }
+]
+```
+
+对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
+
+```json
+"数据集名称": {
+  "file_name": "data.json",
+  "formatting": "sharegpt",
+  "ranking": true,
+  "columns": {
+    "messages": "conversations",
+    "chosen": "chosen",
+    "rejected": "rejected"
+  }
+}
+```
+
+### OpenAI 格式
+
+OpenAI 格式仅仅是 sharegpt 格式的一种特殊情况，其中第一条消息可能是系统提示词。
 
 ```json
 [
@@ -179,7 +316,7 @@
       },
       {
         "role": "user",
-        "content": "用户指令"
+        "content": "人类指令"
       },
       {
         "role": "assistant",
@@ -190,7 +327,7 @@
 ]
 ```
 
-对于上述格式的数据，`dataset_info.json` 中的描述应为：
+对于上述格式的数据，`dataset_info.json` 中的*数据集描述*应为：
 
 ```json
 "数据集名称": {
@@ -209,4 +346,6 @@
 }
 ```
 
-预训练数据集和偏好数据集**尚不支持** sharegpt 格式。
+Sharegpt 格式中的 KTO 数据集和多模态数据集与 alpaca 格式的类似。
+
+预训练数据集**不支持** sharegpt 格式。

data/alpaca_data_en_52k.json.REMOVED.git-id

@@ -1 +0,0 @@
-3779ddbc040543ab1834ef216c983d6fcc06cc9a

data/alpaca_data_zh_51k.json.REMOVED.git-id

@@ -1 +0,0 @@
-a97cf9475291591843976554878568e046d8a46d

data/alpaca_gpt4_data_en.json.REMOVED.git-id

@@ -1 +0,0 @@
-25508714b7879a1e5a6764ba7f979a980f549f1a

data/alpaca_gpt4_data_zh.json.REMOVED.git-id

@@ -1 +0,0 @@
-7cb6a7d11455bddc3d495750a2392683d775b184

data/comparison_gpt4_data_en.json.REMOVED.git-id

@@ -1 +0,0 @@
-f5cb08305ff5dc9c17a09809c54c8c8834aadc70

data/comparison_gpt4_data_zh.json.REMOVED.git-id

@@ -1 +0,0 @@
-aee47b7b443496e37808d7f34ef10403ff99bcc3

data/example_dataset/example_dataset.py

@@ -1,37 +0,0 @@
-import json
-from typing import Any, Dict, Generator, List, Tuple
-
-import datasets
-
-
-_DESCRIPTION = "An example of dataset."
-_CITATION = ""
-_HOMEPAGE = ""
-_LICENSE = ""
-_URL = "examples.json"
-
-
-class ExampleDataset(datasets.GeneratorBasedBuilder):
-    VERSION = datasets.Version("0.0.0")
-
-    def _info(self) -> datasets.DatasetInfo:
-        features = datasets.Features(
-            {
-                "instruction": datasets.Value("string"),
-                "input": datasets.Value("string"),
-                "output": datasets.Value("string"),
-                "history": datasets.Sequence(datasets.Sequence(datasets.Value("string"))),
-            }
-        )
-        return datasets.DatasetInfo(
-            description=_DESCRIPTION, features=features, homepage=_HOMEPAGE, license=_LICENSE, citation=_CITATION
-        )
-
-    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
-        file_path = dl_manager.download(_URL)
-        return [datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": file_path})]
-
-    def _generate_examples(self, filepath: str) -> Generator[Tuple[int, Dict[str, Any]], None, None]:
-        example_dataset = json.load(open(filepath, "r", encoding="utf-8"))
-        for key, example in enumerate(example_dataset):
-            yield key, example

data/glaive_toolcall_10k.json.REMOVED.git-id

@@ -1 +0,0 @@
-4748dff00d1dc42768a5b6cc772143c313017812

data/hh_rlhf_en/hh_rlhf_en.py

@@ -34,7 +34,8 @@ class HhRlhfEn(datasets.GeneratorBasedBuilder):
         features = datasets.Features(
             {
                 "instruction": datasets.Value("string"),
-                "output": datasets.Sequence(datasets.Value("string")),
+                "chosen": datasets.Value("string"),
+                "rejected": datasets.Value("string"),
                 "history": datasets.Sequence(datasets.Sequence(datasets.Value("string"))),
             }
         )
@@ -79,5 +80,5 @@ class HhRlhfEn(datasets.GeneratorBasedBuilder):
                             break
                         prompt = prompt[:human_idx]
 
-                    yield key, {"instruction": query, "output": [r_accept, r_reject], "history": history}
+                    yield key, {"instruction": query, "chosen": r_accept, "rejected": r_reject, "history": history}
                     key += 1

data/orca_rlhf.json.REMOVED.git-id

@@ -1 +0,0 @@
-736bcedea2b24a1414765c6d69cbdafaea839f3c

data/wiki_demo.txt.REMOVED.git-id

@@ -1 +0,0 @@
-c9cf509b7fdac5490cfd6dae72c2d7b8a60af6cb

docker-compose.yml

@@ -1,25 +0,0 @@
-version: '3.8'
-
-services:
-  llama-factory:
-    build:
-      dockerfile: Dockerfile
-      context: .
-    container_name: llama_factory
-    volumes:
-      - ./hf_cache:/root/.cache/huggingface/
-      - ./data:/app/data
-      - ./output:/app/output
-    environment:
-      - CUDA_VISIBLE_DEVICES=0
-    ports:
-      - "7860:7860"
-    ipc: host
-    deploy:
-      resources:
-        reservations:
-          devices:
-          - driver: nvidia
-            count: "all"
-            capabilities: [gpu]
-    restart: unless-stopped

docker/docker-cuda/Dockerfile

@@ -0,0 +1,59 @@
+# Use the NVIDIA official image with PyTorch 2.3.0
+# https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel-24-02.html
+FROM nvcr.io/nvidia/pytorch:24.02-py3
+
+# Define environments
+ENV MAX_JOBS=4
+ENV FLASH_ATTENTION_FORCE_BUILD=TRUE
+ENV VLLM_WORKER_MULTIPROC_METHOD=spawn
+
+# Define installation arguments
+ARG INSTALL_BNB=false
+ARG INSTALL_VLLM=false
+ARG INSTALL_DEEPSPEED=false
+ARG INSTALL_FLASHATTN=false
+ARG PIP_INDEX=https://pypi.org/simple
+
+# Set the working directory
+WORKDIR /app
+
+# Install the requirements
+COPY requirements.txt /app
+RUN pip config set global.index-url "$PIP_INDEX" && \
+    pip config set global.extra-index-url "$PIP_INDEX" && \
+    python -m pip install --upgrade pip && \
+    python -m pip install -r requirements.txt
+
+# Copy the rest of the application into the image
+COPY . /app
+
+# Install the LLaMA Factory
+RUN EXTRA_PACKAGES="metrics"; \
+    if [ "$INSTALL_BNB" == "true" ]; then \
+        EXTRA_PACKAGES="${EXTRA_PACKAGES},bitsandbytes"; \
+    fi; \
+    if [ "$INSTALL_VLLM" == "true" ]; then \
+        EXTRA_PACKAGES="${EXTRA_PACKAGES},vllm"; \
+    fi; \
+    if [ "$INSTALL_DEEPSPEED" == "true" ]; then \
+        EXTRA_PACKAGES="${EXTRA_PACKAGES},deepspeed"; \
+    fi; \
+    pip install -e ".[$EXTRA_PACKAGES]"
+
+# Rebuild flash attention
+RUN pip uninstall -y transformer-engine flash-attn && \
+    if [ "$INSTALL_FLASHATTN" == "true" ]; then \
+        pip uninstall -y ninja && pip install ninja && \
+        pip install --no-cache-dir flash-attn --no-build-isolation; \
+    fi
+
+# Set up volumes
+VOLUME [ "/root/.cache/huggingface", "/root/.cache/modelscope", "/app/data", "/app/output" ]
+
+# Expose port 7860 for the LLaMA Board
+ENV GRADIO_SERVER_PORT 7860
+EXPOSE 7860
+
+# Expose port 8000 for the API service
+ENV API_PORT 8000
+EXPOSE 8000

docker/docker-cuda/docker-compose.yml

@@ -0,0 +1,32 @@
+services:
+  llamafactory:
+    build:
+      dockerfile: ./docker/docker-cuda/Dockerfile
+      context: ../..
+      args:
+        INSTALL_BNB: false
+        INSTALL_VLLM: false
+        INSTALL_DEEPSPEED: false
+        INSTALL_FLASHATTN: false
+        PIP_INDEX: https://pypi.org/simple
+    container_name: llamafactory
+    volumes:
+      - ../../hf_cache:/root/.cache/huggingface
+      - ../../ms_cache:/root/.cache/modelscope
+      - ../../data:/app/data
+      - ../../output:/app/output
+    ports:
+      - "7860:7860"
+      - "8000:8000"
+    ipc: host
+    tty: true
+    stdin_open: true
+    command: bash
+    deploy:
+      resources:
+        reservations:
+          devices:
+          - driver: nvidia
+            count: "all"
+            capabilities: [gpu]
+    restart: unless-stopped

docker/docker-npu/Dockerfile

@@ -0,0 +1,45 @@
+# Use the Ubuntu 22.04 image with CANN 8.0.rc1
+# More versions can be found at https://hub.docker.com/r/cosdt/cann/tags
+# FROM cosdt/cann:8.0.rc1-910-ubuntu22.04
+FROM cosdt/cann:8.0.rc1-910b-ubuntu22.04
+# FROM cosdt/cann:8.0.rc1-910-openeuler22.03
+# FROM cosdt/cann:8.0.rc1-910b-openeuler22.03
+
+# Define environments
+ENV DEBIAN_FRONTEND=noninteractive
+
+# Define installation arguments
+ARG INSTALL_DEEPSPEED=false
+ARG PIP_INDEX=https://pypi.org/simple
+ARG TORCH_INDEX=https://download.pytorch.org/whl/cpu
+
+# Set the working directory
+WORKDIR /app
+
+# Install the requirements
+COPY requirements.txt /app
+RUN pip config set global.index-url "$PIP_INDEX" && \
+    pip config set global.extra-index-url "$TORCH_INDEX" && \
+    python -m pip install --upgrade pip && \
+    python -m pip install -r requirements.txt
+
+# Copy the rest of the application into the image
+COPY . /app
+
+# Install the LLaMA Factory
+RUN EXTRA_PACKAGES="torch-npu,metrics"; \
+    if [ "$INSTALL_DEEPSPEED" == "true" ]; then \
+        EXTRA_PACKAGES="${EXTRA_PACKAGES},deepspeed"; \
+    fi; \
+    pip install -e ".[$EXTRA_PACKAGES]"
+
+# Set up volumes
+VOLUME [ "/root/.cache/huggingface", "/root/.cache/modelscope", "/app/data", "/app/output" ]
+
+# Expose port 7860 for the LLaMA Board
+ENV GRADIO_SERVER_PORT 7860
+EXPOSE 7860
+
+# Expose port 8000 for the API service
+ENV API_PORT 8000
+EXPOSE 8000

docker/docker-npu/docker-compose.yml

@@ -0,0 +1,31 @@
+services:
+  llamafactory:
+    build:
+      dockerfile: ./docker/docker-npu/Dockerfile
+      context: ../..
+      args:
+        INSTALL_DEEPSPEED: false
+        PIP_INDEX: https://pypi.org/simple
+    container_name: llamafactory
+    volumes:
+      - ../../hf_cache:/root/.cache/huggingface
+      - ../../ms_cache:/root/.cache/modelscope
+      - ../../data:/app/data
+      - ../../output:/app/output
+      - /usr/local/dcmi:/usr/local/dcmi
+      - /usr/local/bin/npu-smi:/usr/local/bin/npu-smi
+      - /usr/local/Ascend/driver:/usr/local/Ascend/driver
+      - /etc/ascend_install.info:/etc/ascend_install.info
+    ports:
+      - "7860:7860"
+      - "8000:8000"
+    ipc: host
+    tty: true
+    stdin_open: true
+    command: bash
+    devices:
+      - /dev/davinci0
+      - /dev/davinci_manager
+      - /dev/devmm_svm
+      - /dev/hisi_hdc
+    restart: unless-stopped

evaluation/ceval/ceval.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+
 import os
 
 import datasets

evaluation/cmmlu/cmmlu.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+
 import os
 
 import datasets

evaluation/mmlu/mmlu.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+
 import os
 
 import datasets
@@ -154,7 +155,7 @@ class MMLU(datasets.GeneratorBasedBuilder):
         ]
 
     def _generate_examples(self, filepath):
-        df = pd.read_csv(filepath)
+        df = pd.read_csv(filepath, header=None)
         df.columns = ["question", "A", "B", "C", "D", "answer"]
 
         for i, instance in enumerate(df.to_dict(orient="records")):

examples/README.md

@@ -4,59 +4,59 @@ Make sure to execute these commands in the `LLaMA-Factory` directory.
 
 ## Table of Contents
 
-- [LoRA Fine-Tuning on A Single GPU](#lora-fine-tuning-on-a-single-gpu)
+- [LoRA Fine-Tuning](#lora-fine-tuning)
-- [QLoRA Fine-Tuning on a Single GPU](#qlora-fine-tuning-on-a-single-gpu)
+- [QLoRA Fine-Tuning](#qlora-fine-tuning)
-- [LoRA Fine-Tuning on Multiple GPUs](#lora-fine-tuning-on-multiple-gpus)
+- [Full-Parameter Fine-Tuning](#full-parameter-fine-tuning)
-- [LoRA Fine-Tuning on Multiple NPUs](#lora-fine-tuning-on-multiple-npus)
-- [Full-Parameter Fine-Tuning on Multiple GPUs](#full-parameter-fine-tuning-on-multiple-gpus)
 - [Merging LoRA Adapters and Quantization](#merging-lora-adapters-and-quantization)
 - [Inferring LoRA Fine-Tuned Models](#inferring-lora-fine-tuned-models)
 - [Extras](#extras)
 
+Use `CUDA_VISIBLE_DEVICES` (GPU) or `ASCEND_RT_VISIBLE_DEVICES` (NPU) to choose computing devices.
+
 ## Examples
 
-### LoRA Fine-Tuning on A Single GPU
+### LoRA Fine-Tuning
 
 #### (Continuous) Pre-Training
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_pretrain.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_pretrain.yaml
 ```
 
 #### Supervised Fine-Tuning
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_sft.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
 ```
 
 #### Multimodal Supervised Fine-Tuning
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llava1_5_lora_sft.yaml
+llamafactory-cli train examples/train_lora/llava1_5_lora_sft.yaml
 ```
 
 #### Reward Modeling
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_reward.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_reward.yaml
 ```
 
 #### PPO Training
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_ppo.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_ppo.yaml
 ```
 
-#### DPO Training
+#### DPO/ORPO/SimPO Training
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_dpo.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_dpo.yaml
 ```
 
-#### ORPO Training
+#### KTO Training
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_orpo.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_kto.yaml
 ```
 
 #### Preprocess Dataset
@@ -64,93 +64,79 @@ CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lo
 It is useful for large dataset, use `tokenized_path` in config to load the preprocessed dataset.
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_preprocess.yaml
+llamafactory-cli train examples/train_lora/llama3_preprocess.yaml
 ```
 
 #### Evaluating on MMLU/CMMLU/C-Eval Benchmarks
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli eval examples/lora_single_gpu/llama3_lora_eval.yaml
+llamafactory-cli eval examples/train_lora/llama3_lora_eval.yaml
 ```
 
 #### Batch Predicting and Computing BLEU and ROUGE Scores
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_predict.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_predict.yaml
 ```
 
-### QLoRA Fine-Tuning on a Single GPU
+#### Supervised Fine-Tuning on Multiple Nodes
-
-#### Supervised Fine-Tuning with 4/8-bit Bitsandbytes Quantization (Recommended)
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/qlora_single_gpu/llama3_lora_sft_bitsandbytes.yaml
+FORCE_TORCHRUN=1 NNODES=2 RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
-```
+FORCE_TORCHRUN=1 NNODES=2 RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
-
-#### Supervised Fine-Tuning with 4/8-bit GPTQ Quantization
-
-```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/qlora_single_gpu/llama3_lora_sft_gptq.yaml
-```
-
-#### Supervised Fine-Tuning with 4-bit AWQ Quantization
-
-```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/qlora_single_gpu/llama3_lora_sft_awq.yaml
-```
-
-#### Supervised Fine-Tuning with 2-bit AQLM Quantization
-
-```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/qlora_single_gpu/llama3_lora_sft_aqlm.yaml
-```
-
-### LoRA Fine-Tuning on Multiple GPUs
-
-#### Supervised Fine-Tuning with Accelerate on Single Node
-
-```bash
-bash examples/lora_multi_gpu/single_node.sh
-```
-
-#### Supervised Fine-Tuning with Accelerate on Multiple Nodes
-
-```bash
-bash examples/lora_multi_gpu/multi_node.sh
 ```
 
 #### Supervised Fine-Tuning with DeepSpeed ZeRO-3 (Weight Sharding)
 
 ```bash
-bash examples/lora_multi_gpu/ds_zero3.sh
+FORCE_TORCHRUN=1 llamafactory-cli train examples/train_lora/llama3_lora_sft_ds3.yaml
 ```
 
-### LoRA Fine-Tuning on Multiple NPUs
+### QLoRA Fine-Tuning
 
-#### Supervised Fine-Tuning with DeepSpeed ZeRO-0
+#### Supervised Fine-Tuning with 4/8-bit Bitsandbytes/HQQ/EETQ Quantization (Recommended)
 
 ```bash
-bash examples/lora_multi_npu/ds_zero0.sh
+llamafactory-cli train examples/train_qlora/llama3_lora_sft_otfq.yaml
 ```
 
-### Full-Parameter Fine-Tuning on Multiple GPUs
+#### Supervised Fine-Tuning with 4/8-bit GPTQ Quantization
-
-#### Supervised Fine-Tuning with Accelerate on Single Node
 
 ```bash
-bash examples/full_multi_gpu/single_node.sh
+llamafactory-cli train examples/train_qlora/llama3_lora_sft_gptq.yaml
 ```
 
-#### Supervised Fine-Tuning with Accelerate on Multiple Nodes
+#### Supervised Fine-Tuning with 4-bit AWQ Quantization
 
 ```bash
-bash examples/full_multi_gpu/multi_node.sh
+llamafactory-cli train examples/train_qlora/llama3_lora_sft_awq.yaml
+```
+
+#### Supervised Fine-Tuning with 2-bit AQLM Quantization
+
+```bash
+llamafactory-cli train examples/train_qlora/llama3_lora_sft_aqlm.yaml
+```
+
+### Full-Parameter Fine-Tuning
+
+#### Supervised Fine-Tuning on Single Node
+
+```bash
+FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/llama3_full_sft_ds3.yaml
+```
+
+#### Supervised Fine-Tuning on Multiple Nodes
+
+```bash
+FORCE_TORCHRUN=1 NNODES=2 RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft_ds3.yaml
+FORCE_TORCHRUN=1 NNODES=2 RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft_ds3.yaml
 ```
 
 #### Batch Predicting and Computing BLEU and ROUGE Scores
 
 ```bash
-bash examples/full_multi_gpu/predict.sh
+llamafactory-cli train examples/train_full/llama3_full_predict.yaml
 ```
 
 ### Merging LoRA Adapters and Quantization
@@ -160,13 +146,13 @@ bash examples/full_multi_gpu/predict.sh
 Note: DO NOT use quantized model or `quantization_bit` when merging LoRA adapters.
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
+llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
 ```
 
 #### Quantizing Model using AutoGPTQ
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli export examples/merge_lora/llama3_gptq.yaml
+llamafactory-cli export examples/merge_lora/llama3_gptq.yaml
 ```
 
 ### Inferring LoRA Fine-Tuned Models
@@ -174,19 +160,19 @@ CUDA_VISIBLE_DEVICES=0 llamafactory-cli export examples/merge_lora/llama3_gptq.y
 #### Use CLI
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli chat examples/merge_lora/llama3_lora_sft.yaml
+llamafactory-cli chat examples/inference/llama3_lora_sft.yaml
 ```
 
 #### Use Web UI
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli webchat examples/merge_lora/llama3_lora_sft.yaml
+llamafactory-cli webchat examples/inference/llama3_lora_sft.yaml
 ```
 
 #### Launch OpenAI-style API
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli api examples/merge_lora/llama3_lora_sft.yaml
+llamafactory-cli api examples/inference/llama3_lora_sft.yaml
 ```
 
 ### Extras
@@ -194,36 +180,42 @@ CUDA_VISIBLE_DEVICES=0 llamafactory-cli api examples/merge_lora/llama3_lora_sft.
 #### Full-Parameter Fine-Tuning using GaLore
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/extras/galore/llama3_full_sft.yaml
+llamafactory-cli train examples/extras/galore/llama3_full_sft.yaml
 ```
 
 #### Full-Parameter Fine-Tuning using BAdam
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/extras/badam/llama3_full_sft.yaml
+llamafactory-cli train examples/extras/badam/llama3_full_sft.yaml
 ```
 
 #### LoRA+ Fine-Tuning
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/extras/loraplus/llama3_lora_sft.yaml
+llamafactory-cli train examples/extras/loraplus/llama3_lora_sft.yaml
+```
+
+#### PiSSA Fine-Tuning
+
+```bash
+llamafactory-cli train examples/extras/pissa/llama3_lora_sft.yaml
 ```
 
 #### Mixture-of-Depths Fine-Tuning
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/extras/mod/llama3_full_sft.yaml
+llamafactory-cli train examples/extras/mod/llama3_full_sft.yaml
 ```
 
 #### LLaMA-Pro Fine-Tuning
 
 ```bash
 bash examples/extras/llama_pro/expand.sh
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/extras/llama_pro/llama3_freeze_sft.yaml
+llamafactory-cli train examples/extras/llama_pro/llama3_freeze_sft.yaml
 ```
 
 #### FSDP+QLoRA Fine-Tuning
 
 ```bash
-bash examples/extras/fsdp_qlora/single_node.sh
+bash examples/extras/fsdp_qlora/train.sh
 ```

examples/README_zh.md

@@ -4,59 +4,59 @@
 
 ## 目录
 
-- [单 GPU LoRA 微调](#单-gpu-lora-微调)
+- [LoRA 微调](#lora-微调)
-- [单 GPU QLoRA 微调](#单-gpu-qlora-微调)
+- [QLoRA 微调](#qlora-微调)
-- [多 GPU LoRA 微调](#多-gpu-lora-微调)
+- [全参数微调](#全参数微调)
-- [多 NPU LoRA 微调](#多-npu-lora-微调)
-- [多 GPU 全参数微调](#多-gpu-全参数微调)
 - [合并 LoRA 适配器与模型量化](#合并-lora-适配器与模型量化)
 - [推理 LoRA 模型](#推理-lora-模型)
 - [杂项](#杂项)
 
+使用 `CUDA_VISIBLE_DEVICES`（GPU）或 `ASCEND_RT_VISIBLE_DEVICES`（NPU）选择计算设备。
+
 ## 示例
 
-### 单 GPU LoRA 微调
+### LoRA 微调
 
 #### （增量）预训练
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_pretrain.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_pretrain.yaml
 ```
 
 #### 指令监督微调
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_sft.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
 ```
 
 #### 多模态指令监督微调
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llava1_5_lora_sft.yaml
+llamafactory-cli train examples/train_lora/llava1_5_lora_sft.yaml
 ```
 
 #### 奖励模型训练
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_reward.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_reward.yaml
 ```
 
 #### PPO 训练
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_ppo.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_ppo.yaml
 ```
 
-#### DPO 训练
+#### DPO/ORPO/SimPO 训练
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_dpo.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_dpo.yaml
 ```
 
-#### ORPO 训练
+#### KTO 训练
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_orpo.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_kto.yaml
 ```
 
 #### 预处理数据集
@@ -64,93 +64,79 @@ CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lo
 对于大数据集有帮助，在配置中使用 `tokenized_path` 以加载预处理后的数据集。
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_preprocess.yaml
+llamafactory-cli train examples/train_lora/llama3_preprocess.yaml
 ```
 
 #### 在 MMLU/CMMLU/C-Eval 上评估
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli eval examples/lora_single_gpu/llama3_lora_eval.yaml
+llamafactory-cli eval examples/train_lora/llama3_lora_eval.yaml
 ```
 
 #### 批量预测并计算 BLEU 和 ROUGE 分数
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/lora_single_gpu/llama3_lora_predict.yaml
+llamafactory-cli train examples/train_lora/llama3_lora_predict.yaml
 ```
 
-### 单 GPU QLoRA 微调
+#### 多机指令监督微调
-
-#### 基于 4/8 比特 Bitsandbytes 量化进行指令监督微调（推荐）
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/qlora_single_gpu/llama3_lora_sft_bitsandbytes.yaml
+FORCE_TORCHRUN=1 NNODES=2 RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
-```
+FORCE_TORCHRUN=1 NNODES=2 RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
-
-#### 基于 4/8 比特 GPTQ 量化进行指令监督微调
-
-```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/qlora_single_gpu/llama3_lora_sft_gptq.yaml
-```
-
-#### 基于 4 比特 AWQ 量化进行指令监督微调
-
-```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/qlora_single_gpu/llama3_lora_sft_awq.yaml
-```
-
-#### 基于 2 比特 AQLM 量化进行指令监督微调
-
-```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/qlora_single_gpu/llama3_lora_sft_aqlm.yaml
-```
-
-### 多 GPU LoRA 微调
-
-#### 使用 Accelerate 进行单节点训练
-
-```bash
-bash examples/lora_multi_gpu/single_node.sh
-```
-
-#### 使用 Accelerate 进行多节点训练
-
-```bash
-bash examples/lora_multi_gpu/multi_node.sh
 ```
 
 #### 使用 DeepSpeed ZeRO-3 平均分配显存
 
 ```bash
-bash examples/lora_multi_gpu/ds_zero3.sh
+FORCE_TORCHRUN=1 llamafactory-cli train examples/train_lora/llama3_lora_sft_ds3.yaml
 ```
 
-### 多 NPU LoRA 微调
+### QLoRA 微调
 
-#### 使用 DeepSpeed ZeRO-0 训练
+#### 基于 4/8 比特 Bitsandbytes/HQQ/EETQ 量化进行指令监督微调（推荐）
 
 ```bash
-bash examples/lora_multi_npu/ds_zero0.sh
+llamafactory-cli train examples/train_qlora/llama3_lora_sft_otfq.yaml
 ```
 
-### 多 GPU 全参数微调
+#### 基于 4/8 比特 GPTQ 量化进行指令监督微调
-
-#### 使用 DeepSpeed 进行单节点训练
 
 ```bash
-bash examples/full_multi_gpu/single_node.sh
+llamafactory-cli train examples/train_qlora/llama3_lora_sft_gptq.yaml
 ```
 
-#### 使用 DeepSpeed 进行多节点训练
+#### 基于 4 比特 AWQ 量化进行指令监督微调
 
 ```bash
-bash examples/full_multi_gpu/multi_node.sh
+llamafactory-cli train examples/train_qlora/llama3_lora_sft_awq.yaml
+```
+
+#### 基于 2 比特 AQLM 量化进行指令监督微调
+
+```bash
+llamafactory-cli train examples/train_qlora/llama3_lora_sft_aqlm.yaml
+```
+
+### 全参数微调
+
+#### 在单机上进行指令监督微调
+
+```bash
+FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/llama3_full_sft_ds3.yaml
+```
+
+#### 在多机上进行指令监督微调
+
+```bash
+FORCE_TORCHRUN=1 NNODES=2 RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft_ds3.yaml
+FORCE_TORCHRUN=1 NNODES=2 RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft_ds3.yaml
 ```
 
 #### 批量预测并计算 BLEU 和 ROUGE 分数
 
 ```bash
-bash examples/full_multi_gpu/predict.sh
+llamafactory-cli train examples/train_full/llama3_full_predict.yaml
 ```
 
 ### 合并 LoRA 适配器与模型量化
@@ -160,13 +146,13 @@ bash examples/full_multi_gpu/predict.sh
 注：请勿使用量化后的模型或 `quantization_bit` 参数来合并 LoRA 适配器。
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
+llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
 ```
 
 #### 使用 AutoGPTQ 量化模型
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli export examples/merge_lora/llama3_gptq.yaml
+llamafactory-cli export examples/merge_lora/llama3_gptq.yaml
 ```
 
 ### 推理 LoRA 模型
@@ -174,19 +160,19 @@ CUDA_VISIBLE_DEVICES=0 llamafactory-cli export examples/merge_lora/llama3_gptq.y
 #### 使用命令行接口
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli chat examples/merge_lora/llama3_lora_sft.yaml
+llamafactory-cli chat examples/inference/llama3_lora_sft.yaml
 ```
 
 #### 使用浏览器界面
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli webchat examples/merge_lora/llama3_lora_sft.yaml
+llamafactory-cli webchat examples/inference/llama3_lora_sft.yaml
 ```
 
 #### 启动 OpenAI 风格 API
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli api examples/merge_lora/llama3_lora_sft.yaml
+llamafactory-cli api examples/inference/llama3_lora_sft.yaml
 ```
 
 ### 杂项
@@ -194,36 +180,42 @@ CUDA_VISIBLE_DEVICES=0 llamafactory-cli api examples/merge_lora/llama3_lora_sft.
 #### 使用 GaLore 进行全参数训练
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/extras/galore/llama3_full_sft.yaml
+llamafactory-cli train examples/extras/galore/llama3_full_sft.yaml
 ```
 
 #### 使用 BAdam 进行全参数训练
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/extras/badam/llama3_full_sft.yaml
+llamafactory-cli train examples/extras/badam/llama3_full_sft.yaml
 ```
 
 #### LoRA+ 微调
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/extras/loraplus/llama3_lora_sft.yaml
+llamafactory-cli train examples/extras/loraplus/llama3_lora_sft.yaml
+```
+
+#### PiSSA 微调
+
+```bash
+llamafactory-cli train examples/extras/pissa/llama3_lora_sft.yaml
 ```
 
 #### 深度混合微调
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/extras/mod/llama3_full_sft.yaml
+llamafactory-cli train examples/extras/mod/llama3_full_sft.yaml
 ```
 
 #### LLaMA-Pro 微调
 
 ```bash
 bash examples/extras/llama_pro/expand.sh
-CUDA_VISIBLE_DEVICES=0 llamafactory-cli train examples/extras/llama_pro/llama3_freeze_sft.yaml
+llamafactory-cli train examples/extras/llama_pro/llama3_freeze_sft.yaml
 ```
 
 #### FSDP+QLoRA 微调
 
 ```bash
-bash examples/extras/fsdp_qlora/single_node.sh
+bash examples/extras/fsdp_qlora/train.sh
 ```

examples/accelerate/fsdp_config.yaml

@@ -5,16 +5,16 @@ downcast_bf16: 'no'
 fsdp_config:
   fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
   fsdp_backward_prefetch: BACKWARD_PRE
-  fsdp_cpu_ram_efficient_loading: true
   fsdp_forward_prefetch: false
-  fsdp_offload_params: true
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_offload_params: true # offload may affect training speed
   fsdp_sharding_strategy: FULL_SHARD
   fsdp_state_dict_type: FULL_STATE_DICT
   fsdp_sync_module_states: true
-  fsdp_use_orig_params: false
+  fsdp_use_orig_params: true
 machine_rank: 0
 main_training_function: main
-mixed_precision: fp16
+mixed_precision: fp16 # or bf16
 num_machines: 1 # the number of nodes
 num_processes: 2 # the number of GPUs in all nodes
 rdzv_backend: static

examples/accelerate/master_config.yaml

@@ -1,18 +0,0 @@
-compute_environment: LOCAL_MACHINE
-debug: false
-distributed_type: MULTI_GPU
-downcast_bf16: 'no'
-gpu_ids: all
-machine_rank: 0
-main_process_ip: 192.168.0.1
-main_process_port: 29555
-main_training_function: main
-mixed_precision: fp16
-num_machines: 2 # the number of nodes
-num_processes: 8 # the number of GPUs in all nodes
-rdzv_backend: static
-same_network: true
-tpu_env: []
-tpu_use_cluster: false
-tpu_use_sudo: false
-use_cpu: false

examples/accelerate/single_config.yaml

@@ -1,16 +0,0 @@
-compute_environment: LOCAL_MACHINE
-debug: false
-distributed_type: MULTI_GPU
-downcast_bf16: 'no'
-gpu_ids: all
-machine_rank: 0
-main_training_function: main
-mixed_precision: fp16
-num_machines: 1 # the number of nodes
-num_processes: 4 # the number of GPUs in all nodes
-rdzv_backend: static
-same_network: true
-tpu_env: []
-tpu_use_cluster: false
-tpu_use_sudo: false
-use_cpu: false

examples/accelerate/slave_config.yaml

@@ -1,18 +0,0 @@
-compute_environment: LOCAL_MACHINE
-debug: false
-distributed_type: MULTI_GPU
-downcast_bf16: 'no'
-gpu_ids: all
-machine_rank: 1
-main_process_ip: 192.168.0.1
-main_process_port: 29555
-main_training_function: main
-mixed_precision: fp16
-num_machines: 2 # the number of nodes
-num_processes: 8 # the number of GPUs in all nodes
-rdzv_backend: static
-same_network: true
-tpu_env: []
-tpu_use_cluster: false
-tpu_use_sudo: false
-use_cpu: false

examples/extras/badam/llama3_full_sft.yaml

@@ -1,41 +1,41 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: full
 use_badam: true
-badam_switch_mode: descending
+badam_mode: layer
+badam_switch_mode: ascending
 badam_switch_interval: 50
 badam_verbose: 2
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/full/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-pure_bf16: true
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/extras/badam/llama3_full_sft_ds3.yaml

@@ -0,0 +1,42 @@
+### model
+model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
+
+### method
+stage: sft
+do_train: true
+finetuning_type: full
+use_badam: true
+badam_mode: layer
+badam_switch_mode: ascending
+badam_switch_interval: 50
+badam_verbose: 2
+deepspeed: examples/deepspeed/ds_z3_config.json
+
+### dataset
+dataset: identity,alpaca_en_demo
+template: llama3
+cutoff_len: 1024
+max_samples: 1000
+overwrite_cache: true
+preprocessing_num_workers: 16
+
+### output
+output_dir: saves/llama3-8b/full/sft
+logging_steps: 10
+save_steps: 500
+plot_loss: true
+overwrite_output_dir: true
+
+### train
+per_device_train_batch_size: 1
+gradient_accumulation_steps: 8
+learning_rate: 1.0e-4
+num_train_epochs: 3.0
+lr_scheduler_type: cosine
+warmup_ratio: 0.1
+
+### eval
+val_size: 0.1
+per_device_eval_batch_size: 1
+eval_strategy: steps
+eval_steps: 500

examples/extras/fsdp_qlora/llama3_lora_sft.yaml

@@ -1,42 +1,40 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 quantization_bit: 4
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 
-# ddp
+### dataset
-ddp_timeout: 180000000
+dataset: identity,alpaca_en_demo
-
-# dataset
-dataset: identity,alpaca_gpt4_en
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/extras/fsdp_qlora/train.sh

@@ -1,10 +1,6 @@
 #!/bin/bash
 # DO NOT use GPTQ/AWQ model in FSDP+QLoRA
 
-pip install "transformers>=4.39.1"
-pip install "accelerate>=0.28.0"
-pip install "bitsandbytes>=0.43.0"
-
 CUDA_VISIBLE_DEVICES=0,1 accelerate launch \
     --config_file examples/accelerate/fsdp_config.yaml \
     src/train.py examples/extras/fsdp_qlora/llama3_lora_sft.yaml

examples/extras/galore/llama3_full_sft.yaml

@@ -1,7 +1,7 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: full
@@ -11,32 +11,32 @@ galore_target: mlp,self_attn
 galore_rank: 128
 galore_scale: 2.0
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/full/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 1
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
 pure_bf16: true
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/extras/llama_pro/llama3_freeze_sft.yaml

@@ -1,7 +1,7 @@
-# model
+### model
 model_name_or_path: models/llama3-8b-instruct-pro
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: freeze
@@ -9,32 +9,33 @@ freeze_trainable_layers: 8
 freeze_trainable_modules: all
 use_llama_pro: true
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b-instruct-pro/freeze/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/extras/loraplus/llama3_lora_sft.yaml

@@ -1,39 +1,40 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 loraplus_lr_ratio: 16.0
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/extras/mod/llama3_full_sft.yaml

@@ -1,39 +1,40 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: full
 mixture_of_depths: convert
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b-mod/full/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
 optim: paged_adamw_8bit
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
 pure_bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/extras/pissa/llama3_lora_sft.yaml

@@ -1,39 +1,42 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
-quantization_bit: 4
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
+pissa_init: true
+pissa_iter: 16
+pissa_convert: true
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/full_multi_gpu/multi_node.sh

@@ -1,15 +0,0 @@
-#!/bin/bash
-
-NPROC_PER_NODE=4
-NNODES=2
-RANK=0
-MASTER_ADDR=192.168.0.1
-MASTER_PORT=29500
-
-CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun \
-    --nproc_per_node $NPROC_PER_NODE \
-    --nnodes $NNODES \
-    --node_rank $RANK \
-    --master_addr $MASTER_ADDR \
-    --master_port $MASTER_PORT \
-    src/train.py examples/full_multi_gpu/llama3_full_sft.yaml

examples/full_multi_gpu/predict.sh

@@ -1,5 +0,0 @@
-#!/bin/bash
-
-CUDA_VISIBLE_DEVICES=0,1,2,3 accelerate launch \
-    --config_file examples/accelerate/single_config.yaml \
-    src/train.py examples/full_multi_gpu/llama3_full_predict.yaml

examples/full_multi_gpu/single_node.sh

@@ -1,15 +0,0 @@
-#!/bin/bash
-
-NPROC_PER_NODE=4
-NNODES=1
-RANK=0
-MASTER_ADDR=127.0.0.1
-MASTER_PORT=29500
-
-CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun \
-    --nproc_per_node $NPROC_PER_NODE \
-    --nnodes $NNODES \
-    --node_rank $RANK \
-    --master_addr $MASTER_ADDR \
-    --master_port $MASTER_PORT \
-    src/train.py examples/full_multi_gpu/llama3_full_sft.yaml

examples/inference/llava1_5.yaml

@@ -0,0 +1,3 @@
+model_name_or_path: llava-hf/llava-1.5-7b-hf
+template: vicuna
+visual_inputs: true

examples/lora_multi_gpu/ds_zero3.sh

@@ -1,15 +0,0 @@
-#!/bin/bash
-
-NPROC_PER_NODE=4
-NNODES=1
-RANK=0
-MASTER_ADDR=127.0.0.1
-MASTER_PORT=29500
-
-CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun \
-    --nproc_per_node $NPROC_PER_NODE \
-    --nnodes $NNODES \
-    --node_rank $RANK \
-    --master_addr $MASTER_ADDR \
-    --master_port $MASTER_PORT \
-    src/train.py examples/lora_multi_gpu/llama3_lora_sft_ds.yaml

examples/lora_multi_gpu/multi_node.sh

@@ -1,6 +0,0 @@
-#!/bin/bash
-# also launch it on slave machine using slave_config.yaml
-
-CUDA_VISIBLE_DEVICES=0,1,2,3 accelerate launch \
-    --config_file examples/accelerate/master_config.yaml \
-    src/train.py examples/lora_multi_gpu/llama3_lora_sft.yaml

examples/lora_multi_gpu/single_node.sh

@@ -1,5 +0,0 @@
-#!/bin/bash
-
-CUDA_VISIBLE_DEVICES=0,1,2,3 accelerate launch \
-    --config_file examples/accelerate/single_config.yaml \
-    src/train.py examples/lora_multi_gpu/llama3_lora_sft.yaml

examples/lora_multi_npu/ds_zero0.sh

@@ -1,15 +0,0 @@
-#!/bin/bash
-
-NPROC_PER_NODE=4
-NNODES=1
-RANK=0
-MASTER_ADDR=127.0.0.1
-MASTER_PORT=29500
-
-ASCEND_RT_VISIBLE_DEVICES=0,1,2,3 torchrun \
-    --nproc_per_node $NPROC_PER_NODE \
-    --nnodes $NNODES \
-    --node_rank $RANK \
-    --master_addr $MASTER_ADDR \
-    --master_port $MASTER_PORT \
-    src/train.py examples/lora_multi_npu/llama3_lora_sft_ds.yaml

examples/merge_lora/llama3_gptq.yaml

@@ -1,8 +1,8 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 template: llama3
 
-# export
+### export
 export_dir: models/llama3_gptq
 export_quantization_bit: 4
 export_quantization_dataset: data/c4_demo.json

examples/merge_lora/llama3_lora_sft.yaml

@@ -1,12 +1,12 @@
-# Note: DO NOT use quantized model or quantization_bit when merging lora adapters
+### Note: DO NOT use quantized model or quantization_bit when merging lora adapters
 
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 adapter_name_or_path: saves/llama3-8b/lora/sft
 template: llama3
 finetuning_type: lora
 
-# export
+### export
 export_dir: models/llama3_lora_sft
 export_size: 2
 export_device: cpu

examples/train_full/llama3_full_predict.yaml

@@ -1,23 +1,23 @@
-# model
+### model
 model_name_or_path: saves/llama3-8b/full/sft
 
-# method
+### method
 stage: sft
 do_predict: true
 finetuning_type: full
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 50
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/full/predict
 overwrite_output_dir: true
 
-# eval
+### eval
 per_device_eval_batch_size: 1
 predict_with_generate: true

examples/train_full/llama3_full_sft_ds3.yaml

@@ -1,41 +1,39 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: full
-
-# ddp
-ddp_timeout: 180000000
 deepspeed: examples/deepspeed/ds_z3_config.json
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/full/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 2
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_lora/llama3_lora_dpo.yaml

@@ -1,39 +1,41 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: dpo
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
-dpo_ftx: 1.0
+pref_beta: 0.1
+pref_loss: sigmoid  # choices: [sigmoid (dpo), orpo, simpo]
 
-# dataset
+### dataset
-dataset: orca_rlhf
+dataset: dpo_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/dpo
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.00001
+learning_rate: 5.0e-6
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_lora/llama3_lora_eval.yaml

@@ -1,19 +1,18 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 adapter_name_or_path: saves/llama3-8b/lora/sft
 
-# method
+### method
 finetuning_type: lora
 
-# dataset
+### dataset
-task: mmlu
+task: mmlu_test  # choices: [mmlu_test, ceval_validation, cmmlu_test]
-split: test
 template: fewshot
 lang: en
 n_shot: 5
 
-# output
+### output
 save_dir: saves/llama3-8b/lora/eval
 
-# eval
+### eval
 batch_size: 4

examples/train_lora/llama3_lora_kto.yaml

@@ -1,38 +1,40 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
-stage: orpo
+stage: kto
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
+pref_beta: 0.1
 
-# dataset
+### dataset
-dataset: orca_rlhf
+dataset: kto_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
-output_dir: saves/llama3-8b/lora/orpo
+output_dir: saves/llama3-8b/lora/kto
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.00001
+learning_rate: 5.0e-6
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_lora/llama3_lora_ppo.yaml

@@ -1,38 +1,39 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 reward_model: saves/llama3-8b/lora/reward
 
-# method
+### method
 stage: ppo
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/ppo
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.00001
+learning_rate: 1.0e-5
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# generate
+### generate
 max_new_tokens: 512
 top_k: 0
 top_p: 0.9

examples/train_lora/llama3_lora_predict.yaml

@@ -1,24 +1,25 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 adapter_name_or_path: saves/llama3-8b/lora/sft
 
-# method
+### method
 stage: sft
 do_predict: true
 finetuning_type: lora
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+eval_dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 50
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/predict
 overwrite_output_dir: true
 
-# eval
+### eval
 per_device_eval_batch_size: 1
 predict_with_generate: true
+ddp_timeout: 180000000

examples/train_lora/llama3_lora_pretrain.yaml

@@ -1,37 +1,38 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: pt
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 
-# dataset
+### dataset
 dataset: c4_demo
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
-output_dir: saves/llama3-8b/lora/sft
+output_dir: saves/llama3-8b/lora/pretrain
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_lora/llama3_lora_reward.yaml

@@ -1,38 +1,39 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: rm
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 
-# dataset
+### dataset
-dataset: orca_rlhf
+dataset: dpo_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/reward
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.00001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_lora/llama3_lora_sft.yaml

@@ -1,38 +1,39 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_lora/llama3_lora_sft_ds0.yaml

@@ -1,42 +1,40 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
-
-# ddp
-ddp_timeout: 180000000
 deepspeed: examples/deepspeed/ds_z0_config.json
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 2
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_lora/llama3_lora_sft_ds3.yaml

@@ -1,42 +1,40 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
-
-# ddp
-ddp_timeout: 180000000
 deepspeed: examples/deepspeed/ds_z3_config.json
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 2
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_lora/llama3_preprocess.yaml

@@ -1,14 +1,14 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
@@ -16,6 +16,6 @@ overwrite_cache: true
 preprocessing_num_workers: 16
 tokenized_path: saves/llama3-8b/dataset/sft
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/sft
 overwrite_output_dir: true

examples/train_lora/llava1_5_lora_sft.yaml

@@ -1,14 +1,14 @@
-# model
+### model
 model_name_or_path: llava-hf/llava-1.5-7b-hf
 visual_inputs: true
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 
-# dataset
+### dataset
 dataset: mllm_demo
 template: vicuna
 cutoff_len: 1024
@@ -16,24 +16,25 @@ max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llava1_5-7b/lora/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_qlora/llama3_lora_sft_aqlm.yaml

@@ -1,38 +1,39 @@
-# model
+### model
 model_name_or_path: ISTA-DASLab/Meta-Llama-3-8B-Instruct-AQLM-2Bit-1x16
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_qlora/llama3_lora_sft_awq.yaml

@@ -1,38 +1,39 @@
-# model
+### model
 model_name_or_path: TechxGenus/Meta-Llama-3-8B-Instruct-AWQ
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_qlora/llama3_lora_sft_gptq.yaml

@@ -1,38 +1,39 @@
-# model
+### model
 model_name_or_path: TechxGenus/Meta-Llama-3-8B-Instruct-GPTQ
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 
-# dataset
+### dataset
-dataset: identity,alpaca_gpt4_en
+dataset: identity,alpaca_en_demo
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
 gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

examples/train_qlora/llama3_lora_sft_otfq.yaml

@@ -1,41 +1,41 @@
-# model
+### model
 model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
+quantization_bit: 4
+quantization_method: bitsandbytes  # choices: [bitsandbytes (4/8), hqq (2/3/4/5/6/8), eetq (8)]
 
-# method
+### method
 stage: sft
 do_train: true
 finetuning_type: lora
-lora_target: q_proj,v_proj
+lora_target: all
 
-# ddp
+### dataset
-ddp_timeout: 180000000
+dataset: identity,alpaca_en_demo
-
-# dataset
-dataset: identity,alpaca_gpt4_en
 template: llama3
 cutoff_len: 1024
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16
 
-# output
+### output
 output_dir: saves/llama3-8b/lora/sft
 logging_steps: 10
 save_steps: 500
 plot_loss: true
 overwrite_output_dir: true
 
-# train
+### train
 per_device_train_batch_size: 1
-gradient_accumulation_steps: 2
+gradient_accumulation_steps: 8
-learning_rate: 0.0001
+learning_rate: 1.0e-4
 num_train_epochs: 3.0
 lr_scheduler_type: cosine
-warmup_steps: 0.1
+warmup_ratio: 0.1
-fp16: true
+bf16: true
+ddp_timeout: 180000000
 
-# eval
+### eval
 val_size: 0.1
 per_device_eval_batch_size: 1
-evaluation_strategy: steps
+eval_strategy: steps
 eval_steps: 500

pyproject.toml

@@ -13,7 +13,7 @@ select = ["C", "E", "F", "I", "W"]
 
 [tool.ruff.lint.isort]
 lines-after-imports = 2
-known-first-party = ["llmtuner"]
+known-first-party = ["llamafactory"]
 known-third-party = [
     "accelerate",
     "datasets",

requirements.txt

@@ -1,12 +1,14 @@
-transformers>=4.37.2
+transformers>=4.41.2
-datasets>=2.14.3
+datasets>=2.16.0
-accelerate>=0.27.2
+accelerate>=0.30.1
-peft>=0.10.0
+peft>=0.11.1
-trl>=0.8.1
+trl>=0.8.6
 gradio>=4.0.0
+pandas>=2.0.0
 scipy
 einops
 sentencepiece
+tiktoken
 protobuf
 uvicorn
 pydantic
@@ -16,3 +18,4 @@ matplotlib>=3.7.0
 fire
 packaging
 pyyaml
+numpy<2.0.0

scripts/cal_flops.py

@@ -1,14 +1,27 @@
 # coding=utf-8
-# Calculates the flops of pre-trained models.
+# Copyright 2024 Microsoft Corporation and the LlamaFactory team.
-# Usage: python cal_flops.py --model_name_or_path path_to_model --batch_size 1 --seq_length 512
+#
-# Inspired by: https://www.deepspeed.ai/tutorials/flops-profiler/
+# This code is inspired by the Microsoft's DeepSpeed library.
+# https://www.deepspeed.ai/tutorials/flops-profiler/
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
 
 import fire
 import torch
 from deepspeed.accelerator import get_accelerator  # type: ignore
 from deepspeed.profiling.flops_profiler import get_model_profile  # type: ignore
 
-from llmtuner.chat import ChatModel
+from llamafactory.chat import ChatModel
 
 
 def calculate_flops(
@@ -17,6 +30,10 @@ def calculate_flops(
     seq_length: int = 256,
     flash_attn: str = "auto",
 ):
+    r"""
+    Calculates the flops of pre-trained models.
+    Usage: python cal_flops.py --model_name_or_path path_to_model --batch_size 1 --seq_length 512
+    """
     with get_accelerator().device(0):
         chat_model = ChatModel(dict(model_name_or_path=model_name_or_path, template="empty", flash_attn=flash_attn))
         fake_input = torch.ones((batch_size, seq_length), dtype=torch.long, device=chat_model.model.device)

scripts/cal_lr.py

@@ -1,7 +1,20 @@
 # coding=utf-8
-# Calculates the optimal learning rate for 7B/13B models using LLaMA's hyper-parameters.
+# Copyright 2024 imoneoi and the LlamaFactory team.
-# Usage: python cal_lr.py --model_name_or_path path_to_model --dataset alpaca_en --cutoff_len 1024 --batch_size 16
+#
-# Inspired by: https://github.com/imoneoi/openchat/blob/master/ochat/training_deepspeed/train.py
+# This code is inspired by the imoneoi's OpenChat library.
+# https://github.com/imoneoi/openchat/blob/3.6.0/ochat/training_deepspeed/train.py
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
 
 import math
 from typing import Literal
@@ -12,10 +25,10 @@ from torch.utils.data import DataLoader
 from tqdm import tqdm
 from transformers import DataCollatorForLanguageModeling, DataCollatorForSeq2Seq
 
-from llmtuner.data import get_dataset
+from llamafactory.data import get_dataset
-from llmtuner.extras.constants import IGNORE_INDEX
+from llamafactory.extras.constants import IGNORE_INDEX
-from llmtuner.hparams import get_train_args
+from llamafactory.hparams import get_train_args
-from llmtuner.model import load_tokenizer
+from llamafactory.model import load_tokenizer
 
 
 BASE_LR = 3e-4  # 1.5e-4 for 30B-70B models
@@ -31,7 +44,12 @@ def calculate_lr(
     template: str = "default",
     cutoff_len: int = 1024,  # i.e. maximum input length during training
     is_mistral: bool = False,  # mistral model uses a smaller learning rate,
+    packing: bool = False,
 ):
+    r"""
+    Calculates the optimal learning rate for 7B/13B models using LLaMA's hyper-parameters.
+    Usage: python cal_lr.py --model_name_or_path path_to_model --dataset alpaca_en --cutoff_len 1024 --batch_size 16
+    """
     model_args, data_args, training_args, _, _ = get_train_args(
         dict(
             stage=stage,
@@ -40,19 +58,21 @@ def calculate_lr(
             dataset_dir=dataset_dir,
             template=template,
             cutoff_len=cutoff_len,
+            packing=packing,
             output_dir="dummy_dir",
             overwrite_cache=True,
+            do_train=True,
         )
     )
     tokenizer_module = load_tokenizer(model_args)
     tokenizer = tokenizer_module["tokenizer"]
-    trainset = get_dataset(model_args, data_args, training_args, stage, **tokenizer_module)
+    trainset = get_dataset(model_args, data_args, training_args, stage, **tokenizer_module)["train_dataset"]
     if stage == "pt":
         data_collator = DataCollatorForLanguageModeling(tokenizer=tokenizer, mlm=False)
     elif stage == "sft":
         data_collator = DataCollatorForSeq2Seq(tokenizer=tokenizer, label_pad_token_id=IGNORE_INDEX)
     else:
-        raise NotImplementedError
+        raise NotImplementedError("Stage does not supported: {}.".format(stage))
 
     dataloader = DataLoader(trainset, batch_size, shuffle=False, collate_fn=data_collator, pin_memory=True)
     valid_tokens, total_tokens = 0, 0

scripts/cal_ppl.py

@@ -1,6 +1,17 @@
 # coding=utf-8
-# Calculates the ppl on the dataset of the pre-trained models.
+# Copyright 2024 the LlamaFactory team.
-# Usage: python cal_ppl.py --model_name_or_path path_to_model --save_name ppl.json
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
 
 import json
 from dataclasses import dataclass
@@ -12,10 +23,10 @@ from torch.utils.data import DataLoader
 from tqdm import tqdm
 from transformers import DataCollatorForLanguageModeling, DataCollatorForSeq2Seq
 
-from llmtuner.data import get_dataset
+from llamafactory.data import get_dataset
-from llmtuner.extras.constants import IGNORE_INDEX
+from llamafactory.extras.constants import IGNORE_INDEX
-from llmtuner.hparams import get_train_args
+from llamafactory.hparams import get_train_args
-from llmtuner.model import load_model, load_tokenizer
+from llamafactory.model import load_model, load_tokenizer
 
 
 @dataclass
@@ -56,6 +67,10 @@ def cal_ppl(
     max_samples: Optional[int] = None,
     train_on_prompt: bool = False,
 ):
+    r"""
+    Calculates the ppl on the dataset of the pre-trained models.
+    Usage: python cal_ppl.py --model_name_or_path path_to_model --save_name ppl.json
+    """
     model_args, data_args, training_args, finetuning_args, _ = get_train_args(
         dict(
             stage=stage,
@@ -68,11 +83,12 @@ def cal_ppl(
             train_on_prompt=train_on_prompt,
             output_dir="dummy_dir",
             overwrite_cache=True,
+            do_train=True,
         )
     )
     tokenizer_module = load_tokenizer(model_args)
     tokenizer = tokenizer_module["tokenizer"]
-    trainset = get_dataset(model_args, data_args, training_args, stage, **tokenizer_module)
+    trainset = get_dataset(model_args, data_args, training_args, stage, **tokenizer_module)["train_dataset"]
     model = load_model(tokenizer, model_args, finetuning_args, is_trainable=False)
     if stage == "pt":
         data_collator = DataCollatorForLanguageModeling(tokenizer=tokenizer, mlm=False)
@@ -83,7 +99,7 @@ def cal_ppl(
             tokenizer=tokenizer, label_pad_token_id=IGNORE_INDEX, train_on_prompt=train_on_prompt
         )
     else:
-        raise NotImplementedError
+        raise NotImplementedError("Stage does not supported: {}.".format(stage))
 
     dataloader = DataLoader(trainset, batch_size, shuffle=False, collate_fn=data_collator, pin_memory=True)
     criterion = torch.nn.CrossEntropyLoss(reduction="none")

scripts/length_cdf.py

@@ -1,15 +1,26 @@
 # coding=utf-8
-# Calculates the distribution of the input lengths in the dataset.
+# Copyright 2024 the LlamaFactory team.
-# Usage: python length_cdf.py --model_name_or_path path_to_model --dataset alpaca_en --template default
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
 
 from collections import defaultdict
 
 import fire
 from tqdm import tqdm
 
-from llmtuner.data import get_dataset
+from llamafactory.data import get_dataset
-from llmtuner.hparams import get_train_args
+from llamafactory.hparams import get_train_args
-from llmtuner.model import load_tokenizer
+from llamafactory.model import load_tokenizer
 
 
 def length_cdf(
@@ -19,6 +30,10 @@ def length_cdf(
     template: str = "default",
     interval: int = 1000,
 ):
+    r"""
+    Calculates the distribution of the input lengths in the dataset.
+    Usage: python length_cdf.py --model_name_or_path path_to_model --dataset alpaca_en --template default
+    """
     model_args, data_args, training_args, _, _ = get_train_args(
         dict(
             stage="sft",
@@ -29,10 +44,11 @@ def length_cdf(
             cutoff_len=1_000_000,
             output_dir="dummy_dir",
             overwrite_cache=True,
+            do_train=True,
         )
     )
     tokenizer_module = load_tokenizer(model_args)
-    trainset = get_dataset(model_args, data_args, training_args, stage="sft", **tokenizer_module)
+    trainset = get_dataset(model_args, data_args, training_args, stage="sft", **tokenizer_module)["train_dataset"]
     total_num = len(trainset)
     length_dict = defaultdict(int)
     for sample in tqdm(trainset["input_ids"]):

scripts/llama_pro.py

@@ -1,7 +1,20 @@
 # coding=utf-8
-# Performs block expansion for LLaMA, Mistral, Qwen1.5 or Yi models.
+# Copyright 2024 Tencent Inc. and the LlamaFactory team.
-# Usage: python llama_pro.py --model_name_or_path meta-llama/Llama-2-7b-hf --output_dir llama2_pro --num_expand 8
+#
-# Inspired by: https://github.com/TencentARC/LLaMA-Pro/blob/main/scripts/block_expansion.py
+# This code is inspired by the Tencent's LLaMA-Pro library.
+# https://github.com/TencentARC/LLaMA-Pro/blob/main/scripts/block_expansion.py
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
 
 import json
 import os
@@ -37,6 +50,10 @@ def block_expansion(
     shard_size: Optional[str] = "2GB",
     save_safetensors: Optional[bool] = False,
 ):
+    r"""
+    Performs block expansion for LLaMA, Mistral, Qwen1.5 or Yi models.
+    Usage: python llama_pro.py --model_name_or_path meta-llama/Llama-2-7b-hf --output_dir llama2_pro --num_expand 8
+    """
     config: "PretrainedConfig" = AutoConfig.from_pretrained(model_name_or_path)
     num_layers = getattr(config, "num_hidden_layers")
     setattr(config, "num_hidden_layers", num_layers + num_expand)
@@ -103,11 +120,11 @@ def block_expansion(
             json.dump(index, f, indent=2, sort_keys=True)
         print("Model weights saved in {}".format(output_dir))
 
-    print("Fine-tune this model with:")
+    print("- Fine-tune this model with:")
-    print("  --model_name_or_path {} \\".format(output_dir))
+    print("model_name_or_path: {}".format(output_dir))
-    print("  --finetuning_type freeze \\")
+    print("finetuning_type: freeze")
-    print("  --freeze_trainable_layers {} \\".format(num_expand))
+    print("freeze_trainable_layers: {}".format(num_expand))
-    print("  --use_llama_pro")
+    print("use_llama_pro: true")
 
 
 if __name__ == "__main__":

scripts/llamafy_baichuan2.py

@@ -1,8 +1,17 @@
 # coding=utf-8
-# Converts the Baichuan2-7B model in the same format as LLaMA2-7B.
+# Copyright 2024 the LlamaFactory team.
-# Usage: python llamafy_baichuan2.py --input_dir input --output_dir output
+#
-# Inspired by: https://huggingface.co/fireballoon/baichuan-llama-7b/blob/main/convert_baichuan_to_llama.py
+# Licensed under the Apache License, Version 2.0 (the "License");
-# Converted model: https://huggingface.co/hiyouga/Baichuan2-7B-Base-LLaMAfied
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
 
 import json
 import os
@@ -79,6 +88,11 @@ def save_config(input_dir: str, output_dir: str):
 def llamafy_baichuan2(
     input_dir: str, output_dir: str, shard_size: Optional[str] = "2GB", save_safetensors: Optional[bool] = False
 ):
+    r"""
+    Converts the Baichuan2-7B model in the same format as LLaMA2-7B.
+    Usage: python llamafy_baichuan2.py --input_dir input --output_dir output
+    Converted model: https://huggingface.co/hiyouga/Baichuan2-7B-Base-LLaMAfied
+    """
     try:
         os.makedirs(output_dir, exist_ok=False)
     except Exception as e:

scripts/llamafy_qwen.py

@@ -1,7 +1,17 @@
 # coding=utf-8
-# Converts the Qwen models in the same format as LLaMA2.
+# Copyright 2024 the LlamaFactory team.
-# Usage: python llamafy_qwen.py --input_dir input --output_dir output
+#
-# Converted model: https://huggingface.co/hiyouga/Qwen-14B-Chat-LLaMAfied
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
 
 import json
 import os
@@ -131,6 +141,11 @@ def save_config(input_dir: str, output_dir: str, torch_dtype: str):
 def llamafy_qwen(
     input_dir: str, output_dir: str, shard_size: Optional[str] = "2GB", save_safetensors: Optional[bool] = False
 ):
+    r"""
+    Converts the Qwen models in the same format as LLaMA2.
+    Usage: python llamafy_qwen.py --input_dir input --output_dir output
+    Converted model: https://huggingface.co/hiyouga/Qwen-14B-Chat-LLaMAfied
+    """
     try:
         os.makedirs(output_dir, exist_ok=False)
     except Exception as e:

scripts/loftq_init.py

@@ -1,14 +1,25 @@
 # coding=utf-8
-# Initializes LoRA weights with LoRA-fine-tuning-aware Quantization (LoftQ)
+# Copyright 2024 HuggingFace Inc. and the LlamaFactory team.
-# Usage: python loftq_init.py --model_name_or_path path_to_model --save_dir output_dir
+#
-# Inspired by: https://github.com/huggingface/peft/blob/main/examples/loftq_finetuning/quantize_save_load.py
+# This code is based on the HuggingFace's PEFT library.
+# https://github.com/huggingface/peft/blob/v0.10.0/examples/loftq_finetuning/quantize_save_load.py
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
 
 import os
-from typing import TYPE_CHECKING, Optional
+from typing import TYPE_CHECKING
 
 import fire
-import torch
-import torch.nn as nn
 from peft import LoftQConfig, LoraConfig, TaskType, get_peft_model
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
@@ -17,65 +28,61 @@ if TYPE_CHECKING:
     from transformers import PreTrainedModel
 
 
-class Shell(nn.Module):
-    def __init__(self, weight: torch.Tensor, bias: Optional[torch.Tensor] = None):
-        super().__init__()
-        self.weight = nn.Parameter(weight, requires_grad=False)
-        if bias is not None:
-            self.bias = nn.Parameter(bias, requires_grad=False)
-
-
-def unwrap_model(model: nn.Module, pattern=".base_layer") -> None:
-    for name in {k.split(pattern)[0] for k, _ in model.named_modules() if pattern in k}:
-        parent_name = ".".join(name.split(".")[:-1])
-        child_name = name.split(".")[-1]
-        parent_module = model.get_submodule(parent_name)
-        child_module = getattr(parent_module, child_name)
-        base_layer = getattr(child_module, "base_layer")
-        weight = getattr(base_layer, "weight", None)
-        bias = getattr(base_layer, "bias", None)
-        setattr(parent_module, child_name, Shell(weight, bias))
-
-    print("Model unwrapped.")
-
-
 def quantize_loftq(
     model_name_or_path: str,
-    save_dir: str,
+    output_dir: str,
-    loftq_bits: Optional[int] = 4,
+    loftq_bits: int = 4,
-    loftq_iter: Optional[int] = 1,
+    loftq_iter: int = 4,
-    lora_alpha: Optional[int] = None,
+    lora_alpha: int = None,
-    lora_rank: Optional[int] = 16,
+    lora_rank: int = 16,
-    lora_target: Optional[str] = "q_proj,v_proj",
+    lora_dropout: float = 0,
-    save_safetensors: Optional[bool] = False,
+    lora_target: tuple = ("q_proj", "v_proj"),
+    save_safetensors: bool = True,
 ):
+    r"""
+    Initializes LoRA weights with LoRA-fine-tuning-aware Quantization (LoftQ)
+    Usage: python loftq_init.py --model_name_or_path path_to_model --output_dir output_dir
+    """
+    if isinstance(lora_target, str):
+        lora_target = [name.strip() for name in lora_target.split(",")]
+
     tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, trust_remote_code=True)
     model = AutoModelForCausalLM.from_pretrained(model_name_or_path, trust_remote_code=True, torch_dtype="auto")
+
     loftq_config = LoftQConfig(loftq_bits=loftq_bits, loftq_iter=loftq_iter)
     lora_config = LoraConfig(
         task_type=TaskType.CAUSAL_LM,
         inference_mode=True,
         r=lora_rank,
         lora_alpha=lora_alpha if lora_alpha is not None else lora_rank * 2,
-        lora_dropout=0.1,
+        lora_dropout=lora_dropout,
-        target_modules=[name.strip() for name in lora_target.split(",")],
+        target_modules=lora_target,
         init_lora_weights="loftq",
         loftq_config=loftq_config,
     )
 
     # Init LoftQ model
-    lora_model = get_peft_model(model, lora_config)
+    print("Initializing LoftQ weights, it may be take several minutes, wait patiently.")
-    base_model: "PreTrainedModel" = lora_model.get_base_model()
+    peft_model = get_peft_model(model, lora_config)
+    loftq_dir = os.path.join(output_dir, "loftq_init")
 
     # Save LoftQ model
-    setattr(lora_model.base_model.peft_config["default"], "base_model_name_or_path", save_dir)
+    setattr(peft_model.peft_config["default"], "base_model_name_or_path", output_dir)
-    setattr(lora_model.base_model.peft_config["default"], "init_lora_weights", True)
+    setattr(peft_model.peft_config["default"], "init_lora_weights", True)  # don't apply loftq again
-    lora_model.save_pretrained(os.path.join(save_dir, "adapters"), safe_serialization=save_safetensors)
+    peft_model.save_pretrained(loftq_dir, safe_serialization=save_safetensors)
+    print("Adapter weights saved in {}".format(loftq_dir))
 
     # Save base model
-    unwrap_model(base_model)
+    base_model: "PreTrainedModel" = peft_model.unload()
-    base_model.save_pretrained(save_dir, safe_serialization=save_safetensors)
+    base_model.save_pretrained(output_dir, safe_serialization=save_safetensors)
-    tokenizer.save_pretrained(save_dir)
+    tokenizer.save_pretrained(output_dir)
+    print("Model weights saved in {}".format(output_dir))
+
+    print("- Fine-tune this model with:")
+    print("model_name_or_path: {}".format(output_dir))
+    print("adapter_name_or_path: {}".format(loftq_dir))
+    print("finetuning_type: lora")
+    print("quantization_bit: {}".format(loftq_bits))
 
 
 if __name__ == "__main__":

scripts/pissa_init.py

@@ -0,0 +1,86 @@
+# coding=utf-8
+# Copyright 2024 HuggingFace Inc. and the LlamaFactory team.
+#
+# This code is based on the HuggingFace's PEFT library.
+# https://github.com/huggingface/peft/blob/v0.11.0/examples/pissa_finetuning/preprocess.py
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+from typing import TYPE_CHECKING
+
+import fire
+from peft import LoraConfig, TaskType, get_peft_model
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedModel
+
+
+def quantize_pissa(
+    model_name_or_path: str,
+    output_dir: str,
+    pissa_iter: int = 4,
+    lora_alpha: int = None,
+    lora_rank: int = 16,
+    lora_dropout: float = 0,
+    lora_target: tuple = ("q_proj", "v_proj"),
+    save_safetensors: bool = True,
+):
+    r"""
+    Initializes LoRA weights with Principal Singular values and Singular vectors Adaptation (PiSSA)
+    Usage: python pissa_init.py --model_name_or_path path_to_model --output_dir output_dir
+    """
+    if isinstance(lora_target, str):
+        lora_target = [name.strip() for name in lora_target.split(",")]
+
+    tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, trust_remote_code=True)
+    model = AutoModelForCausalLM.from_pretrained(model_name_or_path, trust_remote_code=True, torch_dtype="auto")
+
+    lora_config = LoraConfig(
+        task_type=TaskType.CAUSAL_LM,
+        r=lora_rank,
+        lora_alpha=lora_alpha if lora_alpha is not None else lora_rank * 2,
+        lora_dropout=lora_dropout,
+        target_modules=lora_target,
+        init_lora_weights="pissa" if pissa_iter == -1 else "pissa_niter_{}".format(pissa_iter),
+    )
+
+    # Init PiSSA model
+    peft_model = get_peft_model(model, lora_config)
+    pissa_dir = os.path.join(output_dir, "pissa_init")
+
+    # Save PiSSA model
+    setattr(peft_model.peft_config["default"], "init_lora_weights", True)  # don't apply pissa again
+    peft_model.save_pretrained(pissa_dir, safe_serialization=save_safetensors)
+    print("Adapter weights saved in {}".format(pissa_dir))
+
+    # Save base model
+    base_model: "PreTrainedModel" = peft_model.unload()
+    base_model.save_pretrained(output_dir, safe_serialization=save_safetensors)
+    tokenizer.save_pretrained(output_dir)
+    print("Model weights saved in {}".format(output_dir))
+
+    print("- Fine-tune this model with:")
+    print("model_name_or_path: {}".format(output_dir))
+    print("adapter_name_or_path: {}".format(pissa_dir))
+    print("finetuning_type: lora")
+    print("pissa_init: false")
+    print("pissa_convert: true")
+    print("- and optionally with:")
+    print("quantization_bit: 4")
+
+
+if __name__ == "__main__":
+    fire.Fire(quantize_pissa)

scripts/test_toolcall.py

@@ -1,3 +1,18 @@
+# coding=utf-8
+# Copyright 2024 the LlamaFactory team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 import json
 import os
 from typing import Sequence
@@ -20,7 +35,7 @@ def calculate_gpa(grades: Sequence[str], hours: Sequence[int]) -> float:
 
 def main():
     client = OpenAI(
-        api_key="0",
+        api_key="{}".format(os.environ.get("API_KEY", "0")),
         base_url="http://localhost:{}/v1".format(os.environ.get("API_PORT", 8000)),
     )
     tools = [

setup.py

@@ -1,3 +1,17 @@
+# Copyright 2024 the LlamaFactory team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 import os
 import re
 
@@ -5,7 +19,7 @@ from setuptools import find_packages, setup
 
 
 def get_version():
-    with open(os.path.join("src", "llmtuner", "cli.py"), "r", encoding="utf-8") as f:
+    with open(os.path.join("src", "llamafactory", "extras", "env.py"), "r", encoding="utf-8") as f:
         file_content = f.read()
         pattern = r"{}\W*=\W*\"([^\"]+)\"".format("VERSION")
         (version,) = re.findall(pattern, file_content)
@@ -21,24 +35,27 @@ def get_requires():
 
 extra_require = {
     "torch": ["torch>=1.13.1"],
+    "torch-npu": ["torch==2.1.0", "torch-npu==2.1.0.post3", "decorator"],
     "metrics": ["nltk", "jieba", "rouge-chinese"],
-    "deepspeed": ["deepspeed>=0.10.0,<=0.14.0"],
+    "deepspeed": ["deepspeed>=0.10.0"],
     "bitsandbytes": ["bitsandbytes>=0.39.0"],
-    "vllm": ["vllm>=0.4.0"],
+    "hqq": ["hqq"],
-    "galore": ["galore-torch"],
+    "eetq": ["eetq"],
-    "badam": ["badam"],
+    "gptq": ["optimum>=1.17.0", "auto-gptq>=0.5.0"],
-    "gptq": ["optimum>=1.16.0", "auto-gptq>=0.5.0"],
     "awq": ["autoawq"],
     "aqlm": ["aqlm[gpu]>=1.1.0"],
-    "qwen": ["tiktoken", "transformers_stream_generator"],
+    "vllm": ["vllm>=0.4.3"],
+    "galore": ["galore-torch"],
+    "badam": ["badam>=1.2.1"],
+    "qwen": ["transformers_stream_generator"],
     "modelscope": ["modelscope"],
-    "quality": ["ruff"],
+    "dev": ["ruff", "pytest"],
 }
 
 
 def main():
     setup(
-        name="llmtuner",
+        name="llamafactory",
         version=get_version(),
         author="hiyouga",
         author_email="hiyouga" "@" "buaa.edu.cn",
@@ -53,7 +70,7 @@ def main():
         python_requires=">=3.8.0",
         install_requires=get_requires(),
         extras_require=extra_require,
-        entry_points={"console_scripts": ["llamafactory-cli = llmtuner.cli:main"]},
+        entry_points={"console_scripts": ["llamafactory-cli = llamafactory.cli:main"]},
         classifiers=[
             "Development Status :: 4 - Beta",
             "Intended Audience :: Developers",

src/api.py

@@ -1,9 +1,23 @@
+# Copyright 2024 the LlamaFactory team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 import os
 
 import uvicorn
 
-from llmtuner.api.app import create_app
+from llamafactory.api.app import create_app
-from llmtuner.chat import ChatModel
+from llamafactory.chat import ChatModel
 
 
 def main():

src/llamafactory/__init__.py

@@ -0,0 +1,41 @@
+# Copyright 2024 the LlamaFactory team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+r"""
+Efficient fine-tuning of large language models.
+
+Level:
+  api, webui > chat, eval, train > data, model > hparams > extras
+
+Dependency graph:
+  main:
+    transformers>=4.41.2
+    datasets>=2.16.0
+    accelerate>=0.30.1
+    peft>=0.11.1
+    trl>=0.8.6
+  attention:
+    transformers>=4.42.4 (gemma+fa2)
+  longlora:
+    transformers>=4.41.2,<=4.42.4
+  packing:
+    transformers>=4.41.2,<=4.42.4
+  patcher:
+    transformers==4.41.2 (chatglm)
+"""
+
+from .cli import VERSION
+
+
+__version__ = VERSION

src/llamafactory/api/app.py

@@ -1,3 +1,17 @@
+# Copyright 2024 the LlamaFactory team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 import os
 from contextlib import asynccontextmanager
 from typing import Optional

src/llamafactory/api/chat.py

@@ -1,10 +1,27 @@
+# Copyright 2024 the LlamaFactory team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import base64
+import io
 import json
+import os
 import uuid
 from typing import TYPE_CHECKING, AsyncGenerator, Dict, List, Optional, Tuple
 
 from ..data import Role as DataRole
 from ..extras.logging import get_logger
-from ..extras.packages import is_fastapi_available
+from ..extras.packages import is_fastapi_available, is_pillow_available, is_requests_available
 from .common import dictify, jsonify
 from .protocol import (
     ChatCompletionMessage,
@@ -25,7 +42,17 @@ if is_fastapi_available():
     from fastapi import HTTPException, status
 
 
+if is_pillow_available():
+    from PIL import Image
+
+
+if is_requests_available():
+    import requests
+
+
 if TYPE_CHECKING:
+    from numpy.typing import NDArray
+
     from ..chat import ChatModel
     from .protocol import ChatCompletionRequest, ScoreEvaluationRequest
 
@@ -40,7 +67,9 @@ ROLE_MAPPING = {
 }
 
 
-def _process_request(request: "ChatCompletionRequest") -> Tuple[List[Dict[str, str]], str, str]:
+def _process_request(
+    request: "ChatCompletionRequest",
+) -> Tuple[List[Dict[str, str]], Optional[str], Optional[str], Optional["NDArray"]]:
     logger.info("==== request ====\n{}".format(json.dumps(dictify(request), indent=2, ensure_ascii=False)))
 
     if len(request.messages) == 0:
@@ -49,12 +78,13 @@ def _process_request(request: "ChatCompletionRequest") -> Tuple[List[Dict[str, s
     if request.messages[0].role == Role.SYSTEM:
         system = request.messages.pop(0).content
     else:
-        system = ""
+        system = None
 
     if len(request.messages) % 2 == 0:
         raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Only supports u/a/u/a/u...")
 
     input_messages = []
+    image = None
     for i, message in enumerate(request.messages):
         if i % 2 == 0 and message.role not in [Role.USER, Role.TOOL]:
             raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Invalid role")
@@ -62,10 +92,27 @@ def _process_request(request: "ChatCompletionRequest") -> Tuple[List[Dict[str, s
             raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Invalid role")
 
         if message.role == Role.ASSISTANT and isinstance(message.tool_calls, list) and len(message.tool_calls):
-            name = message.tool_calls[0].function.name
+            tool_calls = [
-            arguments = message.tool_calls[0].function.arguments
+                {"name": tool_call.function.name, "arguments": tool_call.function.arguments}
-            content = json.dumps({"name": name, "argument": arguments}, ensure_ascii=False)
+                for tool_call in message.tool_calls
+            ]
+            content = json.dumps(tool_calls, ensure_ascii=False)
             input_messages.append({"role": ROLE_MAPPING[Role.FUNCTION], "content": content})
+        elif isinstance(message.content, list):
+            for input_item in message.content:
+                if input_item.type == "text":
+                    input_messages.append({"role": ROLE_MAPPING[message.role], "content": input_item.text})
+                else:
+                    image_url = input_item.image_url.url
+                    if image_url.startswith("data:image"):  # base64 image
+                        image_data = base64.b64decode(image_url.split(",", maxsplit=1)[1])
+                        image_path = io.BytesIO(image_data)
+                    elif os.path.isfile(image_url):  # local file
+                        image_path = open(image_url, "rb")
+                    else:  # web uri
+                        image_path = requests.get(image_url, stream=True).raw
+
+                    image = Image.open(image_path).convert("RGB")
         else:
             input_messages.append({"role": ROLE_MAPPING[message.role], "content": message.content})
 
@@ -73,12 +120,12 @@ def _process_request(request: "ChatCompletionRequest") -> Tuple[List[Dict[str, s
     if isinstance(tool_list, list) and len(tool_list):
         try:
             tools = json.dumps([dictify(tool.function) for tool in tool_list], ensure_ascii=False)
-        except Exception:
+        except json.JSONDecodeError:
             raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Invalid tools")
     else:
-        tools = ""
+        tools = None
 
-    return input_messages, system, tools
+    return input_messages, system, tools, image
 
 
 def _create_stream_chat_completion_chunk(
@@ -97,11 +144,12 @@ async def create_chat_completion_response(
     request: "ChatCompletionRequest", chat_model: "ChatModel"
 ) -> "ChatCompletionResponse":
     completion_id = "chatcmpl-{}".format(uuid.uuid4().hex)
-    input_messages, system, tools = _process_request(request)
+    input_messages, system, tools, image = _process_request(request)
     responses = await chat_model.achat(
         input_messages,
         system,
         tools,
+        image,
         do_sample=request.do_sample,
         temperature=request.temperature,
         top_p=request.top_p,
@@ -114,15 +162,17 @@ async def create_chat_completion_response(
     choices = []
     for i, response in enumerate(responses):
         if tools:
-            result = chat_model.engine.template.format_tools.extract(response.response_text)
+            result = chat_model.engine.template.extract_tool(response.response_text)
         else:
             result = response.response_text
 
-        if isinstance(result, tuple):
+        if isinstance(result, list):
-            name, arguments = result
+            tool_calls = []
-            function = Function(name=name, arguments=arguments)
+            for tool in result:
-            tool_call = FunctionCall(id="call_{}".format(uuid.uuid4().hex), function=function)
+                function = Function(name=tool[0], arguments=tool[1])
-            response_message = ChatCompletionMessage(role=Role.ASSISTANT, tool_calls=[tool_call])
+                tool_calls.append(FunctionCall(id="call_{}".format(uuid.uuid4().hex), function=function))
+
+            response_message = ChatCompletionMessage(role=Role.ASSISTANT, tool_calls=tool_calls)
             finish_reason = Finish.TOOL
         else:
             response_message = ChatCompletionMessage(role=Role.ASSISTANT, content=result)
@@ -145,7 +195,7 @@ async def create_stream_chat_completion_response(
     request: "ChatCompletionRequest", chat_model: "ChatModel"
 ) -> AsyncGenerator[str, None]:
     completion_id = "chatcmpl-{}".format(uuid.uuid4().hex)
-    input_messages, system, tools = _process_request(request)
+    input_messages, system, tools, image = _process_request(request)
     if tools:
         raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Cannot stream function calls.")
 
@@ -159,6 +209,7 @@ async def create_stream_chat_completion_response(
         input_messages,
         system,
         tools,
+        image,
         do_sample=request.do_sample,
         temperature=request.temperature,
         top_p=request.top_p,

src/llamafactory/api/common.py

@@ -0,0 +1,34 @@
+# Copyright 2024 the LlamaFactory team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import json
+from typing import TYPE_CHECKING, Any, Dict
+
+
+if TYPE_CHECKING:
+    from pydantic import BaseModel
+
+
+def dictify(data: "BaseModel") -> Dict[str, Any]:
+    try:  # pydantic v2
+        return data.model_dump(exclude_unset=True)
+    except AttributeError:  # pydantic v1
+        return data.dict(exclude_unset=True)
+
+
+def jsonify(data: "BaseModel") -> str:
+    try:  # pydantic v2
+        return json.dumps(data.model_dump(exclude_unset=True), ensure_ascii=False)
+    except AttributeError:  # pydantic v1
+        return data.json(exclude_unset=True, ensure_ascii=False)

src/llamafactory/api/protocol.py

@@ -1,3 +1,17 @@
+# Copyright 2024 the LlamaFactory team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 import time
 from enum import Enum, unique
 from typing import Any, Dict, List, Optional, Union
@@ -56,9 +70,19 @@ class FunctionCall(BaseModel):
     function: Function
 
 
+class ImageURL(BaseModel):
+    url: str
+
+
+class MultimodalInputItem(BaseModel):
+    type: Literal["text", "image_url"]
+    text: Optional[str] = None
+    image_url: Optional[ImageURL] = None
+
+
 class ChatMessage(BaseModel):
     role: Role
-    content: Optional[str] = None
+    content: Optional[Union[str, List[MultimodalInputItem]]] = None
     tool_calls: Optional[List[FunctionCall]] = None
 
 
@@ -72,7 +96,7 @@ class ChatCompletionRequest(BaseModel):
     model: str
     messages: List[ChatMessage]
     tools: Optional[List[FunctionAvailable]] = None
-    do_sample: bool = True
+    do_sample: Optional[bool] = None
     temperature: Optional[float] = None
     top_p: Optional[float] = None
     n: int = 1

src/llamafactory/chat/__init__.py

@@ -0,0 +1,19 @@
+# Copyright 2024 the LlamaFactory team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from .base_engine import BaseEngine
+from .chat_model import ChatModel
+
+
+__all__ = ["BaseEngine", "ChatModel"]

src/llamafactory/chat/base_engine.py

@@ -1,3 +1,17 @@
+# Copyright 2024 the LlamaFactory team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 from abc import ABC, abstractmethod
 from dataclasses import dataclass
 from typing import TYPE_CHECKING, Any, AsyncGenerator, Dict, List, Literal, Optional, Sequence, Union
@@ -36,11 +50,6 @@ class BaseEngine(ABC):
         generating_args: "GeneratingArguments",
     ) -> None: ...
 
-    @abstractmethod
-    async def start(
-        self,
-    ) -> None: ...
-
     @abstractmethod
     async def chat(
         self,