extra setup so llamafactory can run

.gitignore

@@ -150,3 +150,4 @@ dmypy.json
 /huggingface_tokenizers_cache
 /llama-factory/huggingface_tokenizers_cache
 **/Icon?
+llama-factory/data/mgtv_train.json

llama-factory/README.md

@@ -0,0 +1,645 @@
+![# LLaMA Factory](assets/logo.png)
+
+[![GitHub Repo stars](https://img.shields.io/github/stars/hiyouga/LLaMA-Factory?style=social)](https://github.com/hiyouga/LLaMA-Factory/stargazers)
+[![GitHub Code License](https://img.shields.io/github/license/hiyouga/LLaMA-Factory)](LICENSE)
+[![GitHub last commit](https://img.shields.io/github/last-commit/hiyouga/LLaMA-Factory)](https://github.com/hiyouga/LLaMA-Factory/commits/main)
+[![PyPI](https://img.shields.io/pypi/v/llamafactory)](https://pypi.org/project/llamafactory/)
+[![Citation](https://img.shields.io/badge/citation-72-green)](#projects-using-llama-factory)
+[![GitHub pull request](https://img.shields.io/badge/PRs-welcome-blue)](https://github.com/hiyouga/LLaMA-Factory/pulls)
+[![Discord](https://dcbadge.vercel.app/api/server/rKfvV9r9FK?compact=true&style=flat)](https://discord.gg/rKfvV9r9FK)
+[![Twitter](https://img.shields.io/twitter/follow/llamafactory_ai)](https://twitter.com/llamafactory_ai)
+[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing)
+[![Open in DSW](https://gallery.pai-ml.com/assets/open-in-dsw.svg)](https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory)
+[![Spaces](https://img.shields.io/badge/🤗-Open%20in%20Spaces-blue)](https://huggingface.co/spaces/hiyouga/LLaMA-Board)
+[![Studios](https://img.shields.io/badge/ModelScope-Open%20in%20Studios-blue)](https://modelscope.cn/studios/hiyouga/LLaMA-Board)
+
+[![GitHub Tread](https://trendshift.io/api/badge/repositories/4535)](https://trendshift.io/repositories/4535)
+
+👋 Join our [WeChat](assets/wechat.jpg) or [NPU user group](assets/wechat_npu.jpg).
+
+\[ English | [中文](README_zh.md) \]
+
+**Fine-tuning a large language model can be easy as...**
+
+https://github.com/hiyouga/LLaMA-Factory/assets/16256802/9840a653-7e9c-41c8-ae89-7ace5698baf6
+
+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
+
+- [Features](#features)
+- [Benchmark](#benchmark)
+- [Changelog](#changelog)
+- [Supported Models](#supported-models)
+- [Supported Training Approaches](#supported-training-approaches)
+- [Provided Datasets](#provided-datasets)
+- [Requirement](#requirement)
+- [Getting Started](#getting-started)
+- [Projects using LLaMA Factory](#projects-using-llama-factory)
+- [License](#license)
+- [Citation](#citation)
+- [Acknowledgement](#acknowledgement)
+
+## Features
+
+- **Various models**: LLaMA, LLaVA, Mistral, Mixtral-MoE, Qwen, Yi, Gemma, Baichuan, ChatGLM, Phi, etc.
+- **Integrated methods**: (Continuous) pre-training, (multimodal) supervised fine-tuning, reward modeling, PPO, DPO, KTO, ORPO, etc.
+- **Scalable resources**: 16-bit full-tuning, freeze-tuning, LoRA and 2/3/4/5/6/8-bit QLoRA via AQLM/AWQ/GPTQ/LLM.int8/HQQ/EETQ.
+- **Advanced algorithms**: GaLore, 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.
+
+## Benchmark
+
+Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/ptuning), LLaMA Factory's LoRA tuning offers up to **3.7 times faster** training speed with a better Rouge score on the advertising text generation task. By leveraging 4-bit quantization technique, LLaMA Factory's QLoRA further improves the efficiency regarding the GPU memory.
+
+![benchmark](assets/benchmark.svg)
+
+<details><summary>Definitions</summary>
+
+- **Training Speed**: the number of training samples processed per second during the training. (bs=4, cutoff_len=1024)
+- **Rouge Score**: Rouge-2 score on the development set of the [advertising text generation](https://aclanthology.org/D19-1321.pdf) task. (bs=4, cutoff_len=1024)
+- **GPU Memory**: Peak GPU memory usage in 4-bit quantized training. (bs=1, cutoff_len=1024)
+- We adopt `pre_seq_len=128` for ChatGLM's P-Tuning and `lora_rank=32` for LLaMA Factory's LoRA tuning.
+
+</details>
+
+## Changelog
+
+[24/06/16] We support **[PiSSA](https://arxiv.org/abs/2404.02948)** algorithm. See [examples](examples/README.md) for usage.
+
+[24/06/07] We supported fine-tuning the **[Qwen2](https://qwenlm.github.io/blog/qwen2/)** and **[GLM-4](https://github.com/THUDM/GLM-4)** models.
+
+[24/05/26] We supported **[SimPO](https://arxiv.org/abs/2405.14734)** algorithm for preference learning. See [examples](examples/README.md) for usage.
+
+<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.
+
+[24/04/16] We supported **[BAdam](https://arxiv.org/abs/2404.02827)**. See [examples](examples/README.md) for usage.
+
+[24/04/16] We supported **[unsloth](https://github.com/unslothai/unsloth)**'s long-sequence training (Llama-2-7B-56k within 24GB). It achieves **117%** speed and **50%** memory compared with FlashAttention-2, more benchmarks can be found in [this page](https://github.com/hiyouga/LLaMA-Factory/wiki/Performance-comparison).
+
+[24/03/31] We supported **[ORPO](https://arxiv.org/abs/2403.07691)**. See [examples](examples/README.md) for usage.
+
+[24/03/21] Our paper "[LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models](https://arxiv.org/abs/2403.13372)" is available at arXiv!
+
+[24/03/20] We supported **FSDP+QLoRA** that fine-tunes a 70B model on 2x24GB GPUs. See [examples](examples/README.md) for usage.
+
+[24/03/13] We supported **[LoRA+](https://arxiv.org/abs/2402.12354)**. See [examples](examples/README.md) for usage.
+
+[24/03/07] We supported gradient low-rank projection (**[GaLore](https://arxiv.org/abs/2403.03507)**) algorithm. See [examples](examples/README.md) for usage.
+
+[24/03/07] We integrated **[vLLM](https://github.com/vllm-project/vllm)** for faster and concurrent inference. Try `infer_backend: vllm` to enjoy **270%** inference speed.
+
+[24/02/28] We supported weight-decomposed LoRA (**[DoRA](https://arxiv.org/abs/2402.09353)**). Try `use_dora: true` to activate DoRA training.
+
+[24/02/15] We supported **block expansion** proposed by [LLaMA Pro](https://github.com/TencentARC/LLaMA-Pro). See [examples](examples/README.md) for usage.
+
+[24/02/05] Qwen1.5 (Qwen2 beta version) series models are supported in LLaMA-Factory. Check this [blog post](https://qwenlm.github.io/blog/qwen1.5/) for details.
+
+[24/01/18] We supported **agent tuning** for most models, equipping model with tool using abilities by fine-tuning with `dataset: glaive_toolcall_en`.
+
+[23/12/23] We supported **[unsloth](https://github.com/unslothai/unsloth)**'s implementation to boost LoRA tuning for the LLaMA, Mistral and Yi models. Try `use_unsloth: true` argument to activate unsloth patch. It achieves **170%** speed in our benchmark, check [this page](https://github.com/hiyouga/LLaMA-Factory/wiki/Performance-comparison) for details.
+
+[23/12/12] We supported fine-tuning the latest MoE model **[Mixtral 8x7B](https://huggingface.co/mistralai/Mixtral-8x7B-v0.1)** in our framework. See hardware requirement [here](#hardware-requirement).
+
+[23/12/01] We supported downloading pre-trained models and datasets from the **[ModelScope Hub](https://modelscope.cn/models)** for Chinese mainland users. See [this tutorial](#download-from-modelscope-hub) for usage.
+
+[23/10/21] We supported **[NEFTune](https://arxiv.org/abs/2310.05914)** trick for fine-tuning. Try `neftune_noise_alpha: 5` argument to activate NEFTune.
+
+[23/09/27] We supported **$S^2$-Attn** proposed by [LongLoRA](https://github.com/dvlab-research/LongLoRA) for the LLaMA models. Try `shift_attn: true` argument to enable shift short attention.
+
+[23/09/23] We integrated MMLU, C-Eval and CMMLU benchmarks in this repo. See [examples](examples/README.md) for usage.
+
+[23/09/10] We supported **[FlashAttention-2](https://github.com/Dao-AILab/flash-attention)**. Try `flash_attn: fa2` argument to enable FlashAttention-2 if you are using RTX4090, A100 or H100 GPUs.
+
+[23/08/12] We supported **RoPE scaling** to extend the context length of the LLaMA models. Try `rope_scaling: linear` argument in training and `rope_scaling: dynamic` argument at inference to extrapolate the position embeddings.
+
+[23/08/11] We supported **[DPO training](https://arxiv.org/abs/2305.18290)** for instruction-tuned models. See [examples](examples/README.md) for usage.
+
+[23/07/31] We supported **dataset streaming**. Try `streaming: true` and `max_steps: 10000` arguments to load your dataset in streaming mode.
+
+[23/07/29] We released two instruction-tuned 13B models at Hugging Face. See these Hugging Face Repos ([LLaMA-2](https://huggingface.co/hiyouga/Llama-2-Chinese-13b-chat) / [Baichuan](https://huggingface.co/hiyouga/Baichuan-13B-sft)) for details.
+
+[23/07/18] We developed an **all-in-one Web UI** for training, evaluation and inference. Try `train_web.py` to fine-tune models in your Web browser. Thank [@KanadeSiina](https://github.com/KanadeSiina) and [@codemayq](https://github.com/codemayq) for their efforts in the development.
+
+[23/07/09] We released **[FastEdit](https://github.com/hiyouga/FastEdit)** ⚡🩹, an easy-to-use package for editing the factual knowledge of large language models efficiently. Please follow [FastEdit](https://github.com/hiyouga/FastEdit) if you are interested.
+
+[23/06/29] We provided a **reproducible example** of training a chat model using instruction-following datasets, see [Baichuan-7B-sft](https://huggingface.co/hiyouga/Baichuan-7B-sft) for details.
+
+[23/06/22] We aligned the [demo API](src/api_demo.py) with the [OpenAI's](https://platform.openai.com/docs/api-reference/chat) format where you can insert the fine-tuned model in **arbitrary ChatGPT-based applications**.
+
+[23/06/03] We supported quantized training and inference (aka **[QLoRA](https://github.com/artidoro/qlora)**). See [examples](examples/README.md) for usage.
+
+</details>
+
+## Supported Models
+
+| Model                                                        | Model size                       | Template  |
+| ------------------------------------------------------------ | -------------------------------- | --------- |
+| [Baichuan 2](https://huggingface.co/baichuan-inc)            | 7B/13B                           | baichuan2 |
+| [BLOOM/BLOOMZ](https://huggingface.co/bigscience)            | 560M/1.1B/1.7B/3B/7.1B/176B      | -         |
+| [ChatGLM3](https://huggingface.co/THUDM)                     | 6B                               | chatglm3  |
+| [Command R](https://huggingface.co/CohereForAI)              | 35B/104B                         | cohere    |
+| [DeepSeek (Code/MoE)](https://huggingface.co/deepseek-ai)    | 7B/16B/67B/236B                  | deepseek  |
+| [Falcon](https://huggingface.co/tiiuae)                      | 7B/11B/40B/180B                  | falcon    |
+| [Gemma/Gemma 2/CodeGemma](https://huggingface.co/google)     | 2B/7B/9B/27B                     | gemma     |
+| [GLM-4](https://huggingface.co/THUDM)                        | 9B                               | glm4      |
+| [InternLM2](https://huggingface.co/internlm)                 | 7B/20B                           | intern2   |
+| [Llama](https://github.com/facebookresearch/llama)           | 7B/13B/33B/65B                   | -         |
+| [Llama 2](https://huggingface.co/meta-llama)                 | 7B/13B/70B                       | llama2    |
+| [Llama 3](https://huggingface.co/meta-llama)                 | 8B/70B                           | llama3    |
+| [LLaVA-1.5](https://huggingface.co/llava-hf)                 | 7B/13B                           | vicuna    |
+| [Mistral/Mixtral](https://huggingface.co/mistralai)          | 7B/8x7B/8x22B                    | mistral   |
+| [OLMo](https://huggingface.co/allenai)                       | 1B/7B                            | -         |
+| [PaliGemma](https://huggingface.co/google)                   | 3B                               | gemma     |
+| [Phi-1.5/Phi-2](https://huggingface.co/microsoft)            | 1.3B/2.7B                        | -         |
+| [Phi-3](https://huggingface.co/microsoft)                    | 4B/7B/14B                        | phi       |
+| [Qwen/Qwen1.5/Qwen2 (Code/MoE)](https://huggingface.co/Qwen) | 0.5B/1.5B/4B/7B/14B/32B/72B/110B | qwen      |
+| [StarCoder 2](https://huggingface.co/bigcode)                | 3B/7B/15B                        | -         |
+| [XVERSE](https://huggingface.co/xverse)                      | 7B/13B/65B                       | xverse    |
+| [Yi/Yi-1.5](https://huggingface.co/01-ai)                    | 6B/9B/34B                        | yi        |
+| [Yi-VL](https://huggingface.co/01-ai)                        | 6B/34B                           | yi_vl     |
+| [Yuan 2](https://huggingface.co/IEITYuan)                    | 2B/51B/102B                      | yuan      |
+
+> [!NOTE]
+> For the "base" models, the `template` argument can be chosen from `default`, `alpaca`, `vicuna` etc. But make sure to use the **corresponding template** for the "instruct/chat" models.
+>
+> Remember to use the **SAME** template in training and inference.
+
+Please refer to [constants.py](src/llamafactory/extras/constants.py) for a full list of models we supported.
+
+You also can add a custom chat template to [template.py](src/llamafactory/data/template.py).
+
+## Supported Training Approaches
+
+| Approach               |     Full-tuning    |    Freeze-tuning   |       LoRA         |       QLoRA        |
+| ---------------------- | ------------------ | ------------------ | ------------------ | ------------------ |
+| Pre-Training           | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+| Supervised Fine-Tuning | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+| Reward Modeling        | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+| PPO Training           | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+| DPO Training           | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+| KTO Training           | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+| ORPO Training          | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+| SimPO Training         | :white_check_mark: | :white_check_mark: | :white_check_mark: | :white_check_mark: |
+
+## Provided Datasets
+
+<details><summary>Pre-training datasets</summary>
+
+- [Wiki Demo (en)](data/wiki_demo.txt)
+- [RefinedWeb (en)](https://huggingface.co/datasets/tiiuae/falcon-refinedweb)
+- [RedPajama V2 (en)](https://huggingface.co/datasets/togethercomputer/RedPajama-Data-V2)
+- [Wikipedia (en)](https://huggingface.co/datasets/olm/olm-wikipedia-20221220)
+- [Wikipedia (zh)](https://huggingface.co/datasets/pleisto/wikipedia-cn-20230720-filtered)
+- [Pile (en)](https://huggingface.co/datasets/EleutherAI/pile)
+- [SkyPile (zh)](https://huggingface.co/datasets/Skywork/SkyPile-150B)
+- [FineWeb (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb)
+- [FineWeb-Edu (en)](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu)
+- [The Stack (en)](https://huggingface.co/datasets/bigcode/the-stack)
+- [StarCoder (en)](https://huggingface.co/datasets/bigcode/starcoderdata)
+
+</details>
+
+<details><summary>Supervised fine-tuning datasets</summary>
+
+- [Identity (en&zh)](data/identity.json)
+- [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)
+- [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca-3)
+- [Alpaca GPT4 (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
+- [Glaive Function Calling V2 (en&zh)](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2)
+- [LIMA (en)](https://huggingface.co/datasets/GAIR/lima)
+- [Guanaco Dataset (multilingual)](https://huggingface.co/datasets/JosephusCheung/GuanacoDataset)
+- [BELLE 2M (zh)](https://huggingface.co/datasets/BelleGroup/train_2M_CN)
+- [BELLE 1M (zh)](https://huggingface.co/datasets/BelleGroup/train_1M_CN)
+- [BELLE 0.5M (zh)](https://huggingface.co/datasets/BelleGroup/train_0.5M_CN)
+- [BELLE Dialogue 0.4M (zh)](https://huggingface.co/datasets/BelleGroup/generated_chat_0.4M)
+- [BELLE School Math 0.25M (zh)](https://huggingface.co/datasets/BelleGroup/school_math_0.25M)
+- [BELLE Multiturn Chat 0.8M (zh)](https://huggingface.co/datasets/BelleGroup/multiturn_chat_0.8M)
+- [UltraChat (en)](https://github.com/thunlp/UltraChat)
+- [OpenPlatypus (en)](https://huggingface.co/datasets/garage-bAInd/Open-Platypus)
+- [CodeAlpaca 20k (en)](https://huggingface.co/datasets/sahil2801/CodeAlpaca-20k)
+- [Alpaca CoT (multilingual)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT)
+- [OpenOrca (en)](https://huggingface.co/datasets/Open-Orca/OpenOrca)
+- [SlimOrca (en)](https://huggingface.co/datasets/Open-Orca/SlimOrca)
+- [MathInstruct (en)](https://huggingface.co/datasets/TIGER-Lab/MathInstruct)
+- [Firefly 1.1M (zh)](https://huggingface.co/datasets/YeungNLP/firefly-train-1.1M)
+- [Wiki QA (en)](https://huggingface.co/datasets/wiki_qa)
+- [Web QA (zh)](https://huggingface.co/datasets/suolyer/webqa)
+- [WebNovel (zh)](https://huggingface.co/datasets/zxbsmk/webnovel_cn)
+- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
+- [deepctrl (en&zh)](https://www.modelscope.cn/datasets/deepctrl/deepctrl-sft-data)
+- [Advertise Generating (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)
+- [ShareGPT Hyperfiltered (en)](https://huggingface.co/datasets/totally-not-an-llm/sharegpt-hyperfiltered-3k)
+- [ShareGPT4 (en&zh)](https://huggingface.co/datasets/shibing624/sharegpt_gpt4)
+- [UltraChat 200k (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k)
+- [AgentInstruct (en)](https://huggingface.co/datasets/THUDM/AgentInstruct)
+- [LMSYS Chat 1M (en)](https://huggingface.co/datasets/lmsys/lmsys-chat-1m)
+- [Evol Instruct V2 (en)](https://huggingface.co/datasets/WizardLM/WizardLM_evol_instruct_V2_196k)
+- [Cosmopedia (en)](https://huggingface.co/datasets/HuggingFaceTB/cosmopedia)
+- [STEM (zh)](https://huggingface.co/datasets/hfl/stem_zh_instruction)
+- [Ruozhiba (zh)](https://huggingface.co/datasets/hfl/ruozhiba_gpt4_turbo)
+- [Neo-sft (zh)](https://huggingface.co/datasets/m-a-p/neo_sft_phase2)
+- [WebInstructSub (en)](https://huggingface.co/datasets/TIGER-Lab/WebInstructSub)
+- [Magpie-Pro-300K-Filtered (en)](https://huggingface.co/datasets/Magpie-Align/Magpie-Pro-300K-Filtered)
+- [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)
+- [Alpaca GPT4 (de)](https://huggingface.co/datasets/mayflowergmbh/alpaca-gpt4_de)
+- [OpenSchnabeltier (de)](https://huggingface.co/datasets/mayflowergmbh/openschnabeltier_de)
+- [Evol Instruct (de)](https://huggingface.co/datasets/mayflowergmbh/evol-instruct_de)
+- [Dolphin (de)](https://huggingface.co/datasets/mayflowergmbh/dolphin_de)
+- [Booksum (de)](https://huggingface.co/datasets/mayflowergmbh/booksum_de)
+- [Airoboros (de)](https://huggingface.co/datasets/mayflowergmbh/airoboros-3.0_de)
+- [Ultrachat (de)](https://huggingface.co/datasets/mayflowergmbh/ultra-chat_de)
+
+</details>
+
+<details><summary>Preference datasets</summary>
+
+- [DPO mixed (en&zh)](https://huggingface.co/datasets/hiyouga/DPO-En-Zh-20k)
+- [UltraFeedback (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized)
+- [Orca DPO Pairs (en)](https://huggingface.co/datasets/Intel/orca_dpo_pairs)
+- [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
+- [Nectar (en)](https://huggingface.co/datasets/berkeley-nest/Nectar)
+- [Orca DPO (de)](https://huggingface.co/datasets/mayflowergmbh/intel_orca_dpo_pairs_de)
+- [KTO mixed (en)](https://huggingface.co/datasets/argilla/kto-mix-15k)
+
+</details>
+
+Some datasets require confirmation before using them, so we recommend logging in with your Hugging Face account using these commands.
+
+```bash
+pip install --upgrade huggingface_hub
+huggingface-cli login
+```
+
+## Requirement
+
+| Mandatory    | Minimum | Recommend |
+| ------------ | ------- | --------- |
+| python       | 3.8     | 3.11      |
+| torch        | 1.13.1  | 2.3.0     |
+| transformers | 4.41.2  | 4.41.2    |
+| datasets     | 2.16.0  | 2.19.2    |
+| accelerate   | 0.30.1  | 0.30.1    |
+| peft         | 0.11.1  | 0.11.1    |
+| 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         | 0.4.3   | 0.4.3     |
+| flash-attn   | 2.3.0   | 2.5.9     |
+
+### Hardware Requirement
+
+\* *estimated*
+
+| Method            | Bits |   7B  |  13B  |  30B  |   70B  |  110B  |  8x7B |  8x22B |
+| ----------------- | ---- | ----- | ----- | ----- | ------ | ------ | ----- | ------ |
+| Full              | AMP  | 120GB | 240GB | 600GB | 1200GB | 2000GB | 900GB | 2400GB |
+| Full              |  16  |  60GB | 120GB | 300GB |  600GB |  900GB | 400GB | 1200GB |
+| Freeze            |  16  |  20GB |  40GB |  80GB |  200GB |  360GB | 160GB |  400GB |
+| LoRA/GaLore/BAdam |  16  |  16GB |  32GB |  64GB |  160GB |  240GB | 120GB |  320GB |
+| QLoRA             |   8  |  10GB |  20GB |  40GB |   80GB |  140GB |  60GB |  160GB |
+| QLoRA             |   4  |   6GB |  12GB |  24GB |   48GB |   72GB |  30GB |   96GB |
+| QLoRA             |   2  |   4GB |   8GB |  16GB |   24GB |   48GB |  18GB |   48GB |
+
+## Getting Started
+
+### Installation
+
+> [!IMPORTANT]
+> Installation is mandatory.
+
+```bash
+git clone --depth 1 https://github.com/hiyouga/LLaMA-Factory.git
+cd LLaMA-Factory
+pip install -e ".[torch,metrics]"
+```
+
+Extra dependencies available: torch, torch-npu, metrics, deepspeed, bitsandbytes, hqq, eetq, gptq, awq, aqlm, vllm, galore, badam, qwen, modelscope, quality
+
+> [!TIP]
+> Use `pip install --no-deps -e .` to resolve package conflicts.
+
+<details><summary>For Windows users</summary>
+
+If you want to enable the quantized LoRA (QLoRA) on the Windows platform, you need to install a pre-built version of `bitsandbytes` library, which supports CUDA 11.1 to 12.2, please select the appropriate [release version](https://github.com/jllllll/bitsandbytes-windows-webui/releases/tag/wheels) based on your CUDA version.
+
+```bash
+pip install https://github.com/jllllll/bitsandbytes-windows-webui/releases/download/wheels/bitsandbytes-0.41.2.post2-py3-none-win_amd64.whl
+```
+
+To enable FlashAttention-2 on the Windows platform, you need to install the precompiled `flash-attn` library, which supports CUDA 12.1 to 12.2. Please download the corresponding version from [flash-attention](https://github.com/bdashore3/flash-attention/releases) based on your requirements.
+
+</details>
+
+<details><summary>For Ascend NPU users</summary>
+
+To install LLaMA Factory on Ascend NPU devices, please specify extra dependencies: `pip install -e ".[torch-npu,metrics]"`. Additionally, you need to install the **[Ascend CANN Toolkit and Kernels](https://www.hiascend.com/developer/download/community/result?module=cann)**. Please follow the [installation tutorial](https://www.hiascend.com/document/detail/en/CANNCommunityEdition/600alphaX/softwareinstall/instg/atlasdeploy_03_0031.html) or use the following commands:
+
+```bash
+# replace the url according to your CANN version and devices
+# install CANN Toolkit
+wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C17SPC701/Ascend-cann-toolkit_8.0.RC1.alpha001_linux-"$(uname -i)".run
+bash Ascend-cann-toolkit_8.0.RC1.alpha001_linux-"$(uname -i)".run --install
+
+# install CANN Kernels
+wget https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Milan-ASL/Milan-ASL%20V100R001C17SPC701/Ascend-cann-kernels-910b_8.0.RC1.alpha001_linux.run
+bash Ascend-cann-kernels-910b_8.0.RC1.alpha001_linux.run --install
+
+# set env variables
+source /usr/local/Ascend/ascend-toolkit/set_env.sh
+```
+
+| Requirement  | Minimum | Recommend   |
+| ------------ | ------- | ----------- |
+| CANN         | 8.0.RC1 | 8.0.RC1     |
+| torch        | 2.1.0   | 2.1.0       |
+| torch-npu    | 2.1.0   | 2.1.0.post3 |
+| deepspeed    | 0.13.2  | 0.13.2      |
+
+Remember to use `ASCEND_RT_VISIBLE_DEVICES` instead of `CUDA_VISIBLE_DEVICES` to specify the device to use.
+
+If you cannot infer model on NPU devices, try setting `do_sample: false` in the configurations.
+
+Download the pre-built Docker images: [32GB](http://mirrors.cn-central-221.ovaijisuan.com/detail/130.html) | [64GB](http://mirrors.cn-central-221.ovaijisuan.com/detail/131.html)
+
+</details>
+
+### Data Preparation
+
+Please refer to [data/README.md](data/README.md) for checking the details about the format of dataset files. You can either use datasets on HuggingFace / ModelScope hub or load the dataset in local disk.
+
+> [!NOTE]
+> Please update `data/dataset_info.json` to use your custom dataset.
+
+### Quickstart
+
+Use the following 3 commands to run LoRA **fine-tuning**, **inference** and **merging** of the Llama3-8B-Instruct model, respectively.
+
+```bash
+llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
+llamafactory-cli chat examples/inference/llama3_lora_sft.yaml
+llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml
+```
+
+See [examples/README.md](examples/README.md) for advanced usage (including distributed training).
+
+> [!TIP]
+> Use `llamafactory-cli help` to show help information.
+
+### Fine-Tuning with LLaMA Board GUI (powered by [Gradio](https://github.com/gradio-app/gradio))
+
+```bash
+llamafactory-cli webui
+```
+
+### Build Docker
+
+For CUDA users:
+
+```bash
+cd docker/docker-cuda/
+docker-compose up -d
+docker-compose exec llamafactory bash
+```
+
+For Ascend NPU users:
+
+```bash
+cd docker/docker-npu/
+docker-compose up -d
+docker-compose exec llamafactory bash
+```
+
+<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>
+
+<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.
+- data: Place datasets on this dir of the host machine so that they can be selected on LLaMA Board GUI.
+- output: Set export dir to this location so that the merged result can be accessed directly on the host machine.
+
+</details>
+
+### Deploy with OpenAI-style API and vLLM
+
+```bash
+API_PORT=8000 llamafactory-cli api examples/inference/llama3_vllm.yaml
+```
+
+> [!TIP]
+> Visit https://platform.openai.com/docs/api-reference/chat/create for API document.
+
+### Download from ModelScope Hub
+
+If you have trouble with downloading models and datasets from Hugging Face, you can use ModelScope.
+
+```bash
+export USE_MODELSCOPE_HUB=1 # `set USE_MODELSCOPE_HUB=1` for Windows
+```
+
+Train the model by specifying a model ID of the ModelScope Hub as the `model_name_or_path`. You can find a full list of model IDs at [ModelScope Hub](https://modelscope.cn/models), e.g., `LLM-Research/Meta-Llama-3-8B-Instruct`.
+
+### Use W&B Logger
+
+To use [Weights & Biases](https://wandb.ai) for logging experimental results, you need to add the following arguments to yaml files.
+
+```yaml
+report_to: wandb
+run_name: test_run # optional
+```
+
+Set `WANDB_API_KEY` to [your key](https://wandb.ai/authorize) when launching training tasks to log in with your W&B account.
+
+## Projects using LLaMA Factory
+
+If you have a project that should be incorporated, please contact via email or create a pull request.
+
+<details><summary>Click to show</summary>
+
+1. Wang et al. ESRL: Efficient Sampling-based Reinforcement Learning for Sequence Generation. 2023. [[arxiv]](https://arxiv.org/abs/2308.02223)
+1. Yu et al. Open, Closed, or Small Language Models for Text Classification? 2023. [[arxiv]](https://arxiv.org/abs/2308.10092)
+1. Wang et al. UbiPhysio: Support Daily Functioning, Fitness, and Rehabilitation with Action Understanding and Feedback in Natural Language. 2023. [[arxiv]](https://arxiv.org/abs/2308.10526)
+1. Luceri et al. Leveraging Large Language Models to Detect Influence Campaigns in Social Media. 2023. [[arxiv]](https://arxiv.org/abs/2311.07816)
+1. Zhang et al. Alleviating Hallucinations of Large Language Models through Induced Hallucinations. 2023. [[arxiv]](https://arxiv.org/abs/2312.15710)
+1. Wang et al. Know Your Needs Better: Towards Structured Understanding of Marketer Demands with Analogical Reasoning Augmented LLMs. KDD 2024. [[arxiv]](https://arxiv.org/abs/2401.04319)
+1. Wang et al. CANDLE: Iterative Conceptualization and Instantiation Distillation from Large Language Models for Commonsense Reasoning. ACL 2024. [[arxiv]](https://arxiv.org/abs/2401.07286)
+1. Choi et al. FACT-GPT: Fact-Checking Augmentation via Claim Matching with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2402.05904)
+1. Zhang et al. AutoMathText: Autonomous Data Selection with Language Models for Mathematical Texts. 2024. [[arxiv]](https://arxiv.org/abs/2402.07625)
+1. Lyu et al. KnowTuning: Knowledge-aware Fine-tuning for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11176)
+1. Yang et al. LaCo: Large Language Model Pruning via Layer Collaps. 2024. [[arxiv]](https://arxiv.org/abs/2402.11187)
+1. Bhardwaj et al. Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task Arithmetic. 2024. [[arxiv]](https://arxiv.org/abs/2402.11746)
+1. Yang et al. Enhancing Empathetic Response Generation by Augmenting LLMs with Small-scale Empathetic Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11801)
+1. Yi et al. Generation Meets Verification: Accelerating Large Language Model Inference with Smart Parallel Auto-Correct Decoding. ACL 2024 Findings. [[arxiv]](https://arxiv.org/abs/2402.11809)
+1. Cao et al. Head-wise Shareable Attention for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.11819)
+1. Zhang et al. Enhancing Multilingual Capabilities of Large Language Models through Self-Distillation from Resource-Rich Languages. 2024. [[arxiv]](https://arxiv.org/abs/2402.12204)
+1. Kim et al. Efficient and Effective Vocabulary Expansion Towards Multilingual Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2402.14714)
+1. Yu et al. KIEval: A Knowledge-grounded Interactive Evaluation Framework for Large Language Models. ACL 2024. [[arxiv]](https://arxiv.org/abs/2402.15043)
+1. Huang et al. Key-Point-Driven Data Synthesis with its Enhancement on Mathematical Reasoning. 2024. [[arxiv]](https://arxiv.org/abs/2403.02333)
+1. Duan et al. Negating Negatives: Alignment without Human Positive Samples via Distributional Dispreference Optimization. 2024. [[arxiv]](https://arxiv.org/abs/2403.03419)
+1. Xie and Schwertfeger. Empowering Robotics with Large Language Models: osmAG Map Comprehension with LLMs. 2024. [[arxiv]](https://arxiv.org/abs/2403.08228)
+1. Wu et al. Large Language Models are Parallel Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2403.09073)
+1. Zhang et al. EDT: Improving Large Language Models' Generation by Entropy-based Dynamic Temperature Sampling. 2024. [[arxiv]](https://arxiv.org/abs/2403.14541)
+1. Weller et al. FollowIR: Evaluating and Teaching Information Retrieval Models to Follow Instructions. 2024. [[arxiv]](https://arxiv.org/abs/2403.15246)
+1. Hongbin Na. CBT-LLM: A Chinese Large Language Model for Cognitive Behavioral Therapy-based Mental Health Question Answering. COLING 2024. [[arxiv]](https://arxiv.org/abs/2403.16008)
+1. Zan et al. CodeS: Natural Language to Code Repository via Multi-Layer Sketch. 2024. [[arxiv]](https://arxiv.org/abs/2403.16443)
+1. Liu et al. Extensive Self-Contrast Enables Feedback-Free Language Model Alignment. 2024. [[arxiv]](https://arxiv.org/abs/2404.00604)
+1. Luo et al. BAdam: A Memory Efficient Full Parameter Training Method for Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.02827)
+1. Du et al. Chinese Tiny LLM: Pretraining a Chinese-Centric Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2404.04167)
+1. Ma et al. Parameter Efficient Quasi-Orthogonal Fine-Tuning via Givens Rotation. ICML 2024. [[arxiv]](https://arxiv.org/abs/2404.04316)
+1. Liu et al. Dynamic Generation of Personalities with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.07084)
+1. Shang et al. How Far Have We Gone in Stripped Binary Code Understanding Using Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.09836)
+1. Huang et al. LLMTune: Accelerate Database Knob Tuning with Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2404.11581)
+1. Deng et al. Text-Tuple-Table: Towards Information Integration in Text-to-Table Generation via Global Tuple Extraction. 2024. [[arxiv]](https://arxiv.org/abs/2404.14215)
+1. Acikgoz et al. Hippocrates: An Open-Source Framework for Advancing Large Language Models in Healthcare. 2024. [[arxiv]](https://arxiv.org/abs/2404.16621)
+1. Zhang et al. Small Language Models Need Strong Verifiers to Self-Correct Reasoning. ACL 2024 Findings. [[arxiv]](https://arxiv.org/abs/2404.17140)
+1. Zhou et al. FREB-TQA: A Fine-Grained Robustness Evaluation Benchmark for Table Question Answering. NAACL 2024. [[arxiv]](https://arxiv.org/abs/2404.18585)
+1. Xu et al. Large Language Models for Cyber Security: A Systematic Literature Review. 2024. [[arxiv]](https://arxiv.org/abs/2405.04760)
+1. Dammu et al. "They are uncultured": Unveiling Covert Harms and Social Threats in LLM Generated Conversations. 2024. [[arxiv]](https://arxiv.org/abs/2405.05378)
+1. Yi et al. A safety realignment framework via subspace-oriented model fusion for large language models. 2024. [[arxiv]](https://arxiv.org/abs/2405.09055)
+1. Lou et al. SPO: Multi-Dimensional Preference Sequential Alignment With Implicit Reward Modeling. 2024. [[arxiv]](https://arxiv.org/abs/2405.12739)
+1. Zhang et al. Getting More from Less: Large Language Models are Good Spontaneous Multilingual Learners. 2024. [[arxiv]](https://arxiv.org/abs/2405.13816)
+1. Zhang et al. TS-Align: A Teacher-Student Collaborative Framework for Scalable Iterative Finetuning of Large Language Models. 2024. [[arxiv]](https://arxiv.org/abs/2405.20215)
+1. Zihong Chen. Sentence Segmentation and Sentence Punctuation Based on XunziALLM. 2024. [[paper]](https://aclanthology.org/2024.lt4hala-1.30)
+1. Gao et al. The Best of Both Worlds: Toward an Honest and Helpful Large Language Model. 2024. [[arxiv]](https://arxiv.org/abs/2406.00380)
+1. Wang and Song. MARS: Benchmarking the Metaphysical Reasoning Abilities of Language Models with a Multi-task Evaluation Dataset. 2024. [[arxiv]](https://arxiv.org/abs/2406.02106)
+1. Hu et al. Computational Limits of Low-Rank Adaptation (LoRA) for Transformer-Based Models. 2024. [[arxiv]](https://arxiv.org/abs/2406.03136)
+1. Ge et al. Time Sensitive Knowledge Editing through Efficient Finetuning. ACL 2024. [[arxiv]](https://arxiv.org/abs/2406.04496)
+1. Tan et al. Peer Review as A Multi-Turn and Long-Context Dialogue with Role-Based Interactions. 2024. [[arxiv]](https://arxiv.org/abs/2406.05688)
+1. Song et al. Turbo Sparse: Achieving LLM SOTA Performance with Minimal Activated Parameters. 2024. [[arxiv]](https://arxiv.org/abs/2406.05955)
+1. Gu et al. RWKV-CLIP: A Robust Vision-Language Representation Learner. 2024. [[arxiv]](https://arxiv.org/abs/2406.06973)
+1. Chen et al. Advancing Tool-Augmented Large Language Models: Integrating Insights from Errors in Inference Trees. 2024. [[arxiv]](https://arxiv.org/abs/2406.07115)
+1. Zhu et al. Are Large Language Models Good Statisticians?. 2024. [[arxiv]](https://arxiv.org/abs/2406.07815)
+1. Li et al. Know the Unknown: An Uncertainty-Sensitive Method for LLM Instruction Tuning. 2024. [[arxiv]](https://arxiv.org/abs/2406.10099)
+1. Ding et al. IntentionQA: A Benchmark for Evaluating Purchase Intention Comprehension Abilities of Language Models in E-commerce. 2024. [[arxiv]](https://arxiv.org/abs/2406.10173)
+1. He et al. COMMUNITY-CROSS-INSTRUCT: Unsupervised Instruction Generation for Aligning Large Language Models to Online Communities. 2024. [[arxiv]](https://arxiv.org/abs/2406.12074)
+1. Lin et al. FVEL: Interactive Formal Verification Environment with Large Language Models via Theorem Proving. 2024. [[arxiv]](https://arxiv.org/abs/2406.14408)
+1. Treutlein et al. Connecting the Dots: LLMs can Infer and Verbalize Latent Structure from Disparate Training Data. 2024. [[arxiv]](https://arxiv.org/abs/2406.14546)
+1. Feng et al. SS-Bench: A Benchmark for Social Story Generation and Evaluation. 2024. [[arxiv]](https://arxiv.org/abs/2406.15695)
+1. Feng et al. Self-Constructed Context Decompilation with Fined-grained Alignment Enhancement. 2024. [[arxiv]](https://arxiv.org/abs/2406.17233)
+1. Liu et al. Large Language Models for Cuffless Blood Pressure Measurement From Wearable Biosignals. 2024. [[arxiv]](https://arxiv.org/abs/2406.18069)
+1. Iyer et al. Exploring Very Low-Resource Translation with LLMs: The University of Edinburgh’s Submission to AmericasNLP 2024 Translation Task. AmericasNLP 2024. [[paper]](https://aclanthology.org/2024.americasnlp-1.25)
+1. **[StarWhisper](https://github.com/Yu-Yang-Li/StarWhisper)**: A large language model for Astronomy, based on ChatGLM2-6B and Qwen-14B.
+1. **[DISC-LawLLM](https://github.com/FudanDISC/DISC-LawLLM)**: A large language model specialized in Chinese legal domain, based on Baichuan-13B, is capable of retrieving and reasoning on legal knowledge.
+1. **[Sunsimiao](https://github.com/X-D-Lab/Sunsimiao)**: A large language model specialized in Chinese medical domain, based on Baichuan-7B and ChatGLM-6B.
+1. **[CareGPT](https://github.com/WangRongsheng/CareGPT)**: A series of large language models for Chinese medical domain, based on LLaMA2-7B and Baichuan-13B.
+1. **[MachineMindset](https://github.com/PKU-YuanGroup/Machine-Mindset/)**: A series of MBTI Personality large language models, capable of giving any LLM 16 different personality types based on different datasets and training methods.
+1. **[Luminia-13B-v3](https://huggingface.co/Nekochu/Luminia-13B-v3)**: A large language model specialized in generate metadata for stable diffusion. [[🤗Demo]](https://huggingface.co/spaces/Nekochu/Luminia-13B_SD_Prompt)
+1. **[Chinese-LLaVA-Med](https://github.com/BUAADreamer/Chinese-LLaVA-Med)**: A multimodal large language model specialized in Chinese medical domain, based on LLaVA-1.5-7B.
+1. **[AutoRE](https://github.com/THUDM/AutoRE)**: A document-level relation extraction system based on large language models.
+1. **[NVIDIA RTX AI Toolkit](https://github.com/NVIDIA/RTX-AI-Toolkit)**: SDKs for fine-tuning LLMs on Windows PC for NVIDIA RTX.
+1. **[LazyLLM](https://github.com/LazyAGI/LazyLLM)**: An easy and lazy way for building multi-agent LLMs applications and supports model fine-tuning via LLaMA Factory.
+
+</details>
+
+## License
+
+This repository is licensed under the [Apache-2.0 License](LICENSE).
+
+Please follow the model licenses to use the corresponding model weights: [Baichuan 2](https://huggingface.co/baichuan-inc/Baichuan2-7B-Base/blob/main/Community%20License%20for%20Baichuan%202%20Model.pdf) / [BLOOM](https://huggingface.co/spaces/bigscience/license) / [ChatGLM3](https://github.com/THUDM/ChatGLM3/blob/main/MODEL_LICENSE) / [Command R](https://cohere.com/c4ai-cc-by-nc-license) / [DeepSeek](https://github.com/deepseek-ai/DeepSeek-LLM/blob/main/LICENSE-MODEL) / [Falcon](https://huggingface.co/tiiuae/falcon-180B/blob/main/LICENSE.txt) / [Gemma](https://ai.google.dev/gemma/terms) / [GLM-4](https://huggingface.co/THUDM/glm-4-9b/blob/main/LICENSE) / [InternLM2](https://github.com/InternLM/InternLM#license) / [Llama](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md) / [Llama 2 (LLaVA-1.5)](https://ai.meta.com/llama/license/) / [Llama 3](https://llama.meta.com/llama3/license/) / [Mistral](LICENSE) / [OLMo](LICENSE) / [Phi-1.5/Phi-2](https://huggingface.co/microsoft/phi-1_5/resolve/main/Research%20License.docx) / [Phi-3](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/blob/main/LICENSE) / [Qwen](https://github.com/QwenLM/Qwen/blob/main/Tongyi%20Qianwen%20LICENSE%20AGREEMENT) / [StarCoder 2](https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement) / [XVERSE](https://github.com/xverse-ai/XVERSE-13B/blob/main/MODEL_LICENSE.pdf) / [Yi](https://huggingface.co/01-ai/Yi-6B/blob/main/LICENSE) / [Yi-1.5](LICENSE) / [Yuan 2](https://github.com/IEIT-Yuan/Yuan-2.0/blob/main/LICENSE-Yuan)
+
+## Citation
+
+If this work is helpful, please kindly cite as:
+
+```bibtex
+@inproceedings{zheng2024llamafactory,
+  title={LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models},
+  author={Yaowei Zheng and Richong Zhang and Junhao Zhang and Yanhan Ye and Zheyan Luo and Zhangchi Feng and Yongqiang Ma},
+  booktitle={Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
+  address={Bangkok, Thailand},
+  publisher={Association for Computational Linguistics},
+  year={2024},
+  url={http://arxiv.org/abs/2403.13372}
+}
+```
+
+## Acknowledgement
+
+This repo benefits from [PEFT](https://github.com/huggingface/peft), [TRL](https://github.com/huggingface/trl), [QLoRA](https://github.com/artidoro/qlora) and [FastChat](https://github.com/lm-sys/FastChat). Thanks for their wonderful works.
+
+## Star History
+
+![Star History Chart](https://api.star-history.com/svg?repos=hiyouga/LLaMA-Factory&type=Date)

llama-factory/pyproject.toml

@@ -0,0 +1,33 @@
+[build-system]
+requires = ["setuptools>=61.0"]
+build-backend = "setuptools.build_meta"
+
+[tool.ruff]
+target-version = "py38"
+line-length = 119
+indent-width = 4
+
+[tool.ruff.lint]
+ignore = ["C408", "C901", "E501", "E731", "E741", "W605"]
+select = ["C", "E", "F", "I", "W"]
+
+[tool.ruff.lint.isort]
+lines-after-imports = 2
+known-first-party = ["llamafactory"]
+known-third-party = [
+    "accelerate",
+    "datasets",
+    "gradio",
+    "numpy",
+    "peft",
+    "torch",
+    "transformers",
+    "trl"
+]
+
+[tool.ruff.format]
+quote-style = "double"
+indent-style = "space"
+docstring-code-format = true
+skip-magic-trailing-comma = false
+line-ending = "auto"

llama-factory/requirements.txt

@@ -0,0 +1,21 @@
+transformers>=4.41.2
+datasets>=2.16.0
+accelerate>=0.30.1
+peft>=0.11.1
+trl>=0.8.6
+gradio>=4.0.0
+pandas>=2.0.0
+scipy
+einops
+sentencepiece
+tiktoken
+protobuf
+uvicorn
+pydantic
+fastapi
+sse-starlette
+matplotlib>=3.7.0
+fire
+packaging
+pyyaml
+numpy<2.0.0

llama-factory/setup.py

@@ -0,0 +1,92 @@
+# 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
+
+from setuptools import find_packages, setup
+
+
+def get_version():
+    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)
+        return version
+
+
+def get_requires():
+    with open("requirements.txt", "r", encoding="utf-8") as f:
+        file_content = f.read()
+        lines = [line.strip() for line in file_content.strip().split("\n") if not line.startswith("#")]
+        return lines
+
+
+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"],
+    "bitsandbytes": ["bitsandbytes>=0.39.0"],
+    "hqq": ["hqq"],
+    "eetq": ["eetq"],
+    "gptq": ["optimum>=1.17.0", "auto-gptq>=0.5.0"],
+    "awq": ["autoawq"],
+    "aqlm": ["aqlm[gpu]>=1.1.0"],
+    "vllm": ["vllm>=0.4.3"],
+    "galore": ["galore-torch"],
+    "badam": ["badam>=1.2.1"],
+    "qwen": ["transformers_stream_generator"],
+    "modelscope": ["modelscope"],
+    "dev": ["ruff", "pytest"],
+}
+
+
+def main():
+    setup(
+        name="llamafactory",
+        version=get_version(),
+        author="hiyouga",
+        author_email="hiyouga" "@" "buaa.edu.cn",
+        description="Easy-to-use LLM fine-tuning framework",
+        long_description=open("README.md", "r", encoding="utf-8").read(),
+        long_description_content_type="text/markdown",
+        keywords=["LLaMA", "BLOOM", "Falcon", "LLM", "ChatGPT", "transformer", "pytorch", "deep learning"],
+        license="Apache 2.0 License",
+        url="https://github.com/hiyouga/LLaMA-Factory",
+        package_dir={"": "src"},
+        packages=find_packages("src"),
+        python_requires=">=3.8.0",
+        install_requires=get_requires(),
+        extras_require=extra_require,
+        entry_points={"console_scripts": ["llamafactory-cli = llamafactory.cli:main"]},
+        classifiers=[
+            "Development Status :: 4 - Beta",
+            "Intended Audience :: Developers",
+            "Intended Audience :: Education",
+            "Intended Audience :: Science/Research",
+            "License :: OSI Approved :: Apache Software License",
+            "Operating System :: OS Independent",
+            "Programming Language :: Python :: 3",
+            "Programming Language :: Python :: 3.8",
+            "Programming Language :: Python :: 3.9",
+            "Programming Language :: Python :: 3.10",
+            "Programming Language :: Python :: 3.11",
+            "Topic :: Scientific/Engineering :: Artificial Intelligence",
+        ],
+    )
+
+
+if __name__ == "__main__":
+    main()

llama-factory/src/api.py

@@ -0,0 +1,33 @@
+# 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 llamafactory.api.app import create_app
+from llamafactory.chat import ChatModel
+
+
+def main():
+    chat_model = ChatModel()
+    app = create_app(chat_model)
+    api_host = os.environ.get("API_HOST", "0.0.0.0")
+    api_port = int(os.environ.get("API_PORT", "8000"))
+    print("Visit http://localhost:{}/docs for API document.".format(api_port))
+    uvicorn.run(app, host=api_host, port=api_port)
+
+
+if __name__ == "__main__":
+    main()

llama-factory/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

llama-factory/src/llamafactory/api/app.py

@@ -0,0 +1,122 @@
+# 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
+
+from typing_extensions import Annotated
+
+from ..chat import ChatModel
+from ..extras.misc import torch_gc
+from ..extras.packages import is_fastapi_available, is_starlette_available, is_uvicorn_available
+from .chat import (
+    create_chat_completion_response,
+    create_score_evaluation_response,
+    create_stream_chat_completion_response,
+)
+from .protocol import (
+    ChatCompletionRequest,
+    ChatCompletionResponse,
+    ModelCard,
+    ModelList,
+    ScoreEvaluationRequest,
+    ScoreEvaluationResponse,
+)
+
+
+if is_fastapi_available():
+    from fastapi import Depends, FastAPI, HTTPException, status
+    from fastapi.middleware.cors import CORSMiddleware
+    from fastapi.security.http import HTTPAuthorizationCredentials, HTTPBearer
+
+
+if is_starlette_available():
+    from sse_starlette import EventSourceResponse
+
+
+if is_uvicorn_available():
+    import uvicorn
+
+
+@asynccontextmanager
+async def lifespan(app: "FastAPI"):  # collects GPU memory
+    yield
+    torch_gc()
+
+
+def create_app(chat_model: "ChatModel") -> "FastAPI":
+    app = FastAPI(lifespan=lifespan)
+    app.add_middleware(
+        CORSMiddleware,
+        allow_origins=["*"],
+        allow_credentials=True,
+        allow_methods=["*"],
+        allow_headers=["*"],
+    )
+    api_key = os.environ.get("API_KEY")
+    security = HTTPBearer(auto_error=False)
+
+    async def verify_api_key(auth: Annotated[Optional[HTTPAuthorizationCredentials], Depends(security)]):
+        if api_key and (auth is None or auth.credentials != api_key):
+            raise HTTPException(status_code=status.HTTP_401_UNAUTHORIZED, detail="Invalid API key.")
+
+    @app.get(
+        "/v1/models",
+        response_model=ModelList,
+        status_code=status.HTTP_200_OK,
+        dependencies=[Depends(verify_api_key)],
+    )
+    async def list_models():
+        model_card = ModelCard(id="gpt-3.5-turbo")
+        return ModelList(data=[model_card])
+
+    @app.post(
+        "/v1/chat/completions",
+        response_model=ChatCompletionResponse,
+        status_code=status.HTTP_200_OK,
+        dependencies=[Depends(verify_api_key)],
+    )
+    async def create_chat_completion(request: ChatCompletionRequest):
+        if not chat_model.engine.can_generate:
+            raise HTTPException(status_code=status.HTTP_405_METHOD_NOT_ALLOWED, detail="Not allowed")
+
+        if request.stream:
+            generate = create_stream_chat_completion_response(request, chat_model)
+            return EventSourceResponse(generate, media_type="text/event-stream")
+        else:
+            return await create_chat_completion_response(request, chat_model)
+
+    @app.post(
+        "/v1/score/evaluation",
+        response_model=ScoreEvaluationResponse,
+        status_code=status.HTTP_200_OK,
+        dependencies=[Depends(verify_api_key)],
+    )
+    async def create_score_evaluation(request: ScoreEvaluationRequest):
+        if chat_model.engine.can_generate:
+            raise HTTPException(status_code=status.HTTP_405_METHOD_NOT_ALLOWED, detail="Not allowed")
+
+        return await create_score_evaluation_response(request, chat_model)
+
+    return app
+
+
+def run_api() -> None:
+    chat_model = ChatModel()
+    app = create_app(chat_model)
+    api_host = os.environ.get("API_HOST", "0.0.0.0")
+    api_port = int(os.environ.get("API_PORT", "8000"))
+    print("Visit http://localhost:{}/docs for API document.".format(api_port))
+    uvicorn.run(app, host=api_host, port=api_port)

llama-factory/src/llamafactory/api/chat.py

@@ -0,0 +1,237 @@
+# 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, is_pillow_available, is_requests_available
+from .common import dictify, jsonify
+from .protocol import (
+    ChatCompletionMessage,
+    ChatCompletionResponse,
+    ChatCompletionResponseChoice,
+    ChatCompletionResponseUsage,
+    ChatCompletionStreamResponse,
+    ChatCompletionStreamResponseChoice,
+    Finish,
+    Function,
+    FunctionCall,
+    Role,
+    ScoreEvaluationResponse,
+)
+
+
+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
+
+
+logger = get_logger(__name__)
+ROLE_MAPPING = {
+    Role.USER: DataRole.USER.value,
+    Role.ASSISTANT: DataRole.ASSISTANT.value,
+    Role.SYSTEM: DataRole.SYSTEM.value,
+    Role.FUNCTION: DataRole.FUNCTION.value,
+    Role.TOOL: DataRole.OBSERVATION.value,
+}
+
+
+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:
+        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Invalid length")
+
+    if request.messages[0].role == Role.SYSTEM:
+        system = request.messages.pop(0).content
+    else:
+        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")
+        elif i % 2 == 1 and message.role not in [Role.ASSISTANT, Role.FUNCTION]:
+            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):
+            tool_calls = [
+                {"name": tool_call.function.name, "arguments": tool_call.function.arguments}
+                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})
+
+    tool_list = request.tools
+    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 json.JSONDecodeError:
+            raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Invalid tools")
+    else:
+        tools = None
+
+    return input_messages, system, tools, image
+
+
+def _create_stream_chat_completion_chunk(
+    completion_id: str,
+    model: str,
+    delta: "ChatCompletionMessage",
+    index: Optional[int] = 0,
+    finish_reason: Optional["Finish"] = None,
+) -> str:
+    choice_data = ChatCompletionStreamResponseChoice(index=index, delta=delta, finish_reason=finish_reason)
+    chunk = ChatCompletionStreamResponse(id=completion_id, model=model, choices=[choice_data])
+    return jsonify(chunk)
+
+
+async def create_chat_completion_response(
+    request: "ChatCompletionRequest", chat_model: "ChatModel"
+) -> "ChatCompletionResponse":
+    completion_id = "chatcmpl-{}".format(uuid.uuid4().hex)
+    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,
+        max_new_tokens=request.max_tokens,
+        num_return_sequences=request.n,
+        stop=request.stop,
+    )
+
+    prompt_length, response_length = 0, 0
+    choices = []
+    for i, response in enumerate(responses):
+        if tools:
+            result = chat_model.engine.template.extract_tool(response.response_text)
+        else:
+            result = response.response_text
+
+        if isinstance(result, list):
+            tool_calls = []
+            for tool in result:
+                function = Function(name=tool[0], arguments=tool[1])
+                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)
+            finish_reason = Finish.STOP if response.finish_reason == "stop" else Finish.LENGTH
+
+        choices.append(ChatCompletionResponseChoice(index=i, message=response_message, finish_reason=finish_reason))
+        prompt_length = response.prompt_length
+        response_length += response.response_length
+
+    usage = ChatCompletionResponseUsage(
+        prompt_tokens=prompt_length,
+        completion_tokens=response_length,
+        total_tokens=prompt_length + response_length,
+    )
+
+    return ChatCompletionResponse(id=completion_id, model=request.model, choices=choices, usage=usage)
+
+
+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, image = _process_request(request)
+    if tools:
+        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Cannot stream function calls.")
+
+    if request.n > 1:
+        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Cannot stream multiple responses.")
+
+    yield _create_stream_chat_completion_chunk(
+        completion_id=completion_id, model=request.model, delta=ChatCompletionMessage(role=Role.ASSISTANT, content="")
+    )
+    async for new_token in chat_model.astream_chat(
+        input_messages,
+        system,
+        tools,
+        image,
+        do_sample=request.do_sample,
+        temperature=request.temperature,
+        top_p=request.top_p,
+        max_new_tokens=request.max_tokens,
+        stop=request.stop,
+    ):
+        if len(new_token) != 0:
+            yield _create_stream_chat_completion_chunk(
+                completion_id=completion_id, model=request.model, delta=ChatCompletionMessage(content=new_token)
+            )
+
+    yield _create_stream_chat_completion_chunk(
+        completion_id=completion_id, model=request.model, delta=ChatCompletionMessage(), finish_reason=Finish.STOP
+    )
+    yield "[DONE]"
+
+
+async def create_score_evaluation_response(
+    request: "ScoreEvaluationRequest", chat_model: "ChatModel"
+) -> "ScoreEvaluationResponse":
+    if len(request.messages) == 0:
+        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Invalid request")
+
+    scores = await chat_model.aget_scores(request.messages, max_length=request.max_length)
+    return ScoreEvaluationResponse(model=request.model, scores=scores)

llama-factory/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)

llama-factory/src/llamafactory/api/protocol.py

@@ -0,0 +1,153 @@
+# 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
+
+from pydantic import BaseModel, Field
+from typing_extensions import Literal
+
+
+@unique
+class Role(str, Enum):
+    USER = "user"
+    ASSISTANT = "assistant"
+    SYSTEM = "system"
+    FUNCTION = "function"
+    TOOL = "tool"
+
+
+@unique
+class Finish(str, Enum):
+    STOP = "stop"
+    LENGTH = "length"
+    TOOL = "tool_calls"
+
+
+class ModelCard(BaseModel):
+    id: str
+    object: Literal["model"] = "model"
+    created: int = Field(default_factory=lambda: int(time.time()))
+    owned_by: Literal["owner"] = "owner"
+
+
+class ModelList(BaseModel):
+    object: Literal["list"] = "list"
+    data: List[ModelCard] = []
+
+
+class Function(BaseModel):
+    name: str
+    arguments: str
+
+
+class FunctionDefinition(BaseModel):
+    name: str
+    description: str
+    parameters: Dict[str, Any]
+
+
+class FunctionAvailable(BaseModel):
+    type: Literal["function", "code_interpreter"] = "function"
+    function: Optional[FunctionDefinition] = None
+
+
+class FunctionCall(BaseModel):
+    id: str
+    type: Literal["function"] = "function"
+    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[Union[str, List[MultimodalInputItem]]] = None
+    tool_calls: Optional[List[FunctionCall]] = None
+
+
+class ChatCompletionMessage(BaseModel):
+    role: Optional[Role] = None
+    content: Optional[str] = None
+    tool_calls: Optional[List[FunctionCall]] = None
+
+
+class ChatCompletionRequest(BaseModel):
+    model: str
+    messages: List[ChatMessage]
+    tools: Optional[List[FunctionAvailable]] = None
+    do_sample: Optional[bool] = None
+    temperature: Optional[float] = None
+    top_p: Optional[float] = None
+    n: int = 1
+    max_tokens: Optional[int] = None
+    stop: Optional[Union[str, List[str]]] = None
+    stream: bool = False
+
+
+class ChatCompletionResponseChoice(BaseModel):
+    index: int
+    message: ChatCompletionMessage
+    finish_reason: Finish
+
+
+class ChatCompletionStreamResponseChoice(BaseModel):
+    index: int
+    delta: ChatCompletionMessage
+    finish_reason: Optional[Finish] = None
+
+
+class ChatCompletionResponseUsage(BaseModel):
+    prompt_tokens: int
+    completion_tokens: int
+    total_tokens: int
+
+
+class ChatCompletionResponse(BaseModel):
+    id: str
+    object: Literal["chat.completion"] = "chat.completion"
+    created: int = Field(default_factory=lambda: int(time.time()))
+    model: str
+    choices: List[ChatCompletionResponseChoice]
+    usage: ChatCompletionResponseUsage
+
+
+class ChatCompletionStreamResponse(BaseModel):
+    id: str
+    object: Literal["chat.completion.chunk"] = "chat.completion.chunk"
+    created: int = Field(default_factory=lambda: int(time.time()))
+    model: str
+    choices: List[ChatCompletionStreamResponseChoice]
+
+
+class ScoreEvaluationRequest(BaseModel):
+    model: str
+    messages: List[str]
+    max_length: Optional[int] = None
+
+
+class ScoreEvaluationResponse(BaseModel):
+    id: str
+    object: Literal["score.evaluation"] = "score.evaluation"
+    model: str
+    scores: List[float]

llama-factory/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"]

llama-factory/src/llamafactory/chat/base_engine.py

@@ -0,0 +1,78 @@
+# 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
+
+
+if TYPE_CHECKING:
+    from numpy.typing import NDArray
+    from transformers import PreTrainedModel, PreTrainedTokenizer
+    from vllm import AsyncLLMEngine
+
+    from ..data import Template
+    from ..hparams import DataArguments, FinetuningArguments, GeneratingArguments, ModelArguments
+
+
+@dataclass
+class Response:
+    response_text: str
+    response_length: int
+    prompt_length: int
+    finish_reason: Literal["stop", "length"]
+
+
+class BaseEngine(ABC):
+    model: Union["PreTrainedModel", "AsyncLLMEngine"]
+    tokenizer: "PreTrainedTokenizer"
+    can_generate: bool
+    template: "Template"
+    generating_args: Dict[str, Any]
+
+    @abstractmethod
+    def __init__(
+        self,
+        model_args: "ModelArguments",
+        data_args: "DataArguments",
+        finetuning_args: "FinetuningArguments",
+        generating_args: "GeneratingArguments",
+    ) -> None: ...
+
+    @abstractmethod
+    async def chat(
+        self,
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        **input_kwargs,
+    ) -> List["Response"]: ...
+
+    @abstractmethod
+    async def stream_chat(
+        self,
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        **input_kwargs,
+    ) -> AsyncGenerator[str, None]: ...
+
+    @abstractmethod
+    async def get_scores(
+        self,
+        batch_input: List[str],
+        **input_kwargs,
+    ) -> List[float]: ...

llama-factory/src/llamafactory/chat/chat_model.py

@@ -0,0 +1,155 @@
+# Copyright 2024 THUDM and the LlamaFactory team.
+#
+# This code is inspired by the THUDM's ChatGLM implementation.
+# https://github.com/THUDM/ChatGLM-6B/blob/main/cli_demo.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 asyncio
+import os
+from threading import Thread
+from typing import TYPE_CHECKING, Any, AsyncGenerator, Dict, Generator, List, Optional, Sequence
+
+from ..extras.misc import torch_gc
+from ..hparams import get_infer_args
+from .hf_engine import HuggingfaceEngine
+from .vllm_engine import VllmEngine
+
+
+if TYPE_CHECKING:
+    from numpy.typing import NDArray
+
+    from .base_engine import BaseEngine, Response
+
+
+def _start_background_loop(loop: "asyncio.AbstractEventLoop") -> None:
+    asyncio.set_event_loop(loop)
+    loop.run_forever()
+
+
+class ChatModel:
+    def __init__(self, args: Optional[Dict[str, Any]] = None) -> None:
+        model_args, data_args, finetuning_args, generating_args = get_infer_args(args)
+        if model_args.infer_backend == "huggingface":
+            self.engine: "BaseEngine" = HuggingfaceEngine(model_args, data_args, finetuning_args, generating_args)
+        elif model_args.infer_backend == "vllm":
+            self.engine: "BaseEngine" = VllmEngine(model_args, data_args, finetuning_args, generating_args)
+        else:
+            raise NotImplementedError("Unknown backend: {}".format(model_args.infer_backend))
+
+        self._loop = asyncio.new_event_loop()
+        self._thread = Thread(target=_start_background_loop, args=(self._loop,), daemon=True)
+        self._thread.start()
+
+    def chat(
+        self,
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        **input_kwargs,
+    ) -> List["Response"]:
+        task = asyncio.run_coroutine_threadsafe(self.achat(messages, system, tools, image, **input_kwargs), self._loop)
+        return task.result()
+
+    async def achat(
+        self,
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        **input_kwargs,
+    ) -> List["Response"]:
+        return await self.engine.chat(messages, system, tools, image, **input_kwargs)
+
+    def stream_chat(
+        self,
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        **input_kwargs,
+    ) -> Generator[str, None, None]:
+        generator = self.astream_chat(messages, system, tools, image, **input_kwargs)
+        while True:
+            try:
+                task = asyncio.run_coroutine_threadsafe(generator.__anext__(), self._loop)
+                yield task.result()
+            except StopAsyncIteration:
+                break
+
+    async def astream_chat(
+        self,
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        **input_kwargs,
+    ) -> AsyncGenerator[str, None]:
+        async for new_token in self.engine.stream_chat(messages, system, tools, image, **input_kwargs):
+            yield new_token
+
+    def get_scores(
+        self,
+        batch_input: List[str],
+        **input_kwargs,
+    ) -> List[float]:
+        task = asyncio.run_coroutine_threadsafe(self.aget_scores(batch_input, **input_kwargs), self._loop)
+        return task.result()
+
+    async def aget_scores(
+        self,
+        batch_input: List[str],
+        **input_kwargs,
+    ) -> List[float]:
+        return await self.engine.get_scores(batch_input, **input_kwargs)
+
+
+def run_chat() -> None:
+    if os.name != "nt":
+        try:
+            import readline  # noqa: F401
+        except ImportError:
+            print("Install `readline` for a better experience.")
+
+    chat_model = ChatModel()
+    messages = []
+    print("Welcome to the CLI application, use `clear` to remove the history, use `exit` to exit the application.")
+
+    while True:
+        try:
+            query = input("\nUser: ")
+        except UnicodeDecodeError:
+            print("Detected decoding error at the inputs, please set the terminal encoding to utf-8.")
+            continue
+        except Exception:
+            raise
+
+        if query.strip() == "exit":
+            break
+
+        if query.strip() == "clear":
+            messages = []
+            torch_gc()
+            print("History has been removed.")
+            continue
+
+        messages.append({"role": "user", "content": query})
+        print("Assistant: ", end="", flush=True)
+
+        response = ""
+        for new_text in chat_model.stream_chat(messages):
+            print(new_text, end="", flush=True)
+            response += new_text
+        print()
+        messages.append({"role": "assistant", "content": response})

llama-factory/src/llamafactory/chat/hf_engine.py

@@ -0,0 +1,343 @@
+# 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 asyncio
+import concurrent.futures
+import os
+from threading import Thread
+from typing import TYPE_CHECKING, Any, AsyncGenerator, Callable, Dict, List, Optional, Sequence, Tuple, Union
+
+import torch
+from transformers import GenerationConfig, TextIteratorStreamer
+
+from ..data import get_template_and_fix_tokenizer
+from ..extras.logging import get_logger
+from ..extras.misc import get_logits_processor
+from ..model import load_model, load_tokenizer
+from .base_engine import BaseEngine, Response
+
+
+if TYPE_CHECKING:
+    from numpy.typing import NDArray
+    from transformers import PreTrainedModel, PreTrainedTokenizer, ProcessorMixin
+    from transformers.image_processing_utils import BaseImageProcessor
+    from trl import PreTrainedModelWrapper
+
+    from ..data import Template
+    from ..hparams import DataArguments, FinetuningArguments, GeneratingArguments, ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+class HuggingfaceEngine(BaseEngine):
+    def __init__(
+        self,
+        model_args: "ModelArguments",
+        data_args: "DataArguments",
+        finetuning_args: "FinetuningArguments",
+        generating_args: "GeneratingArguments",
+    ) -> None:
+        self.can_generate = finetuning_args.stage == "sft"
+        tokenizer_module = load_tokenizer(model_args)
+        self.tokenizer = tokenizer_module["tokenizer"]
+        self.processor = tokenizer_module["processor"]
+        self.tokenizer.padding_side = "left" if self.can_generate else "right"
+        self.template = get_template_and_fix_tokenizer(self.tokenizer, data_args.template, data_args.tool_format)
+        self.model = load_model(
+            self.tokenizer, model_args, finetuning_args, is_trainable=False, add_valuehead=(not self.can_generate)
+        )  # must after fixing tokenizer to resize vocab
+        self.generating_args = generating_args.to_dict()
+        try:
+            asyncio.get_event_loop()
+        except RuntimeError:
+            logger.warning("There is no current event loop, creating a new one.")
+            loop = asyncio.new_event_loop()
+            asyncio.set_event_loop(loop)
+
+        self.semaphore = asyncio.Semaphore(int(os.environ.get("MAX_CONCURRENT", "1")))
+
+    @staticmethod
+    def _process_args(
+        model: "PreTrainedModel",
+        tokenizer: "PreTrainedTokenizer",
+        processor: Optional["ProcessorMixin"],
+        template: "Template",
+        generating_args: Dict[str, Any],
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        input_kwargs: Optional[Dict[str, Any]] = {},
+    ) -> Tuple[Dict[str, Any], int]:
+        if (
+            processor is not None
+            and image is not None
+            and not hasattr(processor, "image_seq_length")
+            and template.image_token not in messages[0]["content"]
+        ):  # llava-like models
+            messages[0]["content"] = template.image_token + messages[0]["content"]
+
+        paired_messages = messages + [{"role": "assistant", "content": ""}]
+        system = system or generating_args["default_system"]
+        pixel_values = None
+        prompt_ids, _ = template.encode_oneturn(
+            tokenizer=tokenizer, messages=paired_messages, system=system, tools=tools
+        )
+        if processor is not None and image is not None:  # add image features
+            image_processor: "BaseImageProcessor" = getattr(processor, "image_processor")
+            batch_feature = image_processor(image, return_tensors="pt")
+            pixel_values = batch_feature.to(model.device)["pixel_values"]  # shape (B, C, H, W)
+            if hasattr(processor, "image_seq_length"):  # paligemma models
+                image_token_id = tokenizer.convert_tokens_to_ids(template.image_token)
+                prompt_ids = [image_token_id] * getattr(processor, "image_seq_length") + prompt_ids
+
+        prompt_length = len(prompt_ids)
+        inputs = torch.tensor([prompt_ids], device=model.device)
+        attention_mask = torch.ones_like(inputs, dtype=torch.bool)
+
+        do_sample: Optional[bool] = input_kwargs.pop("do_sample", None)
+        temperature: Optional[float] = input_kwargs.pop("temperature", None)
+        top_p: Optional[float] = input_kwargs.pop("top_p", None)
+        top_k: Optional[float] = input_kwargs.pop("top_k", None)
+        num_return_sequences: int = input_kwargs.pop("num_return_sequences", 1)
+        repetition_penalty: Optional[float] = input_kwargs.pop("repetition_penalty", None)
+        length_penalty: Optional[float] = input_kwargs.pop("length_penalty", None)
+        max_length: Optional[int] = input_kwargs.pop("max_length", None)
+        max_new_tokens: Optional[int] = input_kwargs.pop("max_new_tokens", None)
+        stop: Optional[Union[str, List[str]]] = input_kwargs.pop("stop", None)
+
+        if stop is not None:
+            logger.warning("Stop parameter is not supported by the huggingface engine yet.")
+
+        generating_args = generating_args.copy()
+        generating_args.update(
+            dict(
+                do_sample=do_sample if do_sample is not None else generating_args["do_sample"],
+                temperature=temperature if temperature is not None else generating_args["temperature"],
+                top_p=top_p if top_p is not None else generating_args["top_p"],
+                top_k=top_k if top_k is not None else generating_args["top_k"],
+                num_return_sequences=num_return_sequences,
+                repetition_penalty=repetition_penalty
+                if repetition_penalty is not None
+                else generating_args["repetition_penalty"],
+                length_penalty=length_penalty if length_penalty is not None else generating_args["length_penalty"],
+                eos_token_id=[tokenizer.eos_token_id] + tokenizer.additional_special_tokens_ids,
+                pad_token_id=tokenizer.pad_token_id,
+            )
+        )
+
+        if isinstance(num_return_sequences, int) and num_return_sequences > 1:  # do_sample needs temperature > 0
+            generating_args["do_sample"] = True
+            generating_args["temperature"] = generating_args["temperature"] or 1.0
+
+        if not generating_args["temperature"]:
+            generating_args["do_sample"] = False
+
+        if not generating_args["do_sample"]:
+            generating_args.pop("temperature", None)
+            generating_args.pop("top_p", None)
+
+        if max_length:
+            generating_args.pop("max_new_tokens", None)
+            generating_args["max_length"] = max_length
+
+        if max_new_tokens:
+            generating_args.pop("max_length", None)
+            generating_args["max_new_tokens"] = max_new_tokens
+
+        gen_kwargs = dict(
+            inputs=inputs,
+            attention_mask=attention_mask,
+            generation_config=GenerationConfig(**generating_args),
+            logits_processor=get_logits_processor(),
+        )
+
+        if pixel_values is not None:
+            gen_kwargs["pixel_values"] = pixel_values
+
+        return gen_kwargs, prompt_length
+
+    @staticmethod
+    @torch.inference_mode()
+    def _chat(
+        model: "PreTrainedModel",
+        tokenizer: "PreTrainedTokenizer",
+        processor: Optional["ProcessorMixin"],
+        template: "Template",
+        generating_args: Dict[str, Any],
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        input_kwargs: Optional[Dict[str, Any]] = {},
+    ) -> List["Response"]:
+        gen_kwargs, prompt_length = HuggingfaceEngine._process_args(
+            model, tokenizer, processor, template, generating_args, messages, system, tools, image, input_kwargs
+        )
+        generate_output = model.generate(**gen_kwargs)
+        response_ids = generate_output[:, prompt_length:]
+        response = tokenizer.batch_decode(response_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True)
+        results = []
+        for i in range(len(response)):
+            eos_index = (response_ids[i] == tokenizer.eos_token_id).nonzero()
+            response_length = (eos_index[0].item() + 1) if len(eos_index) else len(response_ids[i])
+            results.append(
+                Response(
+                    response_text=response[i],
+                    response_length=response_length,
+                    prompt_length=prompt_length,
+                    finish_reason="stop" if len(eos_index) else "length",
+                )
+            )
+
+        return results
+
+    @staticmethod
+    @torch.inference_mode()
+    def _stream_chat(
+        model: "PreTrainedModel",
+        tokenizer: "PreTrainedTokenizer",
+        processor: Optional["ProcessorMixin"],
+        template: "Template",
+        generating_args: Dict[str, Any],
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        input_kwargs: Optional[Dict[str, Any]] = {},
+    ) -> Callable[[], str]:
+        gen_kwargs, _ = HuggingfaceEngine._process_args(
+            model, tokenizer, processor, template, generating_args, messages, system, tools, image, input_kwargs
+        )
+        streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
+        gen_kwargs["streamer"] = streamer
+        thread = Thread(target=model.generate, kwargs=gen_kwargs, daemon=True)
+        thread.start()
+
+        def stream():
+            try:
+                return streamer.__next__()
+            except StopIteration:
+                raise StopAsyncIteration()
+
+        return stream
+
+    @staticmethod
+    @torch.inference_mode()
+    def _get_scores(
+        model: "PreTrainedModelWrapper",
+        tokenizer: "PreTrainedTokenizer",
+        batch_input: List[str],
+        input_kwargs: Optional[Dict[str, Any]] = {},
+    ) -> List[float]:
+        max_length = input_kwargs.pop("max_length", None)
+        device = getattr(model.pretrained_model, "device", "cuda")
+        inputs = tokenizer(
+            batch_input,
+            padding=True,
+            truncation=True,
+            max_length=max_length or getattr(model.config, "max_position_embeddings", 1024),
+            return_tensors="pt",
+            add_special_tokens=True,
+        ).to(device)
+
+        input_ids: torch.Tensor = inputs["input_ids"]
+        _, _, values = model(**inputs, output_hidden_states=True, return_dict=True)
+
+        if getattr(model.config, "model_type", None) == "chatglm":
+            values = torch.transpose(values, 0, 1)
+
+        scores = []
+        for i in range(input_ids.size(0)):
+            end_indexes = (input_ids[i] != tokenizer.pad_token_id).nonzero()
+            end_index = end_indexes[-1].item() if len(end_indexes) else 0
+            scores.append(values[i, end_index].nan_to_num().item())
+
+        return scores
+
+    async def chat(
+        self,
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        **input_kwargs,
+    ) -> List["Response"]:
+        if not self.can_generate:
+            raise ValueError("The current model does not support `chat`.")
+
+        loop = asyncio.get_running_loop()
+        input_args = (
+            self.model,
+            self.tokenizer,
+            self.processor,
+            self.template,
+            self.generating_args,
+            messages,
+            system,
+            tools,
+            image,
+            input_kwargs,
+        )
+        async with self.semaphore:
+            with concurrent.futures.ThreadPoolExecutor() as pool:
+                return await loop.run_in_executor(pool, self._chat, *input_args)
+
+    async def stream_chat(
+        self,
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        **input_kwargs,
+    ) -> AsyncGenerator[str, None]:
+        if not self.can_generate:
+            raise ValueError("The current model does not support `stream_chat`.")
+
+        loop = asyncio.get_running_loop()
+        input_args = (
+            self.model,
+            self.tokenizer,
+            self.processor,
+            self.template,
+            self.generating_args,
+            messages,
+            system,
+            tools,
+            image,
+            input_kwargs,
+        )
+        async with self.semaphore:
+            with concurrent.futures.ThreadPoolExecutor() as pool:
+                stream = self._stream_chat(*input_args)
+                while True:
+                    try:
+                        yield await loop.run_in_executor(pool, stream)
+                    except StopAsyncIteration:
+                        break
+
+    async def get_scores(
+        self,
+        batch_input: List[str],
+        **input_kwargs,
+    ) -> List[float]:
+        if self.can_generate:
+            raise ValueError("Cannot get scores using an auto-regressive model.")
+
+        loop = asyncio.get_running_loop()
+        input_args = (self.model, self.tokenizer, batch_input, input_kwargs)
+        async with self.semaphore:
+            with concurrent.futures.ThreadPoolExecutor() as pool:
+                return await loop.run_in_executor(pool, self._get_scores, *input_args)

llama-factory/src/llamafactory/chat/vllm_engine.py

@@ -0,0 +1,242 @@
+# 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 uuid
+from typing import TYPE_CHECKING, Any, AsyncGenerator, AsyncIterator, Dict, List, Optional, Sequence, Union
+
+from ..data import get_template_and_fix_tokenizer
+from ..extras.logging import get_logger
+from ..extras.misc import get_device_count
+from ..extras.packages import is_vllm_available, is_vllm_version_greater_than_0_5, is_vllm_version_greater_than_0_5_1
+from ..model import load_config, load_tokenizer
+from ..model.model_utils.quantization import QuantizationMethod
+from ..model.model_utils.visual import LlavaMultiModalProjectorForYiVLForVLLM
+from .base_engine import BaseEngine, Response
+
+
+if is_vllm_available():
+    from vllm import AsyncEngineArgs, AsyncLLMEngine, RequestOutput, SamplingParams
+    from vllm.lora.request import LoRARequest
+
+    if is_vllm_version_greater_than_0_5_1():
+        pass
+    elif is_vllm_version_greater_than_0_5():
+        from vllm.multimodal.image import ImagePixelData
+    else:
+        from vllm.sequence import MultiModalData
+
+
+if TYPE_CHECKING:
+    from numpy.typing import NDArray
+    from transformers.image_processing_utils import BaseImageProcessor
+
+    from ..hparams import DataArguments, FinetuningArguments, GeneratingArguments, ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+class VllmEngine(BaseEngine):
+    def __init__(
+        self,
+        model_args: "ModelArguments",
+        data_args: "DataArguments",
+        finetuning_args: "FinetuningArguments",
+        generating_args: "GeneratingArguments",
+    ) -> None:
+        config = load_config(model_args)  # may download model from ms hub
+        if getattr(config, "quantization_config", None):  # gptq models should use float16
+            quantization_config: Dict[str, Any] = getattr(config, "quantization_config", None)
+            quant_method = quantization_config.get("quant_method", "")
+            if quant_method == QuantizationMethod.GPTQ and model_args.infer_dtype == "auto":
+                model_args.infer_dtype = "float16"
+
+        self.can_generate = finetuning_args.stage == "sft"
+        tokenizer_module = load_tokenizer(model_args)
+        self.tokenizer = tokenizer_module["tokenizer"]
+        self.processor = tokenizer_module["processor"]
+        self.tokenizer.padding_side = "left"
+        self.template = get_template_and_fix_tokenizer(self.tokenizer, data_args.template, data_args.tool_format)
+        self.generating_args = generating_args.to_dict()
+
+        engine_args = {
+            "model": model_args.model_name_or_path,
+            "trust_remote_code": True,
+            "download_dir": model_args.cache_dir,
+            "dtype": model_args.infer_dtype,
+            "max_model_len": model_args.vllm_maxlen,
+            "tensor_parallel_size": get_device_count() or 1,
+            "gpu_memory_utilization": model_args.vllm_gpu_util,
+            "disable_log_stats": True,
+            "disable_log_requests": True,
+            "enforce_eager": model_args.vllm_enforce_eager,
+            "enable_lora": model_args.adapter_name_or_path is not None,
+            "max_lora_rank": model_args.vllm_max_lora_rank,
+        }
+
+        if model_args.visual_inputs:
+            image_size = config.vision_config.image_size
+            patch_size = config.vision_config.patch_size
+            self.image_feature_size = (image_size // patch_size) ** 2
+            engine_args["image_input_type"] = "pixel_values"
+            engine_args["image_token_id"] = self.tokenizer.convert_tokens_to_ids(self.template.image_token)
+            engine_args["image_input_shape"] = "1,3,{},{}".format(image_size, image_size)
+            engine_args["image_feature_size"] = self.image_feature_size
+            if getattr(config, "is_yi_vl_derived_model", None):
+                import vllm.model_executor.models.llava
+
+                logger.info("Detected Yi-VL model, applying projector patch.")
+                vllm.model_executor.models.llava.LlavaMultiModalProjector = LlavaMultiModalProjectorForYiVLForVLLM
+
+        self.model = AsyncLLMEngine.from_engine_args(AsyncEngineArgs(**engine_args))
+        if model_args.adapter_name_or_path is not None:
+            self.lora_request = LoRARequest("default", 1, model_args.adapter_name_or_path[0])
+        else:
+            self.lora_request = None
+
+    async def _generate(
+        self,
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        **input_kwargs,
+    ) -> AsyncIterator["RequestOutput"]:
+        request_id = "chatcmpl-{}".format(uuid.uuid4().hex)
+
+        if (
+            self.processor is not None
+            and image is not None
+            and not hasattr(self.processor, "image_seq_length")
+            and self.template.image_token not in messages[0]["content"]
+        ):  # llava-like models (TODO: paligemma models)
+            messages[0]["content"] = self.template.image_token * self.image_feature_size + messages[0]["content"]
+
+        paired_messages = messages + [{"role": "assistant", "content": ""}]
+        system = system or self.generating_args["default_system"]
+        prompt_ids, _ = self.template.encode_oneturn(
+            tokenizer=self.tokenizer, messages=paired_messages, system=system, tools=tools
+        )
+
+        if self.processor is not None and image is not None:  # add image features
+            image_processor: "BaseImageProcessor" = getattr(self.processor, "image_processor")
+            pixel_values = image_processor(image, return_tensors="pt")["pixel_values"]
+            if is_vllm_version_greater_than_0_5_1():
+                multi_modal_data = {"image": pixel_values}
+            elif is_vllm_version_greater_than_0_5():
+                multi_modal_data = ImagePixelData(image=pixel_values)
+            else:  # TODO: remove vllm 0.4.3 support
+                multi_modal_data = MultiModalData(type=MultiModalData.Type.IMAGE, data=pixel_values)
+        else:
+            multi_modal_data = None
+
+        prompt_length = len(prompt_ids)
+
+        use_beam_search: bool = self.generating_args["num_beams"] > 1
+        temperature: Optional[float] = input_kwargs.pop("temperature", None)
+        top_p: Optional[float] = input_kwargs.pop("top_p", None)
+        top_k: Optional[float] = input_kwargs.pop("top_k", None)
+        num_return_sequences: int = input_kwargs.pop("num_return_sequences", 1)
+        repetition_penalty: Optional[float] = input_kwargs.pop("repetition_penalty", None)
+        length_penalty: Optional[float] = input_kwargs.pop("length_penalty", None)
+        max_length: Optional[int] = input_kwargs.pop("max_length", None)
+        max_new_tokens: Optional[int] = input_kwargs.pop("max_new_tokens", None)
+        stop: Optional[Union[str, List[str]]] = input_kwargs.pop("stop", None)
+
+        if "max_new_tokens" in self.generating_args:
+            max_tokens = self.generating_args["max_new_tokens"]
+        elif "max_length" in self.generating_args:
+            if self.generating_args["max_length"] > prompt_length:
+                max_tokens = self.generating_args["max_length"] - prompt_length
+            else:
+                max_tokens = 1
+
+        if max_length:
+            max_tokens = max_length - prompt_length if max_length > prompt_length else 1
+
+        if max_new_tokens:
+            max_tokens = max_new_tokens
+
+        sampling_params = SamplingParams(
+            n=num_return_sequences,
+            repetition_penalty=(
+                repetition_penalty if repetition_penalty is not None else self.generating_args["repetition_penalty"]
+            )
+            or 1.0,  # repetition_penalty must > 0
+            temperature=temperature if temperature is not None else self.generating_args["temperature"],
+            top_p=(top_p if top_p is not None else self.generating_args["top_p"]) or 1.0,  # top_p must > 0
+            top_k=top_k if top_k is not None else self.generating_args["top_k"],
+            use_beam_search=use_beam_search,
+            length_penalty=length_penalty if length_penalty is not None else self.generating_args["length_penalty"],
+            stop=stop,
+            stop_token_ids=[self.tokenizer.eos_token_id] + self.tokenizer.additional_special_tokens_ids,
+            max_tokens=max_tokens,
+            skip_special_tokens=True,
+        )
+
+        result_generator = self.model.generate(
+            inputs={"prompt_token_ids": prompt_ids, "multi_modal_data": multi_modal_data},
+            sampling_params=sampling_params,
+            request_id=request_id,
+            lora_request=self.lora_request,
+        )
+        return result_generator
+
+    async def chat(
+        self,
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        **input_kwargs,
+    ) -> List["Response"]:
+        final_output = None
+        generator = await self._generate(messages, system, tools, image, **input_kwargs)
+        async for request_output in generator:
+            final_output = request_output
+
+        results = []
+        for output in final_output.outputs:
+            results.append(
+                Response(
+                    response_text=output.text,
+                    response_length=len(output.token_ids),
+                    prompt_length=len(final_output.prompt_token_ids),
+                    finish_reason=output.finish_reason,
+                )
+            )
+
+        return results
+
+    async def stream_chat(
+        self,
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+        image: Optional["NDArray"] = None,
+        **input_kwargs,
+    ) -> AsyncGenerator[str, None]:
+        generated_text = ""
+        generator = await self._generate(messages, system, tools, image, **input_kwargs)
+        async for result in generator:
+            delta_text = result.outputs[0].text[len(generated_text) :]
+            generated_text = result.outputs[0].text
+            yield delta_text
+
+    async def get_scores(
+        self,
+        batch_input: List[str],
+        **input_kwargs,
+    ) -> List[float]:
+        raise NotImplementedError("vLLM engine does not support get_scores.")

llama-factory/src/llamafactory/cli.py

@@ -0,0 +1,121 @@
+# 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 random
+import subprocess
+import sys
+from enum import Enum, unique
+
+from . import launcher
+from .api.app import run_api
+from .chat.chat_model import run_chat
+from .eval.evaluator import run_eval
+from .extras.env import VERSION, print_env
+from .extras.logging import get_logger
+from .extras.misc import get_device_count
+from .train.tuner import export_model, run_exp
+from .webui.interface import run_web_demo, run_web_ui
+
+
+USAGE = (
+    "-" * 70
+    + "\n"
+    + "| Usage:                                                             |\n"
+    + "|   llamafactory-cli api -h: launch an OpenAI-style API server       |\n"
+    + "|   llamafactory-cli chat -h: launch a chat interface in CLI         |\n"
+    + "|   llamafactory-cli eval -h: evaluate models                        |\n"
+    + "|   llamafactory-cli export -h: merge LoRA adapters and export model |\n"
+    + "|   llamafactory-cli train -h: train models                          |\n"
+    + "|   llamafactory-cli webchat -h: launch a chat interface in Web UI   |\n"
+    + "|   llamafactory-cli webui: launch LlamaBoard                        |\n"
+    + "|   llamafactory-cli version: show version info                      |\n"
+    + "-" * 70
+)
+
+WELCOME = (
+    "-" * 58
+    + "\n"
+    + "| Welcome to LLaMA Factory, version {}".format(VERSION)
+    + " " * (21 - len(VERSION))
+    + "|\n|"
+    + " " * 56
+    + "|\n"
+    + "| Project page: https://github.com/hiyouga/LLaMA-Factory |\n"
+    + "-" * 58
+)
+
+logger = get_logger(__name__)
+
+
+@unique
+class Command(str, Enum):
+    API = "api"
+    CHAT = "chat"
+    ENV = "env"
+    EVAL = "eval"
+    EXPORT = "export"
+    TRAIN = "train"
+    WEBDEMO = "webchat"
+    WEBUI = "webui"
+    VER = "version"
+    HELP = "help"
+
+
+def main():
+    command = sys.argv.pop(1) if len(sys.argv) != 1 else Command.HELP
+    if command == Command.API:
+        run_api()
+    elif command == Command.CHAT:
+        run_chat()
+    elif command == Command.ENV:
+        print_env()
+    elif command == Command.EVAL:
+        run_eval()
+    elif command == Command.EXPORT:
+        export_model()
+    elif command == Command.TRAIN:
+        force_torchrun = os.environ.get("FORCE_TORCHRUN", "0").lower() in ["true", "1"]
+        if force_torchrun or get_device_count() > 1:
+            master_addr = os.environ.get("MASTER_ADDR", "127.0.0.1")
+            master_port = os.environ.get("MASTER_PORT", str(random.randint(20001, 29999)))
+            logger.info("Initializing distributed tasks at: {}:{}".format(master_addr, master_port))
+            process = subprocess.run(
+                (
+                    "torchrun --nnodes {nnodes} --node_rank {node_rank} --nproc_per_node {nproc_per_node} "
+                    "--master_addr {master_addr} --master_port {master_port} {file_name} {args}"
+                ).format(
+                    nnodes=os.environ.get("NNODES", "1"),
+                    node_rank=os.environ.get("RANK", "0"),
+                    nproc_per_node=os.environ.get("NPROC_PER_NODE", str(get_device_count())),
+                    master_addr=master_addr,
+                    master_port=master_port,
+                    file_name=launcher.__file__,
+                    args=" ".join(sys.argv[1:]),
+                ),
+                shell=True,
+            )
+            sys.exit(process.returncode)
+        else:
+            run_exp()
+    elif command == Command.WEBDEMO:
+        run_web_demo()
+    elif command == Command.WEBUI:
+        run_web_ui()
+    elif command == Command.VER:
+        print(WELCOME)
+    elif command == Command.HELP:
+        print(USAGE)
+    else:
+        raise NotImplementedError("Unknown command: {}".format(command))

llama-factory/src/llamafactory/data/__init__.py

@@ -0,0 +1,31 @@
+# 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 .collator import KTODataCollatorWithPadding, PairwiseDataCollatorWithPadding, SFTDataCollatorWith4DAttentionMask
+from .data_utils import Role, split_dataset
+from .loader import get_dataset
+from .template import TEMPLATES, Template, get_template_and_fix_tokenizer
+
+
+__all__ = [
+    "KTODataCollatorWithPadding",
+    "PairwiseDataCollatorWithPadding",
+    "SFTDataCollatorWith4DAttentionMask",
+    "Role",
+    "split_dataset",
+    "get_dataset",
+    "TEMPLATES",
+    "Template",
+    "get_template_and_fix_tokenizer",
+]

llama-factory/src/llamafactory/data/aligner.py

@@ -0,0 +1,239 @@
+# 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 functools import partial
+from typing import TYPE_CHECKING, Any, Dict, List, Union
+
+from datasets import Features
+
+from ..extras.logging import get_logger
+from .data_utils import Role
+
+
+if TYPE_CHECKING:
+    from datasets import Dataset, IterableDataset
+    from transformers import Seq2SeqTrainingArguments
+
+    from ..hparams import DataArguments
+    from .parser import DatasetAttr
+
+
+logger = get_logger(__name__)
+
+
+def _convert_images(images: List[Any], dataset_attr: "DatasetAttr", data_args: "DataArguments") -> List[Any]:
+    r"""
+    Optionally concatenates image path to dataset dir when loading from local disk.
+    """
+    outputs = []
+    if dataset_attr.load_from in ["script", "file"]:
+        for image in images:
+            if isinstance(image, str) and os.path.isfile(os.path.join(data_args.dataset_dir, image)):
+                outputs.append(os.path.join(data_args.dataset_dir, image))
+            else:
+                outputs.append(image)
+
+    return outputs
+
+
+def convert_alpaca(
+    examples: Dict[str, List[Any]], dataset_attr: "DatasetAttr", data_args: "DataArguments"
+) -> Dict[str, List[Any]]:
+    r"""
+    Converts alpaca format dataset to the standard format.
+    """
+    outputs = {"prompt": [], "response": [], "system": [], "tools": [], "images": []}
+    convert_images = partial(_convert_images, dataset_attr=dataset_attr, data_args=data_args)
+    for i in range(len(examples[dataset_attr.prompt])):
+        prompt = []
+        if dataset_attr.history and isinstance(examples[dataset_attr.history][i], list):
+            for old_prompt, old_response in examples[dataset_attr.history][i]:
+                prompt.append({"role": Role.USER.value, "content": old_prompt})
+                prompt.append({"role": Role.ASSISTANT.value, "content": old_response})
+
+        content = []
+        if dataset_attr.prompt and examples[dataset_attr.prompt][i]:
+            content.append(examples[dataset_attr.prompt][i])
+
+        if dataset_attr.query and examples[dataset_attr.query][i]:
+            content.append(examples[dataset_attr.query][i])
+
+        prompt.append({"role": Role.USER.value, "content": "\n".join(content)})  # "prompt\nquery"
+
+        if dataset_attr.kto_tag and isinstance(examples[dataset_attr.kto_tag][i], bool):  # kto example
+            response = [{"role": Role.ASSISTANT.value, "content": examples[dataset_attr.response][i]}]
+            if examples[dataset_attr.kto_tag][i]:
+                response = response + [{"role": Role.ASSISTANT.value, "content": ""}]
+            else:
+                response = [{"role": Role.ASSISTANT.value, "content": ""}] + response
+        elif (
+            dataset_attr.ranking
+            and isinstance(examples[dataset_attr.chosen][i], str)
+            and isinstance(examples[dataset_attr.rejected][i], str)
+        ):  # pairwise example
+            response = [
+                {"role": Role.ASSISTANT.value, "content": examples[dataset_attr.chosen][i]},
+                {"role": Role.ASSISTANT.value, "content": examples[dataset_attr.rejected][i]},
+            ]
+        elif dataset_attr.response and isinstance(examples[dataset_attr.response][i], str):  # normal example
+            response = [{"role": Role.ASSISTANT.value, "content": examples[dataset_attr.response][i]}]
+        else:  # unsupervised
+            response = []
+
+        outputs["prompt"].append(prompt)
+        outputs["response"].append(response)
+        outputs["system"].append(examples[dataset_attr.system][i] if dataset_attr.system else "")
+        outputs["tools"].append(examples[dataset_attr.tools][i] if dataset_attr.tools else "")
+        outputs["images"].append(convert_images(examples[dataset_attr.images][i]) if dataset_attr.images else [])
+
+    return outputs
+
+
+def convert_sharegpt(
+    examples: Dict[str, List[Any]], dataset_attr: "DatasetAttr", data_args: "DataArguments"
+) -> Dict[str, List[Any]]:
+    r"""
+    Converts sharegpt format dataset to the standard format.
+    """
+    outputs = {"prompt": [], "response": [], "system": [], "tools": [], "images": []}
+    convert_images = partial(_convert_images, dataset_attr=dataset_attr, data_args=data_args)
+    tag_mapping = {
+        dataset_attr.user_tag: Role.USER.value,
+        dataset_attr.assistant_tag: Role.ASSISTANT.value,
+        dataset_attr.observation_tag: Role.OBSERVATION.value,
+        dataset_attr.function_tag: Role.FUNCTION.value,
+        dataset_attr.system_tag: Role.SYSTEM.value,
+    }
+    odd_tags = (dataset_attr.user_tag, dataset_attr.observation_tag)
+    even_tags = (dataset_attr.assistant_tag, dataset_attr.function_tag)
+    accept_tags = (odd_tags, even_tags)
+    for i, messages in enumerate(examples[dataset_attr.messages]):
+        if dataset_attr.system_tag and messages[0][dataset_attr.role_tag] == dataset_attr.system_tag:
+            system = messages[0][dataset_attr.content_tag]
+            messages = messages[1:]
+        else:
+            system = examples[dataset_attr.system][i] if dataset_attr.system else ""
+
+        if len(messages) == 0:
+            continue
+
+        aligned_messages = []
+        broken_data = False
+        for turn_idx, message in enumerate(messages):
+            if message[dataset_attr.role_tag] not in accept_tags[turn_idx % 2]:
+                logger.warning("Invalid role tag in {}.".format(messages))
+                broken_data = True
+
+            aligned_messages.append(
+                {"role": tag_mapping[message[dataset_attr.role_tag]], "content": message[dataset_attr.content_tag]}
+            )
+
+        if (not dataset_attr.ranking and len(aligned_messages) % 2 != 0) or (
+            dataset_attr.ranking and len(aligned_messages) % 2 == 0
+        ):
+            logger.warning("Invalid message count in {}.".format(messages))
+            broken_data = True
+
+        if dataset_attr.kto_tag and isinstance(examples[dataset_attr.kto_tag][i], bool):  # kto example
+            prompt = aligned_messages[:-1]
+            response = aligned_messages[-1:]
+            if examples[dataset_attr.kto_tag][i]:
+                response = response + [{"role": Role.ASSISTANT.value, "content": ""}]
+            else:
+                response = [{"role": Role.ASSISTANT.value, "content": ""}] + response
+        elif (
+            dataset_attr.ranking
+            and isinstance(examples[dataset_attr.chosen][i], dict)
+            and isinstance(examples[dataset_attr.rejected][i], dict)
+        ):  # pairwise example
+            chosen = examples[dataset_attr.chosen][i]
+            rejected = examples[dataset_attr.rejected][i]
+            if (
+                chosen[dataset_attr.role_tag] not in accept_tags[-1]
+                or rejected[dataset_attr.role_tag] not in accept_tags[-1]
+            ):
+                logger.warning("Invalid role tag in {}.".format([chosen, rejected]))
+                broken_data = True
+
+            prompt = aligned_messages
+            response = [
+                {"role": tag_mapping[chosen[dataset_attr.role_tag]], "content": chosen[dataset_attr.content_tag]},
+                {"role": tag_mapping[rejected[dataset_attr.role_tag]], "content": rejected[dataset_attr.content_tag]},
+            ]
+        else:  # normal example
+            prompt = aligned_messages[:-1]
+            response = aligned_messages[-1:]
+
+        if broken_data:
+            logger.warning("Skipping this abnormal example.")
+            continue
+
+        outputs["prompt"].append(prompt)
+        outputs["response"].append(response)
+        outputs["system"].append(system)
+        outputs["tools"].append(examples[dataset_attr.tools][i] if dataset_attr.tools else "")
+        outputs["images"].append(convert_images(examples[dataset_attr.images][i]) if dataset_attr.images else [])
+
+    return outputs
+
+
+def align_dataset(
+    dataset: Union["Dataset", "IterableDataset"],
+    dataset_attr: "DatasetAttr",
+    data_args: "DataArguments",
+    training_args: "Seq2SeqTrainingArguments",
+) -> Union["Dataset", "IterableDataset"]:
+    r"""
+    Aligned dataset:
+        prompt: [{"role": "user", "content": "..."}] * (2T - 1)
+        response: [{"role": "assistant", "content": "..."}] * N (N > 1 for ranking dataset)
+        system: "..."
+        tools: "...",
+        images: [],
+    """
+    if dataset_attr.formatting == "alpaca":
+        convert_func = partial(convert_alpaca, dataset_attr=dataset_attr, data_args=data_args)
+    else:
+        convert_func = partial(convert_sharegpt, dataset_attr=dataset_attr, data_args=data_args)
+
+    column_names = list(next(iter(dataset)).keys())
+    features = Features.from_dict(
+        {
+            "prompt": [
+                {"role": {"dtype": "string", "_type": "Value"}, "content": {"dtype": "string", "_type": "Value"}}
+            ],
+            "response": [
+                {"role": {"dtype": "string", "_type": "Value"}, "content": {"dtype": "string", "_type": "Value"}}
+            ],
+            "system": {"dtype": "string", "_type": "Value"},
+            "tools": {"dtype": "string", "_type": "Value"},
+            "images": [{"_type": "Image"}],
+        }
+    )
+    kwargs = {}
+    if not data_args.streaming:
+        kwargs = dict(
+            num_proc=data_args.preprocessing_num_workers,
+            load_from_cache_file=(not data_args.overwrite_cache) or (training_args.local_process_index != 0),
+            desc="Converting format of dataset",
+        )
+
+    return dataset.map(
+        convert_func,
+        batched=True,
+        remove_columns=column_names,
+        features=features,
+        **kwargs,
+    )

llama-factory/src/llamafactory/data/collator.py

@@ -0,0 +1,155 @@
+# Copyright 2024 OpenAccess AI Collective and the LlamaFactory team.
+#
+# This code is inspired by the OpenAccess AI Collective's axolotl library.
+# https://github.com/OpenAccess-AI-Collective/axolotl/blob/main/src/axolotl/monkeypatch/utils.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.
+
+from dataclasses import dataclass
+from typing import Any, Dict, Literal, Sequence
+
+import torch
+from transformers import DataCollatorForSeq2Seq
+
+
+def prepare_4d_attention_mask(attention_mask_with_indices: "torch.Tensor", dtype: "torch.dtype") -> "torch.Tensor":
+    r"""
+    Expands the attention mask with indices from (batch_size, seq_len) to (batch_size, 1, seq_len, seq_len),
+    while handles packed sequences and transforms the mask to lower triangular form to prevent future peeking.
+
+    e.g.
+    ```python
+    # input
+    [[1, 1, 2, 2, 2, 0]]
+    # output
+    [
+        [
+            [
+                [o, x, x, x, x, x],
+                [o, o, x, x, x, x],
+                [x, x, o, x, x, x],
+                [x, x, o, o, x, x],
+                [x, x, o, o, o, x],
+                [x, x, x, x, x, x],
+            ]
+        ]
+    ]
+    ```
+    where `o` equals to `0.0`, `x` equals to `min_dtype`.
+    """
+    bsz, seq_len = attention_mask_with_indices.size()
+    min_dtype = torch.finfo(dtype).min
+    expanded_mask = attention_mask_with_indices[:, None, None, :].expand(bsz, 1, seq_len, seq_len)
+    # Create a binary mask from the original mask where zeros remain zeros and all other values are set to one
+    padding_mask = torch.where(expanded_mask != 0, 1, 0)
+    # Create a block-diagonal mask.
+    attention_mask_4d = torch.eq(expanded_mask, expanded_mask.transpose(-1, -2)).int() * padding_mask
+    # Use the lower triangular mask to zero out the upper triangular part
+    attention_mask_4d *= torch.tril(torch.ones((seq_len, seq_len), dtype=torch.long))
+    # Invert the attention mask.
+    attention_mask_4d = torch.where(attention_mask_4d != 0, torch.tensor(0, dtype=dtype), min_dtype)
+    return attention_mask_4d
+
+
+@dataclass
+class SFTDataCollatorWith4DAttentionMask(DataCollatorForSeq2Seq):
+    r"""
+    Data collator for 4d attention mask.
+    """
+
+    block_diag_attn: bool = False
+    attn_implementation: Literal["eager", "sdpa", "flash_attention_2"] = "eager"
+    compute_dtype: "torch.dtype" = torch.float32
+
+    def __call__(self, features: Sequence[Dict[str, Any]]) -> Dict[str, "torch.Tensor"]:
+        features = super().__call__(features)
+        if self.block_diag_attn and self.attn_implementation != "flash_attention_2":
+            features["attention_mask"] = prepare_4d_attention_mask(features["attention_mask"], self.compute_dtype)
+
+        return features
+
+
+@dataclass
+class PairwiseDataCollatorWithPadding(DataCollatorForSeq2Seq):
+    r"""
+    Data collator for pairwise data.
+    """
+
+    def __call__(self, features: Sequence[Dict[str, Any]]) -> Dict[str, "torch.Tensor"]:
+        r"""
+        Pads batched data to the longest sequence in the batch.
+
+        We generate 2 * n examples where the first n examples represent chosen examples and
+        the last n examples represent rejected examples.
+        """
+        concatenated_features = []
+        for key in ("chosen", "rejected"):
+            for feature in features:
+                target_feature = {
+                    "input_ids": feature["{}_input_ids".format(key)],
+                    "attention_mask": feature["{}_attention_mask".format(key)],
+                    "labels": feature["{}_labels".format(key)],
+                }
+                if "pixel_values" in feature:
+                    target_feature["pixel_values"] = feature["pixel_values"]
+
+                if "{}_token_type_ids".format(key) in feature:
+                    target_feature["token_type_ids"] = feature["{}_token_type_ids".format(key)]
+
+                concatenated_features.append(target_feature)
+
+        return super().__call__(concatenated_features)
+
+
+@dataclass
+class KTODataCollatorWithPadding(DataCollatorForSeq2Seq):
+    r"""
+    Data collator for KTO data.
+    """
+
+    def __call__(self, features: Sequence[Dict[str, Any]]) -> Dict[str, "torch.Tensor"]:
+        target_features = []
+        kl_features = []
+        kto_tags = []
+        for feature in features:
+            target_feature = {
+                "input_ids": feature["input_ids"],
+                "attention_mask": feature["attention_mask"],
+                "labels": feature["labels"],
+            }
+            kl_feature = {
+                "input_ids": feature["kl_input_ids"],
+                "attention_mask": feature["kl_attention_mask"],
+                "labels": feature["kl_labels"],
+            }
+            if "pixel_values" in feature:
+                target_feature["pixel_values"] = feature["pixel_values"]
+
+            if "token_type_ids" in feature:
+                target_feature["token_type_ids"] = feature["token_type_ids"]
+                kl_feature["token_type_ids"] = feature["kl_token_type_ids"]
+
+            target_features.append(target_feature)
+            kl_features.append(kl_feature)
+            kto_tags.append(feature["kto_tags"])
+
+        batch = super().__call__(target_features)
+        kl_batch = super().__call__(kl_features)
+        batch["kl_input_ids"] = kl_batch["input_ids"]
+        batch["kl_attention_mask"] = kl_batch["attention_mask"]
+        batch["kl_labels"] = kl_batch["labels"]
+        if "token_type_ids" in batch:
+            batch["kl_token_type_ids"] = kl_batch["token_type_ids"]
+
+        batch["kto_tags"] = torch.tensor(kto_tags)
+        return batch

llama-factory/src/llamafactory/data/data_utils.py

@@ -0,0 +1,87 @@
+# 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 enum import Enum, unique
+from typing import TYPE_CHECKING, Dict, List, Optional, Sequence, Set, TypedDict, Union
+
+from datasets import DatasetDict, concatenate_datasets, interleave_datasets
+
+from ..extras.logging import get_logger
+
+
+if TYPE_CHECKING:
+    from datasets import Dataset, IterableDataset
+
+    from ..hparams import DataArguments
+
+
+logger = get_logger(__name__)
+
+
+SLOTS = Sequence[Union[str, Set[str], Dict[str, str]]]
+
+
+@unique
+class Role(str, Enum):
+    USER = "user"
+    ASSISTANT = "assistant"
+    SYSTEM = "system"
+    FUNCTION = "function"
+    OBSERVATION = "observation"
+
+
+class DatasetModule(TypedDict):
+    train_dataset: Optional[Union["Dataset", "IterableDataset"]]
+    eval_dataset: Optional[Union["Dataset", "IterableDataset"]]
+
+
+def merge_dataset(
+    all_datasets: List[Union["Dataset", "IterableDataset"]], data_args: "DataArguments", seed: int
+) -> Union["Dataset", "IterableDataset"]:
+    if len(all_datasets) == 1:
+        return all_datasets[0]
+    elif data_args.mix_strategy == "concat":
+        if data_args.streaming:
+            logger.warning("The samples between different datasets will not be mixed in streaming mode.")
+
+        return concatenate_datasets(all_datasets)
+    elif data_args.mix_strategy.startswith("interleave"):
+        if not data_args.streaming:
+            logger.warning("We recommend using `mix_strategy=concat` in non-streaming mode.")
+
+        return interleave_datasets(
+            datasets=all_datasets,
+            probabilities=data_args.interleave_probs,
+            seed=seed,
+            stopping_strategy="first_exhausted" if data_args.mix_strategy.endswith("under") else "all_exhausted",
+        )
+    else:
+        raise ValueError("Unknown mixing strategy.")
+
+
+def split_dataset(
+    dataset: Union["Dataset", "IterableDataset"], data_args: "DataArguments", seed: int
+) -> "DatasetDict":
+    r"""
+    Splits the dataset and returns a dataset dict containing train set (required) and validation set (optional).
+    """
+    if data_args.streaming:
+        dataset = dataset.shuffle(buffer_size=data_args.buffer_size, seed=seed)
+        val_set = dataset.take(int(data_args.val_size))
+        train_set = dataset.skip(int(data_args.val_size))
+        return DatasetDict({"train": train_set, "validation": val_set})
+    else:
+        val_size = int(data_args.val_size) if data_args.val_size > 1 else data_args.val_size
+        dataset = dataset.train_test_split(test_size=val_size, seed=seed)
+        return DatasetDict({"train": dataset["train"], "validation": dataset["test"]})

llama-factory/src/llamafactory/data/formatter.py

@@ -0,0 +1,140 @@
+# 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 re
+from abc import ABC, abstractmethod
+from dataclasses import dataclass, field
+from typing import List, Literal, Optional, Tuple, Union
+
+from .data_utils import SLOTS
+from .tool_utils import DefaultToolUtils, GLM4ToolUtils
+
+
+@dataclass
+class Formatter(ABC):
+    slots: SLOTS = field(default_factory=list)
+    tool_format: Optional[Literal["default", "glm4"]] = None
+
+    @abstractmethod
+    def apply(self, **kwargs) -> SLOTS: ...
+
+    def extract(self, content: str) -> Union[str, List[Tuple[str, str]]]:
+        raise NotImplementedError
+
+
+@dataclass
+class EmptyFormatter(Formatter):
+    def __post_init__(self):
+        has_placeholder = False
+        for slot in filter(lambda s: isinstance(s, str), self.slots):
+            if re.search(r"\{\{[a-zA-Z_][a-zA-Z0-9_]*\}\}", slot):
+                has_placeholder = True
+
+        if has_placeholder:
+            raise ValueError("Empty formatter should not contain any placeholder.")
+
+    def apply(self, **kwargs) -> SLOTS:
+        return self.slots
+
+
+@dataclass
+class StringFormatter(Formatter):
+    def __post_init__(self):
+        has_placeholder = False
+        for slot in filter(lambda s: isinstance(s, str), self.slots):
+            if re.search(r"\{\{[a-zA-Z_][a-zA-Z0-9_]*\}\}", slot):
+                has_placeholder = True
+
+        if not has_placeholder:
+            raise ValueError("A placeholder is required in the string formatter.")
+
+    def apply(self, **kwargs) -> SLOTS:
+        elements = []
+        for slot in self.slots:
+            if isinstance(slot, str):
+                for name, value in kwargs.items():
+                    if not isinstance(value, str):
+                        raise RuntimeError("Expected a string, got {}".format(value))
+
+                    slot = slot.replace("{{" + name + "}}", value, 1)
+                elements.append(slot)
+            elif isinstance(slot, (dict, set)):
+                elements.append(slot)
+            else:
+                raise RuntimeError("Input must be string, set[str] or dict[str, str], got {}".format(type(slot)))
+
+        return elements
+
+
+@dataclass
+class FunctionFormatter(Formatter):
+    def __post_init__(self):
+        if self.tool_format == "default":
+            self.slots = DefaultToolUtils.get_function_slots() + self.slots
+        elif self.tool_format == "glm4":
+            self.slots = GLM4ToolUtils.get_function_slots() + self.slots
+        else:
+            raise NotImplementedError("Tool format {} was not found.".format(self.tool_format))
+
+    def apply(self, **kwargs) -> SLOTS:
+        content = kwargs.pop("content")
+        functions: List[Tuple[str, str]] = []
+        try:
+            tool_calls = json.loads(content)
+            if not isinstance(tool_calls, list):  # parallel function call
+                tool_calls = [tool_calls]
+
+            for tool_call in tool_calls:
+                functions.append((tool_call["name"], json.dumps(tool_call["arguments"], ensure_ascii=False)))
+
+        except json.JSONDecodeError:
+            functions = []
+
+        elements = []
+        for name, arguments in functions:
+            for slot in self.slots:
+                if isinstance(slot, str):
+                    slot = slot.replace("{{name}}", name).replace("{{arguments}}", arguments)
+                    elements.append(slot)
+                elif isinstance(slot, (dict, set)):
+                    elements.append(slot)
+                else:
+                    raise RuntimeError("Input must be string, set[str] or dict[str, str], got {}".format(type(slot)))
+
+        return elements
+
+
+@dataclass
+class ToolFormatter(Formatter):
+    def __post_init__(self):
+        if self.tool_format == "default":
+            self._tool_formatter = DefaultToolUtils.tool_formatter
+            self._tool_extractor = DefaultToolUtils.tool_extractor
+        elif self.tool_format == "glm4":
+            self._tool_formatter = GLM4ToolUtils.tool_formatter
+            self._tool_extractor = GLM4ToolUtils.tool_extractor
+        else:
+            raise NotImplementedError("Tool format {} was not found.".format(self.tool_format))
+
+    def apply(self, **kwargs) -> SLOTS:
+        content = kwargs.pop("content")
+        try:
+            tools = json.loads(content)
+            return [self._tool_formatter(tools) if len(tools) != 0 else ""]
+        except json.JSONDecodeError:
+            return [""]
+
+    def extract(self, content: str) -> Union[str, List[Tuple[str, str]]]:
+        return self._tool_extractor(content)

llama-factory/src/llamafactory/data/loader.py

@@ -0,0 +1,276 @@
+# 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 sys
+from typing import TYPE_CHECKING, Dict, Literal, Optional, Sequence, Union
+
+import numpy as np
+from datasets import DatasetDict, load_dataset, load_from_disk
+from transformers.utils.versions import require_version
+
+from ..extras.constants import FILEEXT2TYPE
+from ..extras.logging import get_logger
+from ..extras.misc import has_tokenized_data
+from .aligner import align_dataset
+from .data_utils import merge_dataset, split_dataset
+from .parser import get_dataset_list
+from .preprocess import get_preprocess_and_print_func
+from .template import get_template_and_fix_tokenizer
+
+
+if TYPE_CHECKING:
+    from datasets import Dataset, IterableDataset
+    from transformers import PreTrainedTokenizer, ProcessorMixin, Seq2SeqTrainingArguments
+
+    from ..hparams import DataArguments, ModelArguments
+    from .data_utils import DatasetModule
+    from .parser import DatasetAttr
+    from .template import Template
+
+
+logger = get_logger(__name__)
+
+
+def _load_single_dataset(
+    dataset_attr: "DatasetAttr",
+    model_args: "ModelArguments",
+    data_args: "DataArguments",
+    training_args: "Seq2SeqTrainingArguments",
+) -> Union["Dataset", "IterableDataset"]:
+    logger.info("Loading dataset {}...".format(dataset_attr))
+    data_path, data_name, data_dir, data_files = None, None, None, None
+    if dataset_attr.load_from in ["hf_hub", "ms_hub"]:
+        data_path = dataset_attr.dataset_name
+        data_name = dataset_attr.subset
+        data_dir = dataset_attr.folder
+
+    elif dataset_attr.load_from == "script":
+        data_path = os.path.join(data_args.dataset_dir, dataset_attr.dataset_name)
+        data_name = dataset_attr.subset
+        data_dir = dataset_attr.folder
+
+    elif dataset_attr.load_from == "file":
+        data_files = []
+        local_path = os.path.join(data_args.dataset_dir, dataset_attr.dataset_name)
+        if os.path.isdir(local_path):  # is directory
+            for file_name in os.listdir(local_path):
+                data_files.append(os.path.join(local_path, file_name))
+                if data_path is None:
+                    data_path = FILEEXT2TYPE.get(file_name.split(".")[-1], None)
+                elif data_path != FILEEXT2TYPE.get(file_name.split(".")[-1], None):
+                    raise ValueError("File types should be identical.")
+        elif os.path.isfile(local_path):  # is file
+            data_files.append(local_path)
+            data_path = FILEEXT2TYPE.get(local_path.split(".")[-1], None)
+        else:
+            raise ValueError("File {} not found.".format(local_path))
+
+        if data_path is None:
+            raise ValueError("Allowed file types: {}.".format(",".join(FILEEXT2TYPE.keys())))
+    else:
+        raise NotImplementedError("Unknown load type: {}.".format(dataset_attr.load_from))
+
+    if dataset_attr.load_from == "ms_hub":
+        require_version("modelscope>=1.11.0", "To fix: pip install modelscope>=1.11.0")
+        from modelscope import MsDataset
+        from modelscope.utils.config_ds import MS_DATASETS_CACHE
+
+        cache_dir = model_args.cache_dir or MS_DATASETS_CACHE
+        dataset = MsDataset.load(
+            dataset_name=data_path,
+            subset_name=data_name,
+            data_dir=data_dir,
+            data_files=data_files,
+            split=dataset_attr.split,
+            cache_dir=cache_dir,
+            token=model_args.ms_hub_token,
+            use_streaming=(data_args.streaming and (dataset_attr.load_from != "file")),
+        )
+        if isinstance(dataset, MsDataset):
+            dataset = dataset.to_hf_dataset()
+    else:
+        dataset = load_dataset(
+            path=data_path,
+            name=data_name,
+            data_dir=data_dir,
+            data_files=data_files,
+            split=dataset_attr.split,
+            cache_dir=model_args.cache_dir,
+            token=model_args.hf_hub_token,
+            streaming=(data_args.streaming and (dataset_attr.load_from != "file")),
+            trust_remote_code=True,
+        )
+
+    if data_args.streaming and (dataset_attr.load_from == "file"):  # faster than specifying streaming=True
+        dataset = dataset.to_iterable_dataset()  # TODO: add num shards parameter
+
+    if dataset_attr.num_samples is not None and not data_args.streaming:
+        target_num = dataset_attr.num_samples
+        indexes = np.random.permutation(len(dataset))[:target_num]
+        target_num -= len(indexes)
+        if target_num > 0:
+            expand_indexes = np.random.choice(len(dataset), target_num)
+            indexes = np.concatenate((indexes, expand_indexes), axis=0)
+
+        assert len(indexes) == dataset_attr.num_samples, "Sample num mismatched."
+        dataset = dataset.select(indexes)
+        logger.info("Sampled {} examples from dataset {}.".format(dataset_attr.num_samples, dataset_attr))
+
+    if data_args.max_samples is not None:  # truncate dataset
+        max_samples = min(data_args.max_samples, len(dataset))
+        dataset = dataset.select(range(max_samples))
+
+    return align_dataset(dataset, dataset_attr, data_args, training_args)
+
+
+def _get_merged_dataset(
+    dataset_names: Optional[Sequence[str]],
+    model_args: "ModelArguments",
+    data_args: "DataArguments",
+    training_args: "Seq2SeqTrainingArguments",
+    stage: Literal["pt", "sft", "rm", "ppo", "kto"],
+) -> Optional[Union["Dataset", "IterableDataset"]]:
+    if dataset_names is None:
+        return None
+
+    datasets = []
+    for dataset_attr in get_dataset_list(dataset_names, data_args.dataset_dir):
+        if (stage == "rm" and dataset_attr.ranking is False) or (stage != "rm" and dataset_attr.ranking is True):
+            raise ValueError("The dataset is not applicable in the current training stage.")
+
+        datasets.append(_load_single_dataset(dataset_attr, model_args, data_args, training_args))
+
+    return merge_dataset(datasets, data_args, seed=training_args.seed)
+
+
+def _get_preprocessed_dataset(
+    dataset: Optional[Union["Dataset", "IterableDataset"]],
+    data_args: "DataArguments",
+    training_args: "Seq2SeqTrainingArguments",
+    stage: Literal["pt", "sft", "rm", "ppo", "kto"],
+    template: "Template",
+    tokenizer: "PreTrainedTokenizer",
+    processor: Optional["ProcessorMixin"] = None,
+    is_eval: bool = False,
+) -> Optional[Union["Dataset", "IterableDataset"]]:
+    if dataset is None:
+        return None
+
+    preprocess_func, print_function = get_preprocess_and_print_func(
+        data_args, stage, template, tokenizer, processor, do_generate=(training_args.predict_with_generate and is_eval)
+    )
+    column_names = list(next(iter(dataset)).keys())
+    kwargs = {}
+    if not data_args.streaming:
+        kwargs = dict(
+            num_proc=data_args.preprocessing_num_workers,
+            load_from_cache_file=(not data_args.overwrite_cache) or (training_args.local_process_index != 0),
+            desc="Running tokenizer on dataset",
+        )
+
+    dataset = dataset.map(preprocess_func, batched=True, remove_columns=column_names, **kwargs)
+
+    if training_args.should_log:
+        try:
+            print("eval example:" if is_eval else "training example:")
+            print_function(next(iter(dataset)))
+        except StopIteration:
+            if stage == "pt":
+                raise RuntimeError("Cannot find sufficient samples, consider increasing dataset size.")
+            else:
+                raise RuntimeError("Cannot find valid samples, check `data/README.md` for the data format.")
+
+    return dataset
+
+
+def get_dataset(
+    model_args: "ModelArguments",
+    data_args: "DataArguments",
+    training_args: "Seq2SeqTrainingArguments",
+    stage: Literal["pt", "sft", "rm", "ppo", "kto"],
+    tokenizer: "PreTrainedTokenizer",
+    processor: Optional["ProcessorMixin"] = None,
+) -> "DatasetModule":
+    template = get_template_and_fix_tokenizer(tokenizer, data_args.template, data_args.tool_format)
+    if data_args.train_on_prompt and template.efficient_eos:
+        raise ValueError("Current template does not support `train_on_prompt`.")
+
+    # Load tokenized dataset
+    if data_args.tokenized_path is not None:
+        if has_tokenized_data(data_args.tokenized_path):
+            logger.warning("Loading dataset from disk will ignore other data arguments.")
+            dataset_dict: "DatasetDict" = load_from_disk(data_args.tokenized_path)
+            logger.info("Loaded tokenized dataset from {}.".format(data_args.tokenized_path))
+
+            dataset_module: Dict[str, "Dataset"] = {}
+            if "train" in dataset_dict:
+                dataset_module["train_dataset"] = dataset_dict["train"]
+            if "validation" in dataset_dict:
+                dataset_module["eval_dataset"] = dataset_dict["validation"]
+
+            if data_args.streaming:
+                dataset_module = {k: v.to_iterable_dataset() for k, v in dataset_module.items()}
+
+            return dataset_module
+
+        if data_args.streaming:
+            raise ValueError("Turn off `streaming` when saving dataset to disk.")
+
+    # Load and preprocess dataset
+    with training_args.main_process_first(desc="load dataset"):
+        dataset = _get_merged_dataset(data_args.dataset, model_args, data_args, training_args, stage)
+        eval_dataset = _get_merged_dataset(data_args.eval_dataset, model_args, data_args, training_args, stage)
+
+    with training_args.main_process_first(desc="pre-process dataset"):
+        dataset = _get_preprocessed_dataset(
+            dataset, data_args, training_args, stage, template, tokenizer, processor, is_eval=False
+        )
+        eval_dataset = _get_preprocessed_dataset(
+            eval_dataset, data_args, training_args, stage, template, tokenizer, processor, is_eval=True
+        )
+
+        if data_args.val_size > 1e-6:
+            dataset_dict = split_dataset(dataset, data_args, seed=training_args.seed)
+        else:
+            dataset_dict = {}
+            if dataset is not None:
+                if data_args.streaming:
+                    dataset = dataset.shuffle(buffer_size=data_args.buffer_size, seed=training_args.seed)
+
+                dataset_dict["train"] = dataset
+
+            if eval_dataset is not None:
+                if data_args.streaming:
+                    eval_dataset = eval_dataset.shuffle(buffer_size=data_args.buffer_size, seed=training_args.seed)
+
+                dataset_dict["validation"] = eval_dataset
+
+            dataset_dict = DatasetDict(dataset_dict)
+
+        if data_args.tokenized_path is not None:
+            if training_args.should_save:
+                dataset_dict.save_to_disk(data_args.tokenized_path)
+                logger.info("Tokenized dataset saved at {}.".format(data_args.tokenized_path))
+                logger.info("Please restart the training with `tokenized_path: {}`.".format(data_args.tokenized_path))
+
+            sys.exit(0)
+
+        dataset_module = {}
+        if "train" in dataset_dict:
+            dataset_module["train_dataset"] = dataset_dict["train"]
+        if "validation" in dataset_dict:
+            dataset_module["eval_dataset"] = dataset_dict["validation"]
+
+        return dataset_module

llama-factory/src/llamafactory/data/parser.py

@@ -0,0 +1,153 @@
+# 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 dataclasses import dataclass
+from typing import Any, Dict, List, Literal, Optional, Sequence
+
+from transformers.utils import cached_file
+
+from ..extras.constants import DATA_CONFIG
+from ..extras.misc import use_modelscope
+
+
+@dataclass
+class DatasetAttr:
+    r"""
+    Dataset attributes.
+    """
+
+    # basic configs
+    load_from: Literal["hf_hub", "ms_hub", "script", "file"]
+    dataset_name: str
+    formatting: Literal["alpaca", "sharegpt"] = "alpaca"
+    ranking: bool = False
+    # extra configs
+    subset: Optional[str] = None
+    split: str = "train"
+    folder: Optional[str] = None
+    num_samples: Optional[int] = None
+    # common columns
+    system: Optional[str] = None
+    tools: Optional[str] = None
+    images: Optional[str] = None
+    # rlhf columns
+    chosen: Optional[str] = None
+    rejected: Optional[str] = None
+    kto_tag: Optional[str] = None
+    # alpaca columns
+    prompt: Optional[str] = "instruction"
+    query: Optional[str] = "input"
+    response: Optional[str] = "output"
+    history: Optional[str] = None
+    # sharegpt columns
+    messages: Optional[str] = "conversations"
+    # sharegpt tags
+    role_tag: Optional[str] = "from"
+    content_tag: Optional[str] = "value"
+    user_tag: Optional[str] = "human"
+    assistant_tag: Optional[str] = "gpt"
+    observation_tag: Optional[str] = "observation"
+    function_tag: Optional[str] = "function_call"
+    system_tag: Optional[str] = "system"
+
+    def __repr__(self) -> str:
+        return self.dataset_name
+
+    def set_attr(self, key: str, obj: Dict[str, Any], default: Optional[Any] = None) -> None:
+        setattr(self, key, obj.get(key, default))
+
+
+def get_dataset_list(dataset_names: Optional[Sequence[str]], dataset_dir: str) -> List["DatasetAttr"]:
+    r"""
+    Gets the attributes of the datasets.
+    """
+    if dataset_names is None:
+        dataset_names = []
+
+    if dataset_dir == "ONLINE":
+        dataset_info = None
+    else:
+        if dataset_dir.startswith("REMOTE:"):
+            config_path = cached_file(path_or_repo_id=dataset_dir[7:], filename=DATA_CONFIG, repo_type="dataset")
+        else:
+            config_path = os.path.join(dataset_dir, DATA_CONFIG)
+
+        try:
+            with open(config_path, "r") as f:
+                dataset_info = json.load(f)
+        except Exception as err:
+            if len(dataset_names) != 0:
+                raise ValueError("Cannot open {} due to {}.".format(config_path, str(err)))
+
+            dataset_info = None
+
+    dataset_list: List["DatasetAttr"] = []
+    for name in dataset_names:
+        if dataset_info is None:  # dataset_dir is ONLINE
+            load_from = "ms_hub" if use_modelscope() else "hf_hub"
+            dataset_attr = DatasetAttr(load_from, dataset_name=name)
+            dataset_list.append(dataset_attr)
+            continue
+
+        if name not in dataset_info:
+            raise ValueError("Undefined dataset {} in {}.".format(name, DATA_CONFIG))
+
+        has_hf_url = "hf_hub_url" in dataset_info[name]
+        has_ms_url = "ms_hub_url" in dataset_info[name]
+
+        if has_hf_url or has_ms_url:
+            if (use_modelscope() and has_ms_url) or (not has_hf_url):
+                dataset_attr = DatasetAttr("ms_hub", dataset_name=dataset_info[name]["ms_hub_url"])
+            else:
+                dataset_attr = DatasetAttr("hf_hub", dataset_name=dataset_info[name]["hf_hub_url"])
+        elif "script_url" in dataset_info[name]:
+            dataset_attr = DatasetAttr("script", dataset_name=dataset_info[name]["script_url"])
+        else:
+            dataset_attr = DatasetAttr("file", dataset_name=dataset_info[name]["file_name"])
+
+        dataset_attr.set_attr("formatting", dataset_info[name], default="alpaca")
+        dataset_attr.set_attr("ranking", dataset_info[name], default=False)
+        dataset_attr.set_attr("subset", dataset_info[name])
+        dataset_attr.set_attr("split", dataset_info[name], default="train")
+        dataset_attr.set_attr("folder", dataset_info[name])
+        dataset_attr.set_attr("num_samples", dataset_info[name])
+
+        if "columns" in dataset_info[name]:
+            column_names = ["system", "tools", "images", "chosen", "rejected", "kto_tag"]
+            if dataset_attr.formatting == "alpaca":
+                column_names.extend(["prompt", "query", "response", "history"])
+            else:
+                column_names.extend(["messages"])
+
+            for column_name in column_names:
+                dataset_attr.set_attr(column_name, dataset_info[name]["columns"])
+
+        if dataset_attr.formatting == "sharegpt" and "tags" in dataset_info[name]:
+            tag_names = (
+                "role_tag",
+                "content_tag",
+                "user_tag",
+                "assistant_tag",
+                "observation_tag",
+                "function_tag",
+                "system_tag",
+            )
+            for tag in tag_names:
+                dataset_attr.set_attr(tag, dataset_info[name]["tags"])
+
+        dataset_list.append(dataset_attr)
+
+    return dataset_list

llama-factory/src/llamafactory/data/preprocess.py

@@ -0,0 +1,110 @@
+# 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 functools import partial
+from typing import TYPE_CHECKING, Callable, Literal, Optional, Tuple
+
+from .processors.feedback import preprocess_feedback_dataset
+from .processors.pairwise import preprocess_pairwise_dataset, print_pairwise_dataset_example
+from .processors.pretrain import preprocess_pretrain_dataset
+from .processors.supervised import (
+    preprocess_packed_supervised_dataset,
+    preprocess_supervised_dataset,
+    print_supervised_dataset_example,
+)
+from .processors.unsupervised import preprocess_unsupervised_dataset, print_unsupervised_dataset_example
+
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedTokenizer, ProcessorMixin
+
+    from ..hparams import DataArguments
+    from .template import Template
+
+
+def get_preprocess_and_print_func(
+    data_args: "DataArguments",
+    stage: Literal["pt", "sft", "rm", "ppo", "kto"],
+    template: "Template",
+    tokenizer: "PreTrainedTokenizer",
+    processor: Optional["ProcessorMixin"],
+    do_generate: bool = False,
+) -> Tuple[Callable, Callable]:
+    if stage == "pt":
+        preprocess_func = partial(
+            preprocess_pretrain_dataset,
+            tokenizer=tokenizer,
+            data_args=data_args,
+        )
+        print_function = partial(print_unsupervised_dataset_example, tokenizer=tokenizer)
+    elif stage == "sft" and not do_generate:
+        if data_args.packing:
+            if data_args.neat_packing:
+                from datasets.arrow_writer import OptimizedTypedSequence, TypedSequence
+
+                def __init__(self, data, **kwargs):
+                    return TypedSequence.__init__(
+                        self,
+                        data,
+                        type=kwargs.pop("type", None),
+                        try_type=kwargs.pop("try_type", None),
+                        optimized_int_type=kwargs.pop("optimized_int_type", None),
+                    )
+
+                OptimizedTypedSequence.__init__ = __init__
+            preprocess_func = partial(
+                preprocess_packed_supervised_dataset,
+                template=template,
+                tokenizer=tokenizer,
+                data_args=data_args,
+            )
+        else:
+            preprocess_func = partial(
+                preprocess_supervised_dataset,
+                template=template,
+                tokenizer=tokenizer,
+                processor=processor,
+                data_args=data_args,
+            )
+
+        print_function = partial(print_supervised_dataset_example, tokenizer=tokenizer)
+    elif stage == "rm":
+        preprocess_func = partial(
+            preprocess_pairwise_dataset,
+            template=template,
+            tokenizer=tokenizer,
+            processor=processor,
+            data_args=data_args,
+        )
+        print_function = partial(print_pairwise_dataset_example, tokenizer=tokenizer)
+    elif stage == "kto":
+        preprocess_func = partial(
+            preprocess_feedback_dataset,
+            template=template,
+            tokenizer=tokenizer,
+            processor=processor,
+            data_args=data_args,
+        )
+        print_function = partial(print_supervised_dataset_example, tokenizer=tokenizer)
+    else:
+        preprocess_func = partial(
+            preprocess_unsupervised_dataset,
+            template=template,
+            tokenizer=tokenizer,
+            processor=processor,
+            data_args=data_args,
+        )
+        print_function = partial(print_unsupervised_dataset_example, tokenizer=tokenizer)
+
+    return preprocess_func, print_function

llama-factory/src/llamafactory/data/processors/feedback.py

@@ -0,0 +1,143 @@
+# 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 typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence, Tuple
+
+from ...extras.constants import IGNORE_INDEX
+from ...extras.logging import get_logger
+from .processor_utils import get_paligemma_token_type_ids, get_pixel_values, infer_seqlen
+
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedTokenizer, ProcessorMixin
+
+    from ...hparams import DataArguments
+    from ..template import Template
+
+
+logger = get_logger(__name__)
+
+
+def _encode_feedback_example(
+    prompt: Sequence[Dict[str, str]],
+    response: Sequence[Dict[str, str]],
+    kl_response: Sequence[Dict[str, str]],
+    system: Optional[str],
+    tools: Optional[str],
+    template: "Template",
+    tokenizer: "PreTrainedTokenizer",
+    processor: Optional["ProcessorMixin"],
+    data_args: "DataArguments",
+) -> Tuple[List[int], List[int], List[int], List[int], bool]:
+    if processor is not None and not hasattr(processor, "image_seq_length"):  # llava-like models
+        prompt[0]["content"] = template.image_token + prompt[0]["content"]
+
+    if response[0]["content"]:  # desired example
+        kto_tag = True
+        messages = prompt + [response[0]]
+    else:  # undesired example
+        kto_tag = False
+        messages = prompt + [response[1]]
+
+    if kl_response[0]["content"]:
+        kl_messages = prompt + [kl_response[0]]
+    else:
+        kl_messages = prompt + [kl_response[1]]
+
+    prompt_ids, response_ids = template.encode_oneturn(tokenizer, messages, system, tools)
+    kl_prompt_ids, kl_response_ids = template.encode_oneturn(tokenizer, kl_messages, system, tools)
+
+    if template.efficient_eos:
+        response_ids += [tokenizer.eos_token_id]
+        kl_response_ids += [tokenizer.eos_token_id]
+
+    if processor is not None and hasattr(processor, "image_seq_length"):  # paligemma models
+        image_token_id = tokenizer.convert_tokens_to_ids(template.image_token)
+        prompt_ids = [image_token_id] * getattr(processor, "image_seq_length") + prompt_ids
+        kl_prompt_ids = [image_token_id] * getattr(processor, "image_seq_length") + kl_prompt_ids
+
+    source_len, target_len = infer_seqlen(len(prompt_ids), len(response_ids), data_args.cutoff_len)
+    prompt_ids = prompt_ids[:source_len]
+    response_ids = response_ids[:target_len]
+    kl_source_len, kl_target_len = infer_seqlen(len(kl_prompt_ids), len(kl_response_ids), data_args.cutoff_len)
+    kl_prompt_ids = kl_prompt_ids[:kl_source_len]
+    kl_response_ids = kl_response_ids[:kl_target_len]
+
+    input_ids = prompt_ids + response_ids
+    labels = [IGNORE_INDEX] * source_len + response_ids
+    kl_input_ids = kl_prompt_ids + kl_response_ids
+    kl_labels = [IGNORE_INDEX] * kl_source_len + kl_response_ids
+
+    return input_ids, labels, kl_input_ids, kl_labels, kto_tag
+
+
+def preprocess_feedback_dataset(
+    examples: Dict[str, List[Any]],
+    template: "Template",
+    tokenizer: "PreTrainedTokenizer",
+    processor: Optional["ProcessorMixin"],
+    data_args: "DataArguments",
+) -> Dict[str, List[List[int]]]:
+    # create unrelated input-output pairs for estimating the KL term by flipping the matched pairs
+    kl_response = examples["response"][::-1]
+    model_inputs = {
+        "input_ids": [],
+        "attention_mask": [],
+        "labels": [],
+        "kl_input_ids": [],
+        "kl_attention_mask": [],
+        "kl_labels": [],
+        "kto_tags": [],
+    }
+    if processor is not None:
+        model_inputs["pixel_values"] = []
+        if hasattr(processor, "image_seq_length"):  # paligemma models
+            model_inputs["token_type_ids"] = []
+            model_inputs["kl_token_type_ids"] = []
+
+    for i in range(len(examples["prompt"])):
+        if len(examples["prompt"][i]) % 2 != 1 or len(examples["response"][i]) < 2:
+            logger.warning("Dropped invalid example: {}".format(examples["prompt"][i] + examples["response"][i]))
+            continue
+
+        input_ids, labels, kl_input_ids, kl_labels, kto_tag = _encode_feedback_example(
+            prompt=examples["prompt"][i],
+            response=examples["response"][i],
+            kl_response=kl_response[i],
+            system=examples["system"][i],
+            tools=examples["tools"][i],
+            template=template,
+            tokenizer=tokenizer,
+            processor=processor,
+            data_args=data_args,
+        )
+        model_inputs["input_ids"].append(input_ids)
+        model_inputs["attention_mask"].append([1] * len(input_ids))
+        model_inputs["labels"].append(labels)
+        model_inputs["kl_input_ids"].append(kl_input_ids)
+        model_inputs["kl_attention_mask"].append([1] * len(kl_input_ids))
+        model_inputs["kl_labels"].append(kl_labels)
+        model_inputs["kto_tags"].append(kto_tag)
+        if processor is not None:
+            model_inputs["pixel_values"].append(get_pixel_values(examples["images"][i], processor))
+            if hasattr(processor, "image_seq_length"):  # paligemma models
+                model_inputs["token_type_ids"].append(get_paligemma_token_type_ids(len(input_ids), processor))
+                model_inputs["kl_token_type_ids"].append(get_paligemma_token_type_ids(len(kl_input_ids), processor))
+
+    desirable_num = sum([1 for tag in model_inputs["kto_tags"] if tag])
+    undesirable_num = len(model_inputs["kto_tags"]) - desirable_num
+    if desirable_num == 0 or undesirable_num == 0:
+        logger.warning("Your dataset only has one preference type.")
+
+    return model_inputs

llama-factory/src/llamafactory/data/processors/pairwise.py

@@ -0,0 +1,139 @@
+# 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 typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence, Tuple
+
+from ...extras.constants import IGNORE_INDEX
+from ...extras.logging import get_logger
+from .processor_utils import get_paligemma_token_type_ids, get_pixel_values, infer_seqlen
+
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedTokenizer, ProcessorMixin
+
+    from ...hparams import DataArguments
+    from ..template import Template
+
+
+logger = get_logger(__name__)
+
+
+def _encode_pairwise_example(
+    prompt: Sequence[Dict[str, str]],
+    response: Sequence[Dict[str, str]],
+    system: Optional[str],
+    tools: Optional[str],
+    template: "Template",
+    tokenizer: "PreTrainedTokenizer",
+    processor: Optional["ProcessorMixin"],
+    data_args: "DataArguments",
+) -> Tuple[List[int], List[int], List[int], List[int]]:
+    if processor is not None and not hasattr(processor, "image_seq_length"):  # llava-like models
+        prompt[0]["content"] = template.image_token + prompt[0]["content"]
+
+    chosen_messages = prompt + [response[0]]
+    rejected_messages = prompt + [response[1]]
+    prompt_ids, chosen_ids = template.encode_oneturn(tokenizer, chosen_messages, system, tools)
+    _, rejected_ids = template.encode_oneturn(tokenizer, rejected_messages, system, tools)
+
+    if template.efficient_eos:
+        chosen_ids += [tokenizer.eos_token_id]
+        rejected_ids += [tokenizer.eos_token_id]
+
+    if processor is not None and hasattr(processor, "image_seq_length"):  # paligemma models
+        image_token_id = tokenizer.convert_tokens_to_ids(template.image_token)
+        prompt_ids = [image_token_id] * getattr(processor, "image_seq_length") + prompt_ids
+
+    source_len, target_len = infer_seqlen(
+        len(prompt_ids), max(len(chosen_ids), len(rejected_ids)), data_args.cutoff_len
+    )  # consider the response is more important
+    prompt_ids = prompt_ids[:source_len]
+    chosen_ids = chosen_ids[:target_len]
+    rejected_ids = rejected_ids[:target_len]
+
+    chosen_input_ids = prompt_ids + chosen_ids
+    chosen_labels = [IGNORE_INDEX] * source_len + chosen_ids
+    rejected_input_ids = prompt_ids + rejected_ids
+    rejected_labels = [IGNORE_INDEX] * source_len + rejected_ids
+
+    return chosen_input_ids, chosen_labels, rejected_input_ids, rejected_labels
+
+
+def preprocess_pairwise_dataset(
+    examples: Dict[str, List[Any]],
+    template: "Template",
+    tokenizer: "PreTrainedTokenizer",
+    processor: Optional["ProcessorMixin"],
+    data_args: "DataArguments",
+) -> Dict[str, List[List[int]]]:
+    # build input pairs with format `<bos> X`, `Y1 <eos>` and `Y2 <eos>`
+    model_inputs = {
+        "chosen_input_ids": [],
+        "chosen_attention_mask": [],
+        "chosen_labels": [],
+        "rejected_input_ids": [],
+        "rejected_attention_mask": [],
+        "rejected_labels": [],
+    }
+    if processor is not None:
+        model_inputs["pixel_values"] = []
+        if hasattr(processor, "image_seq_length"):  # paligemma models
+            model_inputs["chosen_token_type_ids"] = []
+            model_inputs["rejected_token_type_ids"] = []
+
+    for i in range(len(examples["prompt"])):
+        if len(examples["prompt"][i]) % 2 != 1 or len(examples["response"][i]) < 2:
+            logger.warning("Dropped invalid example: {}".format(examples["prompt"][i] + examples["response"][i]))
+            continue
+
+        chosen_input_ids, chosen_labels, rejected_input_ids, rejected_labels = _encode_pairwise_example(
+            prompt=examples["prompt"][i],
+            response=examples["response"][i],
+            system=examples["system"][i],
+            tools=examples["tools"][i],
+            template=template,
+            tokenizer=tokenizer,
+            processor=processor,
+            data_args=data_args,
+        )
+        model_inputs["chosen_input_ids"].append(chosen_input_ids)
+        model_inputs["chosen_attention_mask"].append([1] * len(chosen_input_ids))
+        model_inputs["chosen_labels"].append(chosen_labels)
+        model_inputs["rejected_input_ids"].append(rejected_input_ids)
+        model_inputs["rejected_attention_mask"].append([1] * len(rejected_input_ids))
+        model_inputs["rejected_labels"].append(rejected_labels)
+        if processor is not None:
+            model_inputs["pixel_values"].append(get_pixel_values(examples["images"][i], processor))
+            if hasattr(processor, "image_seq_length"):  # paligemma models
+                model_inputs["chosen_token_type_ids"].append(
+                    get_paligemma_token_type_ids(len(chosen_input_ids), processor)
+                )
+                model_inputs["rejected_token_type_ids"].append(
+                    get_paligemma_token_type_ids(len(rejected_input_ids), processor)
+                )
+
+    return model_inputs
+
+
+def print_pairwise_dataset_example(example: Dict[str, List[int]], tokenizer: "PreTrainedTokenizer") -> None:
+    valid_chosen_labels = list(filter(lambda x: x != IGNORE_INDEX, example["chosen_labels"]))
+    valid_rejected_labels = list(filter(lambda x: x != IGNORE_INDEX, example["rejected_labels"]))
+    print("chosen_input_ids:\n{}".format(example["chosen_input_ids"]))
+    print("chosen_inputs:\n{}".format(tokenizer.decode(example["chosen_input_ids"], skip_special_tokens=False)))
+    print("chosen_label_ids:\n{}".format(example["chosen_labels"]))
+    print("chosen_labels:\n{}".format(tokenizer.decode(valid_chosen_labels, skip_special_tokens=False)))
+    print("rejected_input_ids:\n{}".format(example["rejected_input_ids"]))
+    print("rejected_inputs:\n{}".format(tokenizer.decode(example["rejected_input_ids"], skip_special_tokens=False)))
+    print("rejected_label_ids:\n{}".format(example["rejected_labels"]))
+    print("rejected_labels:\n{}".format(tokenizer.decode(valid_rejected_labels, skip_special_tokens=False)))

llama-factory/src/llamafactory/data/processors/pretrain.py

@@ -0,0 +1,54 @@
+# Copyright 2024 HuggingFace Inc. and the LlamaFactory team.
+#
+# This code is inspired by the HuggingFace's transformers library.
+# https://github.com/huggingface/transformers/blob/v4.40.0/examples/pytorch/language-modeling/run_clm.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.
+
+from itertools import chain
+from typing import TYPE_CHECKING, Any, Dict, List
+
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedTokenizer
+
+    from ...hparams import DataArguments
+
+
+def preprocess_pretrain_dataset(
+    examples: Dict[str, List[Any]], tokenizer: "PreTrainedTokenizer", data_args: "DataArguments"
+) -> Dict[str, List[List[int]]]:
+    # build grouped texts with format `X1 X2 X3 ...` if packing is enabled
+    eos_token = "<|end_of_text|>" if data_args.template == "llama3" else tokenizer.eos_token
+    text_examples = [messages[0]["content"] + eos_token for messages in examples["prompt"]]
+
+    if not data_args.packing:
+        if data_args.template == "gemma":
+            text_examples = [tokenizer.bos_token + example for example in text_examples]
+
+        result = tokenizer(text_examples, add_special_tokens=False, max_length=data_args.cutoff_len, truncation=True)
+    else:
+        tokenized_examples = tokenizer(text_examples, add_special_tokens=False)
+        concatenated_examples = {k: list(chain(*tokenized_examples[k])) for k in tokenized_examples.keys()}
+        total_length = len(concatenated_examples[list(concatenated_examples.keys())[0]])
+        block_size = data_args.cutoff_len
+        total_length = (total_length // block_size) * block_size
+        result = {
+            k: [t[i : i + block_size] for i in range(0, total_length, block_size)]
+            for k, t in concatenated_examples.items()
+        }
+        if data_args.template == "gemma":
+            for i in range(len(result["input_ids"])):
+                result["input_ids"][i][0] = tokenizer.bos_token_id
+
+    return result

llama-factory/src/llamafactory/data/processors/processor_utils.py

@@ -0,0 +1,95 @@
+# 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 bisect
+from typing import TYPE_CHECKING, List, Sequence, Tuple
+
+from ...extras.packages import is_pillow_available
+
+
+if is_pillow_available():
+    from PIL import Image
+
+
+if TYPE_CHECKING:
+    from numpy.typing import NDArray
+    from PIL.Image import Image as ImageObject
+    from transformers import ProcessorMixin
+    from transformers.image_processing_utils import BaseImageProcessor
+
+
+def search_for_fit(numbers: Sequence[int], capacity: int) -> int:
+    r"""
+    Finds the index of largest number that fits into the knapsack with the given capacity.
+    """
+    index = bisect.bisect(numbers, capacity)
+    return -1 if index == 0 else (index - 1)
+
+
+def greedy_knapsack(numbers: List[int], capacity: int) -> List[List[int]]:
+    r"""
+    An efficient greedy algorithm with binary search for the knapsack problem.
+    """
+    numbers.sort()  # sort numbers in ascending order for binary search
+    knapsacks = []
+
+    while numbers:
+        current_knapsack = []
+        remaining_capacity = capacity
+
+        while True:
+            index = search_for_fit(numbers, remaining_capacity)
+            if index == -1:
+                break  # no more numbers fit in this knapsack
+
+            remaining_capacity -= numbers[index]  # update the remaining capacity
+            current_knapsack.append(numbers.pop(index))  # add the number to knapsack
+
+        knapsacks.append(current_knapsack)
+
+    return knapsacks
+
+
+def get_pixel_values(images: Sequence["ImageObject"], processor: "ProcessorMixin") -> "NDArray":
+    r"""
+    Processes visual inputs. (currently only supports a single image)
+    """
+    image_processor: "BaseImageProcessor" = getattr(processor, "image_processor")
+    image = images[0] if len(images) != 0 else Image.new("RGB", (100, 100), (255, 255, 255))
+    return image_processor(image, return_tensors="pt")["pixel_values"][0]  # shape (C, H, W)
+
+
+def get_paligemma_token_type_ids(input_len: int, processor: "ProcessorMixin") -> List[int]:
+    r"""
+    Gets paligemma token type ids for computing loss.
+    """
+    image_seq_length = getattr(processor, "image_seq_length")
+    return [0] * image_seq_length + [1] * (input_len - image_seq_length)
+
+
+def infer_seqlen(source_len: int, target_len: int, cutoff_len: int) -> Tuple[int, int]:
+    r"""
+    Computes the real sequence length after truncation by the cutoff_len.
+    """
+    if target_len * 2 < cutoff_len:  # truncate source
+        max_target_len = cutoff_len
+    elif source_len * 2 < cutoff_len:  # truncate target
+        max_target_len = cutoff_len - source_len
+    else:  # truncate both
+        max_target_len = int(cutoff_len * (target_len / (source_len + target_len)))
+
+    new_target_len = min(max_target_len, target_len)
+    max_source_len = max(cutoff_len - new_target_len, 0)
+    new_source_len = min(max_source_len, source_len)
+    return new_source_len, new_target_len

llama-factory/src/llamafactory/data/processors/supervised.py

@@ -0,0 +1,202 @@
+# 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 collections import defaultdict
+from typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence, Tuple
+
+from ...extras.constants import IGNORE_INDEX
+from ...extras.logging import get_logger
+from .processor_utils import get_paligemma_token_type_ids, get_pixel_values, greedy_knapsack, infer_seqlen
+
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedTokenizer, ProcessorMixin
+
+    from ...hparams import DataArguments
+    from ..template import Template
+
+
+logger = get_logger(__name__)
+
+
+def _encode_supervised_example(
+    prompt: Sequence[Dict[str, str]],
+    response: Sequence[Dict[str, str]],
+    system: Optional[str],
+    tools: Optional[str],
+    template: "Template",
+    tokenizer: "PreTrainedTokenizer",
+    processor: Optional["ProcessorMixin"],
+    data_args: "DataArguments",
+) -> Tuple[List[int], List[int]]:
+    if processor is not None and not hasattr(processor, "image_seq_length"):  # llava-like models
+        prompt[0]["content"] = template.image_token + prompt[0]["content"]
+
+    messages = prompt + response
+    input_ids, labels = [], []
+
+    if processor is not None and hasattr(processor, "image_seq_length"):  # paligemma models
+        image_token_id = tokenizer.convert_tokens_to_ids(template.image_token)
+        input_ids += [image_token_id] * getattr(processor, "image_seq_length")
+        labels += [IGNORE_INDEX] * getattr(processor, "image_seq_length")
+
+    encoded_pairs = template.encode_multiturn(tokenizer, messages, system, tools)
+    total_length = 1 if template.efficient_eos else 0
+    for turn_idx, (source_ids, target_ids) in enumerate(encoded_pairs):
+        if total_length >= data_args.cutoff_len:
+            break
+
+        source_len, target_len = infer_seqlen(len(source_ids), len(target_ids), data_args.cutoff_len - total_length)
+        source_ids = source_ids[:source_len]
+        target_ids = target_ids[:target_len]
+        total_length += source_len + target_len
+
+        if data_args.train_on_prompt:
+            source_label = source_ids
+        elif turn_idx != 0 and template.efficient_eos:
+            source_label = [tokenizer.eos_token_id] + [IGNORE_INDEX] * (source_len - 1)
+        else:
+            source_label = [IGNORE_INDEX] * source_len
+
+        if data_args.mask_history and turn_idx != len(encoded_pairs) - 1:
+            target_label = [IGNORE_INDEX] * target_len
+        else:
+            target_label = target_ids
+
+        input_ids += source_ids + target_ids
+        labels += source_label + target_label
+
+    if template.efficient_eos:
+        input_ids += [tokenizer.eos_token_id]
+        labels += [tokenizer.eos_token_id]
+
+    return input_ids, labels
+
+
+def preprocess_supervised_dataset(
+    examples: Dict[str, List[Any]],
+    template: "Template",
+    tokenizer: "PreTrainedTokenizer",
+    processor: Optional["ProcessorMixin"],
+    data_args: "DataArguments",
+) -> Dict[str, List[List[int]]]:
+    # build inputs with format `<bos> X Y <eos>` and labels with format `<ignore> ... <ignore> Y <eos>`
+    # for multiturn examples, we only mask the prompt part in each prompt-response pair.
+    model_inputs = {"input_ids": [], "attention_mask": [], "labels": []}
+    if processor is not None:
+        model_inputs["pixel_values"] = []
+        if hasattr(processor, "image_seq_length"):  # paligemma models
+            model_inputs["token_type_ids"] = []
+
+    for i in range(len(examples["prompt"])):
+        if len(examples["prompt"][i]) % 2 != 1 or len(examples["response"][i]) != 1:
+            logger.warning("Dropped invalid example: {}".format(examples["prompt"][i] + examples["response"][i]))
+            continue
+
+        input_ids, labels = _encode_supervised_example(
+            prompt=examples["prompt"][i],
+            response=examples["response"][i],
+            system=examples["system"][i],
+            tools=examples["tools"][i],
+            template=template,
+            tokenizer=tokenizer,
+            processor=processor,
+            data_args=data_args,
+        )
+        model_inputs["input_ids"].append(input_ids)
+        model_inputs["attention_mask"].append([1] * len(input_ids))
+        model_inputs["labels"].append(labels)
+        if processor is not None:
+            model_inputs["pixel_values"].append(get_pixel_values(examples["images"][i], processor))
+            if hasattr(processor, "image_seq_length"):  # paligemma models
+                model_inputs["token_type_ids"].append(get_paligemma_token_type_ids(len(input_ids), processor))
+
+    return model_inputs
+
+
+def preprocess_packed_supervised_dataset(
+    examples: Dict[str, List[Any]],
+    template: "Template",
+    tokenizer: "PreTrainedTokenizer",
+    data_args: "DataArguments",
+) -> Dict[str, List[List[int]]]:
+    # build inputs with format `<bos> X1 Y1 <eos> <bos> X2 Y2 <eos>`
+    # and labels with format `<ignore> ... <ignore> Y1 <eos> <ignore> ... <ignore> Y2 <eos>`
+    valid_num = 0
+    batch_input_ids, batch_labels = [], []
+    lengths = []
+    length2indexes = defaultdict(list)
+    for i in range(len(examples["prompt"])):
+        if len(examples["prompt"][i]) % 2 != 1 or len(examples["response"][i]) != 1:
+            logger.warning("Dropped invalid example: {}".format(examples["prompt"][i] + examples["response"][i]))
+            continue
+
+        input_ids, labels = _encode_supervised_example(
+            prompt=examples["prompt"][i],
+            response=examples["response"][i],
+            system=examples["system"][i],
+            tools=examples["tools"][i],
+            template=template,
+            tokenizer=tokenizer,
+            processor=None,
+            data_args=data_args,
+        )
+        length = len(input_ids)
+        if length > data_args.cutoff_len:
+            logger.warning("Dropped lengthy example with length {} > {}.".format(length, data_args.cutoff_len))
+        else:
+            lengths.append(length)
+            length2indexes[length].append(valid_num)
+            batch_input_ids.append(input_ids)
+            batch_labels.append(labels)
+            valid_num += 1
+
+    model_inputs = {"input_ids": [], "attention_mask": [], "labels": []}
+    knapsacks = greedy_knapsack(lengths, data_args.cutoff_len)
+    for knapsack in knapsacks:
+        packed_input_ids, packed_attention_masks, packed_labels = [], [], []
+        for i, length in enumerate(knapsack):
+            index = length2indexes[length].pop()
+            packed_input_ids += batch_input_ids[index]
+            packed_labels += batch_labels[index]
+            if data_args.neat_packing:
+                packed_attention_masks += [i + 1] * len(batch_input_ids[index])  # start from 1
+            else:
+                packed_attention_masks += [1] * len(batch_input_ids[index])
+
+        if len(packed_input_ids) < data_args.cutoff_len:
+            pad_length = data_args.cutoff_len - len(packed_input_ids)
+            packed_input_ids += [tokenizer.pad_token_id] * pad_length
+            packed_labels += [IGNORE_INDEX] * pad_length
+            if data_args.neat_packing:
+                packed_attention_masks += [0] * pad_length
+            else:
+                packed_attention_masks += [1] * pad_length  # more efficient flash_attn
+
+        if len(packed_input_ids) != data_args.cutoff_len:
+            raise ValueError("The length of packed example should be identical to the cutoff length.")
+
+        model_inputs["input_ids"].append(packed_input_ids)
+        model_inputs["attention_mask"].append(packed_attention_masks)
+        model_inputs["labels"].append(packed_labels)
+
+    return model_inputs
+
+
+def print_supervised_dataset_example(example: Dict[str, List[int]], tokenizer: "PreTrainedTokenizer") -> None:
+    valid_labels = list(filter(lambda x: x != IGNORE_INDEX, example["labels"]))
+    print("input_ids:\n{}".format(example["input_ids"]))
+    print("inputs:\n{}".format(tokenizer.decode(example["input_ids"], skip_special_tokens=False)))
+    print("label_ids:\n{}".format(example["labels"]))
+    print("labels:\n{}".format(tokenizer.decode(valid_labels, skip_special_tokens=False)))

llama-factory/src/llamafactory/data/processors/unsupervised.py

@@ -0,0 +1,106 @@
+# 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 typing import TYPE_CHECKING, Any, Dict, List, Optional, Sequence, Tuple
+
+from ...extras.logging import get_logger
+from ..data_utils import Role
+from .processor_utils import get_paligemma_token_type_ids, get_pixel_values, infer_seqlen
+
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedTokenizer, ProcessorMixin
+
+    from ...hparams import DataArguments
+    from ..template import Template
+
+
+logger = get_logger(__name__)
+
+
+def _encode_unsupervised_example(
+    prompt: Sequence[Dict[str, str]],
+    response: Sequence[Dict[str, str]],
+    system: Optional[str],
+    tools: Optional[str],
+    template: "Template",
+    tokenizer: "PreTrainedTokenizer",
+    processor: Optional["ProcessorMixin"],
+    data_args: "DataArguments",
+) -> Tuple[List[int], List[int]]:
+    if processor is not None and not hasattr(processor, "image_seq_length"):  # llava-like models
+        prompt[0]["content"] = template.image_token + prompt[0]["content"]
+
+    if len(response) == 1:
+        messages = prompt + response
+    else:
+        messages = prompt + [{"role": Role.ASSISTANT.value, "content": ""}]
+
+    input_ids, labels = template.encode_oneturn(tokenizer, messages, system, tools)
+    if template.efficient_eos:
+        labels += [tokenizer.eos_token_id]
+
+    if processor is not None and hasattr(processor, "image_seq_length"):  # paligemma models
+        image_token_id = tokenizer.convert_tokens_to_ids(template.image_token)
+        input_ids = [image_token_id] * getattr(processor, "image_seq_length") + input_ids
+
+    source_len, target_len = infer_seqlen(len(input_ids), len(labels), data_args.cutoff_len)
+    input_ids = input_ids[:source_len]
+    labels = labels[:target_len]
+    return input_ids, labels
+
+
+def preprocess_unsupervised_dataset(
+    examples: Dict[str, List[Any]],
+    template: "Template",
+    tokenizer: "PreTrainedTokenizer",
+    processor: Optional["ProcessorMixin"],
+    data_args: "DataArguments",
+) -> Dict[str, List[List[int]]]:
+    # build inputs with format `<bos> X` and labels with format `Y <eos>`
+    model_inputs = {"input_ids": [], "attention_mask": [], "labels": []}
+    if processor is not None:
+        model_inputs["pixel_values"] = []
+        if hasattr(processor, "image_seq_length"):  # paligemma models
+            model_inputs["token_type_ids"] = []
+
+    for i in range(len(examples["prompt"])):
+        if len(examples["prompt"][i]) % 2 != 1:
+            logger.warning("Dropped invalid example: {}".format(examples["prompt"][i] + examples["response"][i]))
+            continue
+
+        input_ids, labels = _encode_unsupervised_example(
+            prompt=examples["prompt"][i],
+            response=examples["response"][i],
+            system=examples["system"][i],
+            tools=examples["tools"][i],
+            template=template,
+            tokenizer=tokenizer,
+            processor=processor,
+            data_args=data_args,
+        )
+        model_inputs["input_ids"].append(input_ids)
+        model_inputs["attention_mask"].append([1] * len(input_ids))
+        model_inputs["labels"].append(labels)
+        if processor is not None:
+            model_inputs["pixel_values"].append(get_pixel_values(examples["images"][i], processor))
+            if hasattr(processor, "image_seq_length"):  # paligemma models
+                model_inputs["token_type_ids"].append(get_paligemma_token_type_ids(len(input_ids), processor))
+
+    return model_inputs
+
+
+def print_unsupervised_dataset_example(example: Dict[str, List[int]], tokenizer: "PreTrainedTokenizer") -> None:
+    print("input_ids:\n{}".format(example["input_ids"]))
+    print("inputs:\n{}".format(tokenizer.decode(example["input_ids"], skip_special_tokens=False)))

llama-factory/src/llamafactory/data/template.py

@@ -0,0 +1,905 @@
+# 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 dataclasses import dataclass
+from typing import TYPE_CHECKING, Dict, List, Optional, Sequence, Tuple, Union
+
+from ..extras.logging import get_logger
+from .data_utils import Role
+from .formatter import EmptyFormatter, FunctionFormatter, StringFormatter, ToolFormatter
+
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedTokenizer
+
+    from .formatter import SLOTS, Formatter
+
+
+logger = get_logger(__name__)
+
+
+@dataclass
+class Template:
+    format_user: "Formatter"
+    format_assistant: "Formatter"
+    format_system: "Formatter"
+    format_function: "Formatter"
+    format_observation: "Formatter"
+    format_tools: "Formatter"
+    format_separator: "Formatter"
+    format_prefix: "Formatter"
+    default_system: str
+    stop_words: List[str]
+    image_token: str
+    efficient_eos: bool
+    replace_eos: bool
+
+    def encode_oneturn(
+        self,
+        tokenizer: "PreTrainedTokenizer",
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+    ) -> Tuple[List[int], List[int]]:
+        r"""
+        Returns a single pair of token ids representing prompt and response respectively.
+        """
+        encoded_messages = self._encode(tokenizer, messages, system, tools)
+        prompt_ids = []
+        for encoded_ids in encoded_messages[:-1]:
+            prompt_ids += encoded_ids
+
+        answer_ids = encoded_messages[-1]
+        return prompt_ids, answer_ids
+
+    def encode_multiturn(
+        self,
+        tokenizer: "PreTrainedTokenizer",
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str] = None,
+        tools: Optional[str] = None,
+    ) -> List[Tuple[List[int], List[int]]]:
+        r"""
+        Returns multiple pairs of token ids representing prompts and responses respectively.
+        """
+        encoded_messages = self._encode(tokenizer, messages, system, tools)
+        return [(encoded_messages[i], encoded_messages[i + 1]) for i in range(0, len(encoded_messages), 2)]
+
+    def extract_tool(self, content: str) -> Union[str, List[Tuple[str, str]]]:
+        r"""
+        Extracts tool message.
+        """
+        return self.format_tools.extract(content)
+
+    def _encode(
+        self,
+        tokenizer: "PreTrainedTokenizer",
+        messages: Sequence[Dict[str, str]],
+        system: Optional[str],
+        tools: Optional[str],
+    ) -> List[List[int]]:
+        r"""
+        Encodes formatted inputs to pairs of token ids.
+        Turn 0: prefix + system + query        resp
+        Turn t: sep + query                    resp
+        """
+        system = system or self.default_system
+        encoded_messages = []
+        for i, message in enumerate(messages):
+            elements = []
+
+            if i == 0:
+                elements += self.format_prefix.apply()
+                if system or tools:
+                    tool_text = self.format_tools.apply(content=tools)[0] if tools else ""
+                    elements += self.format_system.apply(content=(system + tool_text))
+
+            if i > 0 and i % 2 == 0:
+                elements += self.format_separator.apply()
+
+            if message["role"] == Role.USER.value:
+                elements += self.format_user.apply(content=message["content"], idx=str(i // 2))
+            elif message["role"] == Role.ASSISTANT.value:
+                elements += self.format_assistant.apply(content=message["content"])
+            elif message["role"] == Role.OBSERVATION.value:
+                elements += self.format_observation.apply(content=message["content"])
+            elif message["role"] == Role.FUNCTION.value:
+                elements += self.format_function.apply(content=message["content"])
+            else:
+                raise NotImplementedError("Unexpected role: {}".format(message["role"]))
+
+            encoded_messages.append(self._convert_elements_to_ids(tokenizer, elements))
+
+        return encoded_messages
+
+    def _convert_elements_to_ids(self, tokenizer: "PreTrainedTokenizer", elements: "SLOTS") -> List[int]:
+        r"""
+        Converts elements to token ids.
+        """
+        token_ids = []
+        for elem in elements:
+            if isinstance(elem, str):
+                if len(elem) != 0:
+                    token_ids += tokenizer.encode(elem, add_special_tokens=False)
+            elif isinstance(elem, dict):
+                token_ids += [tokenizer.convert_tokens_to_ids(elem.get("token"))]
+            elif isinstance(elem, set):
+                if "bos_token" in elem and tokenizer.bos_token_id is not None:
+                    token_ids += [tokenizer.bos_token_id]
+                elif "eos_token" in elem and tokenizer.eos_token_id is not None:
+                    token_ids += [tokenizer.eos_token_id]
+            else:
+                raise ValueError("Input must be string, set[str] or dict[str, str], got {}".format(type(elem)))
+
+        return token_ids
+
+
+@dataclass
+class Llama2Template(Template):
+    def _encode(
+        self,
+        tokenizer: "PreTrainedTokenizer",
+        messages: Sequence[Dict[str, str]],
+        system: str,
+        tools: str,
+    ) -> List[List[int]]:
+        r"""
+        Encodes formatted inputs to pairs of token ids.
+        Turn 0: prefix + system + query        resp
+        Turn t: sep + query                    resp
+        """
+        system = system or self.default_system
+        encoded_messages = []
+        for i, message in enumerate(messages):
+            elements = []
+
+            system_text = ""
+            if i == 0:
+                elements += self.format_prefix.apply()
+                if system or tools:
+                    tool_text = self.format_tools.apply(content=tools)[0] if tools else ""
+                    system_text = self.format_system.apply(content=(system + tool_text))[0]
+
+            if i > 0 and i % 2 == 0:
+                elements += self.format_separator.apply()
+
+            if message["role"] == Role.USER.value:
+                elements += self.format_user.apply(content=system_text + message["content"])
+            elif message["role"] == Role.ASSISTANT.value:
+                elements += self.format_assistant.apply(content=message["content"])
+            elif message["role"] == Role.OBSERVATION.value:
+                elements += self.format_observation.apply(content=message["content"])
+            elif message["role"] == Role.FUNCTION.value:
+                elements += self.format_function.apply(content=message["content"])
+            else:
+                raise NotImplementedError("Unexpected role: {}".format(message["role"]))
+
+            encoded_messages.append(self._convert_elements_to_ids(tokenizer, elements))
+
+        return encoded_messages
+
+
+TEMPLATES: Dict[str, Template] = {}
+
+
+def _register_template(
+    name: str,
+    format_user: Optional["Formatter"] = None,
+    format_assistant: Optional["Formatter"] = None,
+    format_system: Optional["Formatter"] = None,
+    format_function: Optional["Formatter"] = None,
+    format_observation: Optional["Formatter"] = None,
+    format_tools: Optional["Formatter"] = None,
+    format_separator: Optional["Formatter"] = None,
+    format_prefix: Optional["Formatter"] = None,
+    default_system: str = "",
+    stop_words: Sequence[str] = [],
+    image_token: str = "<image>",
+    efficient_eos: bool = False,
+    replace_eos: bool = False,
+) -> None:
+    r"""
+    Registers a chat template.
+
+    To add the following chat template:
+    ```
+    [HUMAN]:
+    user prompt here
+    [AI]:
+    model response here
+
+    [HUMAN]:
+    user prompt here
+    [AI]:
+    model response here
+    ```
+
+    The corresponding code should be:
+    ```
+    _register_template(
+        name="custom",
+        format_user=StringFormatter(slots=["[HUMAN]:\n{{content}}\n[AI]:\n"]),
+        format_separator=EmptyFormatter(slots=["\n\n"]),
+        efficient_eos=True,
+    )
+    ```
+    """
+    eos_slots = [] if efficient_eos else [{"eos_token"}]
+    template_class = Llama2Template if name.startswith("llama2") else Template
+    default_user_formatter = StringFormatter(slots=["{{content}}"])
+    default_assistant_formatter = StringFormatter(slots=["{{content}}"] + eos_slots)
+    default_function_formatter = FunctionFormatter(slots=eos_slots, tool_format="default")
+    default_tool_formatter = ToolFormatter(tool_format="default")
+    default_separator_formatter = EmptyFormatter()
+    default_prefix_formatter = EmptyFormatter()
+    TEMPLATES[name] = template_class(
+        format_user=format_user or default_user_formatter,
+        format_assistant=format_assistant or default_assistant_formatter,
+        format_system=format_system or default_user_formatter,
+        format_function=format_function or default_function_formatter,
+        format_observation=format_observation or format_user or default_user_formatter,
+        format_tools=format_tools or default_tool_formatter,
+        format_separator=format_separator or default_separator_formatter,
+        format_prefix=format_prefix or default_prefix_formatter,
+        default_system=default_system,
+        stop_words=stop_words,
+        image_token=image_token,
+        efficient_eos=efficient_eos,
+        replace_eos=replace_eos,
+    )
+
+
+def _add_or_replace_eos_token(tokenizer: "PreTrainedTokenizer", eos_token: str) -> None:
+    is_added = tokenizer.eos_token_id is None
+    num_added_tokens = tokenizer.add_special_tokens({"eos_token": eos_token})
+
+    if is_added:
+        logger.info("Add eos token: {}".format(tokenizer.eos_token))
+    else:
+        logger.info("Replace eos token: {}".format(tokenizer.eos_token))
+
+    if num_added_tokens > 0:
+        logger.warning("New tokens have been added, make sure `resize_vocab` is True.")
+
+
+def _jinja_escape(content: str) -> str:
+    return content.replace("'", r"\'")
+
+
+def _convert_slots_to_jinja(slots: "SLOTS", tokenizer: "PreTrainedTokenizer", placeholder: str = "content") -> str:
+    slot_items = []
+    for slot in slots:
+        if isinstance(slot, str):
+            slot_pieces = slot.split("{{content}}")
+            if slot_pieces[0]:
+                slot_items.append("'" + _jinja_escape(slot_pieces[0]) + "'")
+            if len(slot_pieces) > 1:
+                slot_items.append(placeholder)
+                if slot_pieces[1]:
+                    slot_items.append("'" + _jinja_escape(slot_pieces[1]) + "'")
+        elif isinstance(slot, set):  # do not use {{ eos_token }} since it may be replaced
+            if "bos_token" in slot and tokenizer.bos_token_id is not None:
+                slot_items.append("'" + tokenizer.bos_token + "'")
+            elif "eos_token" in slot and tokenizer.eos_token_id is not None:
+                slot_items.append("'" + tokenizer.eos_token + "'")
+        elif isinstance(slot, dict):
+            raise ValueError("Dict is not supported.")
+
+    return " + ".join(slot_items)
+
+
+def _get_jinja_template(template: "Template", tokenizer: "PreTrainedTokenizer") -> str:
+    jinja_template = ""
+
+    prefix = _convert_slots_to_jinja(template.format_prefix.apply(), tokenizer)
+    if prefix:
+        jinja_template += "{{ " + prefix + " }}"
+
+    if template.default_system:
+        jinja_template += "{% set system_message = '" + _jinja_escape(template.default_system) + "' %}"
+
+    jinja_template += (
+        "{% if messages[0]['role'] == 'system' %}{% set system_message = messages[0]['content'] %}{% endif %}"
+    )
+
+    system_message = _convert_slots_to_jinja(template.format_system.apply(), tokenizer, placeholder="system_message")
+    if not isinstance(template, Llama2Template):
+        jinja_template += "{% if system_message is defined %}{{ " + system_message + " }}{% endif %}"
+
+    jinja_template += "{% for message in messages %}"
+    jinja_template += "{% set content = message['content'] %}"
+    if isinstance(template, Llama2Template):
+        jinja_template += "{% if loop.index0 == 0 and system_message is defined %}"
+        jinja_template += "{% set content = " + system_message + " + message['content'] %}"
+        jinja_template += "{% endif %}"
+
+    jinja_template += "{% if message['role'] == 'user' %}"
+    user_message = _convert_slots_to_jinja(template.format_user.apply(), tokenizer)
+    jinja_template += "{{ " + user_message + " }}"
+
+    jinja_template += "{% elif message['role'] == 'assistant' %}"
+    assistant_message = _convert_slots_to_jinja(
+        template.format_assistant.apply() + template.format_separator.apply(), tokenizer
+    )
+    jinja_template += "{{ " + assistant_message + " }}"
+    jinja_template += "{% endif %}"
+    jinja_template += "{% endfor %}"
+    return jinja_template
+
+
+def get_template_and_fix_tokenizer(
+    tokenizer: "PreTrainedTokenizer",
+    name: Optional[str] = None,
+    tool_format: Optional[str] = None,
+) -> Template:
+    if name is None:
+        template = TEMPLATES["empty"]  # placeholder
+    else:
+        template = TEMPLATES.get(name, None)
+        if template is None:
+            raise ValueError("Template {} does not exist.".format(name))
+
+    if tool_format is not None:
+        logger.info("Using tool format: {}.".format(tool_format))
+        eos_slots = [] if template.efficient_eos else [{"eos_token"}]
+        template.format_tools = ToolFormatter(tool_format=tool_format)
+        template.format_function = FunctionFormatter(slots=eos_slots, tool_format=tool_format)
+
+    stop_words = template.stop_words
+    if template.replace_eos:
+        if not stop_words:
+            raise ValueError("Stop words are required to replace the EOS token.")
+
+        _add_or_replace_eos_token(tokenizer, eos_token=stop_words[0])
+        stop_words = stop_words[1:]
+
+    if tokenizer.eos_token_id is None:
+        _add_or_replace_eos_token(tokenizer, eos_token="<|endoftext|>")
+
+    if tokenizer.pad_token_id is None:
+        tokenizer.pad_token = tokenizer.eos_token
+        logger.info("Add pad token: {}".format(tokenizer.pad_token))
+
+    if stop_words:
+        num_added_tokens = tokenizer.add_special_tokens(
+            dict(additional_special_tokens=stop_words), replace_additional_special_tokens=False
+        )
+        logger.info("Add {} to stop words.".format(",".join(stop_words)))
+        if num_added_tokens > 0:
+            logger.warning("New tokens have been added, make sure `resize_vocab` is True.")
+
+    try:
+        tokenizer.chat_template = _get_jinja_template(template, tokenizer)
+    except ValueError:
+        logger.info("Cannot add this chat template to tokenizer.")
+
+    return template
+
+
+_register_template(
+    name="alpaca",
+    format_user=StringFormatter(slots=["### Instruction:\n{{content}}\n\n### Response:\n"]),
+    format_separator=EmptyFormatter(slots=["\n\n"]),
+    default_system=(
+        "Below is an instruction that describes a task. "
+        "Write a response that appropriately completes the request.\n\n"
+    ),
+)
+
+
+_register_template(
+    name="aquila",
+    format_user=StringFormatter(slots=["Human: {{content}}###Assistant:"]),
+    format_separator=EmptyFormatter(slots=["###"]),
+    default_system=(
+        "A chat between a curious human and an artificial intelligence assistant. "
+        "The assistant gives helpful, detailed, and polite answers to the human's questions."
+    ),
+    stop_words=["</s>"],
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="atom",
+    format_user=StringFormatter(
+        slots=[{"bos_token"}, "Human: {{content}}\n", {"eos_token"}, {"bos_token"}, "Assistant:"]
+    ),
+    format_assistant=StringFormatter(slots=["{{content}}\n", {"eos_token"}]),
+)
+
+
+_register_template(
+    name="baichuan",
+    format_user=StringFormatter(slots=[{"token": "<reserved_102>"}, "{{content}}", {"token": "<reserved_103>"}]),
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="baichuan2",
+    format_user=StringFormatter(slots=["<reserved_106>{{content}}<reserved_107>"]),
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="belle",
+    format_user=StringFormatter(slots=["Human: {{content}}\n\nBelle: "]),
+    format_separator=EmptyFormatter(slots=["\n\n"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+)
+
+
+_register_template(
+    name="bluelm",
+    format_user=StringFormatter(slots=[{"token": "[|Human|]:"}, "{{content}}", {"token": "[|AI|]:"}]),
+)
+
+
+_register_template(
+    name="breeze",
+    format_user=StringFormatter(slots=["[INST] {{content}} [/INST] "]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="chatglm2",
+    format_user=StringFormatter(slots=["[Round {{idx}}]\n\n问：{{content}}\n\n答："]),
+    format_separator=EmptyFormatter(slots=["\n\n"]),
+    format_prefix=EmptyFormatter(slots=[{"token": "[gMASK]"}, {"token": "sop"}]),
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="chatglm3",
+    format_user=StringFormatter(slots=[{"token": "<|user|>"}, "\n", "{{content}}", {"token": "<|assistant|>"}]),
+    format_assistant=StringFormatter(slots=["\n", "{{content}}"]),
+    format_system=StringFormatter(slots=[{"token": "<|system|>"}, "\n", "{{content}}"]),
+    format_function=FunctionFormatter(slots=[], tool_format="glm4"),
+    format_observation=StringFormatter(
+        slots=[{"token": "<|observation|>"}, "\n", "{{content}}", {"token": "<|assistant|>"}]
+    ),
+    format_tools=ToolFormatter(tool_format="glm4"),
+    format_prefix=EmptyFormatter(slots=[{"token": "[gMASK]"}, {"token": "sop"}]),
+    stop_words=["<|user|>", "<|observation|>"],
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="chatml",
+    format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
+    format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
+    format_observation=StringFormatter(slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+    stop_words=["<|im_end|>", "<|im_start|>"],
+    replace_eos=True,
+)
+
+
+_register_template(
+    name="chatml_de",
+    format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
+    format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
+    format_observation=StringFormatter(slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+    default_system="Du bist ein freundlicher und hilfsbereiter KI-Assistent.",
+    stop_words=["<|im_end|>", "<|im_start|>"],
+    replace_eos=True,
+)
+
+
+_register_template(
+    name="codegeex2",
+    format_prefix=EmptyFormatter(slots=[{"token": "[gMASK]"}, {"token": "sop"}]),
+)
+
+
+_register_template(
+    name="codegeex4",
+    format_user=StringFormatter(slots=["<|user|>\n{{content}}<|assistant|>\n"]),
+    format_system=StringFormatter(slots=["<|system|>\n{{content}}"]),
+    format_function=FunctionFormatter(slots=[], tool_format="glm4"),
+    format_observation=StringFormatter(slots=["<|observation|>\n{{content}}<|assistant|>\n"]),
+    format_tools=ToolFormatter(tool_format="glm4"),
+    format_prefix=EmptyFormatter(slots=["[gMASK]<sop>"]),
+    default_system=(
+        "你是一位智能编程助手，你叫CodeGeeX。你会为用户回答关于编程、代码、计算机方面的任何问题，"
+        "并提供格式规范、可以执行、准确安全的代码，并在必要时提供详细的解释。"
+    ),
+    stop_words=["<|user|>", "<|observation|>"],
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="cohere",
+    format_user=StringFormatter(
+        slots=[
+            (
+                "<|START_OF_TURN_TOKEN|><|USER_TOKEN|>{{content}}<|END_OF_TURN_TOKEN|>"
+                "<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>"
+            )
+        ]
+    ),
+    format_system=StringFormatter(slots=["<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>{{content}}<|END_OF_TURN_TOKEN|>"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+)
+
+
+_register_template(
+    name="cpm",
+    format_user=StringFormatter(slots=["<用户>{{content}}<AI>"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+)
+
+
+_register_template(
+    name="dbrx",
+    format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
+    format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
+    format_observation=StringFormatter(slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+    default_system=(
+        "You are DBRX, created by Databricks. You were last updated in December 2023. "
+        "You answer questions based on information available up to that point.\n"
+        "YOU PROVIDE SHORT RESPONSES TO SHORT QUESTIONS OR STATEMENTS, but provide thorough "
+        "responses to more complex and open-ended questions.\nYou assist with various tasks, "
+        "from writing to coding (using markdown for code blocks — remember to use ``` with "
+        "code, JSON, and tables).\n(You do not have real-time data access or code execution "
+        "capabilities. You avoid stereotyping and provide balanced perspectives on "
+        "controversial topics. You do not provide song lyrics, poems, or news articles and "
+        "do not divulge details of your training data.)\nThis is your system prompt, "
+        "guiding your responses. Do not reference it, just respond to the user. If you find "
+        "yourself talking about this message, stop. You should be responding appropriately "
+        "and usually that means not mentioning this.\nYOU DO NOT MENTION ANY OF THIS INFORMATION "
+        "ABOUT YOURSELF UNLESS THE INFORMATION IS DIRECTLY PERTINENT TO THE USER'S QUERY."
+    ),
+    stop_words=["<|im_end|>"],
+    replace_eos=True,
+)
+
+
+_register_template(
+    name="deepseek",
+    format_user=StringFormatter(slots=["User: {{content}}\n\nAssistant:"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+)
+
+
+_register_template(
+    name="deepseekcoder",
+    format_user=StringFormatter(slots=["### Instruction:\n{{content}}\n### Response:"]),
+    format_assistant=StringFormatter(slots=["\n{{content}}\n"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+    default_system=(
+        "You are an AI programming assistant, utilizing the Deepseek Coder model, "
+        "developed by Deepseek Company, and you only answer questions related to computer science. "
+        "For politically sensitive questions, security and privacy issues, "
+        "and other non-computer science questions, you will refuse to answer\n"
+    ),
+)
+
+
+_register_template(
+    name="default",
+    format_user=StringFormatter(slots=["Human: {{content}}\nAssistant:"]),
+    format_system=StringFormatter(slots=["{{content}}\n"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+)
+
+
+_register_template(
+    name="empty",
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="falcon",
+    format_user=StringFormatter(slots=["User: {{content}}\nFalcon:"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="fewshot",
+    format_separator=EmptyFormatter(slots=["\n\n"]),
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="gemma",
+    format_user=StringFormatter(slots=["<start_of_turn>user\n{{content}}<end_of_turn>\n<start_of_turn>model\n"]),
+    format_observation=StringFormatter(
+        slots=["<start_of_turn>tool\n{{content}}<end_of_turn>\n<start_of_turn>model\n"]
+    ),
+    format_separator=EmptyFormatter(slots=["<end_of_turn>\n"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="glm4",
+    format_user=StringFormatter(slots=["<|user|>\n{{content}}<|assistant|>"]),
+    format_assistant=StringFormatter(slots=["\n{{content}}"]),
+    format_system=StringFormatter(slots=["<|system|>\n{{content}}"]),
+    format_function=FunctionFormatter(slots=[], tool_format="glm4"),
+    format_observation=StringFormatter(slots=["<|observation|>\n{{content}}<|assistant|>"]),
+    format_tools=ToolFormatter(tool_format="glm4"),
+    format_prefix=EmptyFormatter(slots=["[gMASK]<sop>"]),
+    stop_words=["<|user|>", "<|observation|>"],
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="intern",
+    format_user=StringFormatter(slots=["<|User|>:{{content}}\n<|Bot|>:"]),
+    format_system=StringFormatter(slots=["<|System|>:{{content}}\n"]),
+    format_separator=EmptyFormatter(slots=["<eoa>\n"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+    stop_words=["<eoa>"],
+    efficient_eos=True,  # internlm tokenizer cannot set eos_token_id
+)
+
+
+_register_template(
+    name="intern2",
+    format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
+    format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
+    format_separator=EmptyFormatter(slots=["<|im_end|>\n"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+    stop_words=["<|im_end|>"],
+    efficient_eos=True,  # internlm2 tokenizer cannot set eos_token_id
+)
+
+
+_register_template(
+    name="llama2",
+    format_user=StringFormatter(slots=[{"bos_token"}, "[INST] {{content}} [/INST]"]),
+    format_system=StringFormatter(slots=["<<SYS>>\n{{content}}\n<</SYS>>\n\n"]),
+)
+
+
+_register_template(
+    name="llama2_zh",
+    format_user=StringFormatter(slots=[{"bos_token"}, "[INST] {{content}} [/INST]"]),
+    format_system=StringFormatter(slots=["<<SYS>>\n{{content}}\n<</SYS>>\n\n"]),
+    default_system="You are a helpful assistant. 你是一个乐于助人的助手。",
+)
+
+
+_register_template(
+    name="llama3",
+    format_user=StringFormatter(
+        slots=[
+            (
+                "<|start_header_id|>user<|end_header_id|>\n\n{{content}}<|eot_id|>"
+                "<|start_header_id|>assistant<|end_header_id|>\n\n"
+            )
+        ]
+    ),
+    format_system=StringFormatter(slots=["<|start_header_id|>system<|end_header_id|>\n\n{{content}}<|eot_id|>"]),
+    format_observation=StringFormatter(
+        slots=[
+            (
+                "<|start_header_id|>tool<|end_header_id|>\n\n{{content}}<|eot_id|>"
+                "<|start_header_id|>assistant<|end_header_id|>\n\n"
+            )
+        ]
+    ),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+    stop_words=["<|eot_id|>"],
+    replace_eos=True,
+)
+
+
+_register_template(
+    name="mistral",
+    format_user=StringFormatter(slots=["[INST] {{content}} [/INST]"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+)
+
+
+_register_template(
+    name="olmo",
+    format_user=StringFormatter(slots=["<|user|>\n{{content}}<|assistant|>\n"]),
+    format_prefix=EmptyFormatter(slots=[{"eos_token"}]),
+)
+
+
+_register_template(
+    name="openchat",
+    format_user=StringFormatter(slots=["GPT4 Correct User: {{content}}", {"eos_token"}, "GPT4 Correct Assistant:"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+)
+
+
+_register_template(
+    name="openchat-3.6",
+    format_user=StringFormatter(
+        slots=[
+            (
+                "<|start_header_id|>GPT4 Correct User<|end_header_id|>\n\n{{content}}<|eot_id|>"
+                "<|start_header_id|>GPT4 Correct Assistant<|end_header_id|>\n\n"
+            )
+        ]
+    ),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+    stop_words=["<|eot_id|>"],
+    replace_eos=True,
+)
+
+
+_register_template(
+    name="orion",
+    format_user=StringFormatter(slots=["Human: {{content}}\n\nAssistant: ", {"eos_token"}]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+)
+
+
+_register_template(
+    name="phi",
+    format_user=StringFormatter(slots=["<|user|>\n{{content}}<|end|>\n<|assistant|>\n"]),
+    format_system=StringFormatter(slots=["<|system|>\n{{content}}<|end|>\n"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+    format_prefix=EmptyFormatter(slots=[{"bos_token"}]),
+    stop_words=["<|end|>"],
+    replace_eos=True,
+)
+
+
+_register_template(
+    name="qwen",
+    format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
+    format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
+    format_observation=StringFormatter(slots=["<|im_start|>tool\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+    default_system="You are a helpful assistant.",
+    stop_words=["<|im_end|>"],
+    replace_eos=True,
+)
+
+
+_register_template(
+    name="solar",
+    format_user=StringFormatter(slots=["### User:\n{{content}}\n\n### Assistant:\n"]),
+    format_system=StringFormatter(slots=["### System:\n{{content}}\n\n"]),
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="starchat",
+    format_user=StringFormatter(slots=["<|user|>\n{{content}}<|end|>\n<|assistant|>"]),
+    format_system=StringFormatter(slots=["<|system|>\n{{content}}<|end|>\n"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+    stop_words=["<|end|>"],
+    replace_eos=True,
+)
+
+
+_register_template(
+    name="telechat",
+    format_user=StringFormatter(slots=["<_user>{{content}}<_bot>"]),
+    format_system=StringFormatter(slots=["<_system>{{content}}<_end>"]),
+    stop_words=["<_end>"],
+    replace_eos=True,
+)
+
+
+_register_template(
+    name="vicuna",
+    format_user=StringFormatter(slots=["USER: {{content}} ASSISTANT:"]),
+    default_system=(
+        "A chat between a curious user and an artificial intelligence assistant. "
+        "The assistant gives helpful, detailed, and polite answers to the user's questions."
+    ),
+)
+
+
+_register_template(
+    name="xuanyuan",
+    format_user=StringFormatter(slots=["Human: {{content}} Assistant:"]),
+    default_system=(
+        "以下是用户和人工智能助手之间的对话。用户以Human开头，人工智能助手以Assistant开头，"
+        "会对人类提出的问题给出有帮助、高质量、详细和礼貌的回答，并且总是拒绝参与与不道德、"
+        "不安全、有争议、政治敏感等相关的话题、问题和指示。\n"
+    ),
+)
+
+
+_register_template(
+    name="xverse",
+    format_user=StringFormatter(slots=["Human: {{content}}\n\nAssistant: "]),
+)
+
+
+_register_template(
+    name="yayi",
+    format_user=StringFormatter(slots=[{"token": "<|Human|>"}, ":\n{{content}}\n\n", {"token": "<|YaYi|>"}, ":"]),
+    format_system=StringFormatter(slots=[{"token": "<|System|>"}, ":\n{{content}}\n\n"]),
+    format_separator=EmptyFormatter(slots=["\n\n"]),
+    default_system=(
+        "You are a helpful, respectful and honest assistant named YaYi "
+        "developed by Beijing Wenge Technology Co.,Ltd. "
+        "Always answer as helpfully as possible, while being safe.  "
+        "Your answers should not include any harmful, unethical, "
+        "racist, sexist, toxic, dangerous, or illegal content. "
+        "Please ensure that your responses are socially unbiased and positive in nature.\n\n"
+        "If a question does not make any sense, or is not factually coherent, "
+        "explain why instead of answering something not correct. "
+        "If you don't know the answer to a question, please don't share false information."
+    ),
+    stop_words=["<|End|>"],
+)
+
+
+_register_template(
+    name="yi",
+    format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
+    format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+    stop_words=["<|im_end|>"],
+    replace_eos=True,
+)
+
+
+_register_template(
+    name="yi_vl",
+    format_user=StringFormatter(slots=["### Human: {{content}}\n### Assistant:"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+    default_system=(
+        "This is a chat between an inquisitive human and an AI assistant. "
+        "Assume the role of the AI assistant. Read all the images carefully, "
+        "and respond to the human's questions with informative, helpful, detailed and polite answers. "
+        "这是一个好奇的人类和一个人工智能助手之间的对话。假设你扮演这个AI助手的角色。"
+        "仔细阅读所有的图像，并对人类的问题做出信息丰富、有帮助、详细的和礼貌的回答。\n\n"
+    ),
+    stop_words=["###"],
+    efficient_eos=True,
+)
+
+
+_register_template(
+    name="yuan",
+    format_user=StringFormatter(slots=["{{content}}", {"token": "<sep>"}]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+    stop_words=["<eod>"],
+    replace_eos=True,
+)
+
+
+_register_template(
+    name="zephyr",
+    format_user=StringFormatter(slots=["<|user|>\n{{content}}", {"eos_token"}, "<|assistant|>\n"]),
+    format_system=StringFormatter(slots=["<|system|>\n{{content}}", {"eos_token"}]),
+    default_system="You are Zephyr, a helpful assistant.",
+)
+
+
+_register_template(
+    name="ziya",
+    format_user=StringFormatter(slots=["<human>:{{content}}\n<bot>:"]),
+    format_separator=EmptyFormatter(slots=["\n"]),
+)

llama-factory/src/llamafactory/data/tool_utils.py

@@ -0,0 +1,140 @@
+# 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 re
+from abc import ABC, abstractmethod
+from dataclasses import dataclass
+from typing import Any, Dict, List, Tuple, Union
+
+from .data_utils import SLOTS
+
+
+DEFAULT_TOOL_PROMPT = (
+    "You have access to the following tools:\n{tool_text}"
+    "Use the following format if using a tool:\n"
+    "```\n"
+    "Action: tool name (one of [{tool_names}])\n"
+    "Action Input: the input to the tool, in a JSON format representing the kwargs "
+    """(e.g. ```{{"input": "hello world", "num_beams": 5}}```)\n"""
+    "```\n"
+)
+
+
+GLM4_TOOL_PROMPT = (
+    "你是一个名为 ChatGLM 的人工智能助手。你是基于智谱AI训练的语言模型 GLM-4 模型开发的，"
+    "你的任务是针对用户的问题和要求提供适当的答复和支持。# 可用工具{tool_text}"
+)
+
+
+@dataclass
+class ToolUtils(ABC):
+    @staticmethod
+    @abstractmethod
+    def get_function_slots() -> SLOTS: ...
+
+    @staticmethod
+    @abstractmethod
+    def tool_formatter(tools: List[Dict[str, Any]]) -> str: ...
+
+    @staticmethod
+    @abstractmethod
+    def tool_extractor(content: str) -> Union[str, List[Tuple[str, str]]]: ...
+
+
+class DefaultToolUtils(ToolUtils):
+    @staticmethod
+    def get_function_slots() -> SLOTS:
+        return ["Action: {{name}}\nAction Input: {{arguments}}\n"]
+
+    @staticmethod
+    def tool_formatter(tools: List[Dict[str, Any]]) -> str:
+        tool_text = ""
+        tool_names = []
+        for tool in tools:
+            param_text = ""
+            for name, param in tool["parameters"]["properties"].items():
+                required, enum, items = "", "", ""
+                if name in tool["parameters"].get("required", []):
+                    required = ", required"
+
+                if param.get("enum", None):
+                    enum = ", should be one of [{}]".format(", ".join(param["enum"]))
+
+                if param.get("items", None):
+                    items = ", where each item should be {}".format(param["items"].get("type", ""))
+
+                param_text += "  - {name} ({type}{required}): {desc}{enum}{items}\n".format(
+                    name=name,
+                    type=param.get("type", ""),
+                    required=required,
+                    desc=param.get("description", ""),
+                    enum=enum,
+                    items=items,
+                )
+
+            tool_text += "> Tool Name: {name}\nTool Description: {desc}\nTool Args:\n{args}\n".format(
+                name=tool["name"], desc=tool.get("description", ""), args=param_text
+            )
+            tool_names.append(tool["name"])
+
+        return DEFAULT_TOOL_PROMPT.format(tool_text=tool_text, tool_names=", ".join(tool_names))
+
+    @staticmethod
+    def tool_extractor(content: str) -> Union[str, List[Tuple[str, str]]]:
+        regex = re.compile(r"Action:\s*([a-zA-Z0-9_]+)\s*Action Input:\s*(.+?)(?=\s*Action:|\s*$)", re.DOTALL)
+        action_match: List[Tuple[str, str]] = re.findall(regex, content)
+        if not action_match:
+            return content
+
+        results = []
+        for match in action_match:
+            tool_name = match[0].strip()
+            tool_input = match[1].strip().strip('"').strip("```")
+            try:
+                arguments = json.loads(tool_input)
+                results.append((tool_name, json.dumps(arguments, ensure_ascii=False)))
+            except json.JSONDecodeError:
+                return content
+
+        return results
+
+
+class GLM4ToolUtils(ToolUtils):
+    @staticmethod
+    def get_function_slots() -> SLOTS:
+        return ["{{name}}\n{{arguments}}"]
+
+    @staticmethod
+    def tool_formatter(tools: List[Dict[str, Any]]) -> str:
+        tool_text = ""
+        for tool in tools:
+            tool_text += "\n\n## {name}\n\n{body}\n在调用上述函数时，请使用 Json 格式表示调用的参数。".format(
+                name=tool["name"], body=json.dumps(tool, indent=4, ensure_ascii=False)
+            )
+
+        return GLM4_TOOL_PROMPT.format(tool_text=tool_text)
+
+    @staticmethod
+    def tool_extractor(content: str) -> Union[str, List[Tuple[str, str]]]:
+        if "\n" not in content:
+            return content
+
+        tool_name, tool_input = content.split("\n", maxsplit=1)
+        try:
+            arguments = json.loads(tool_input)
+        except json.JSONDecodeError:
+            return content
+
+        return [(tool_name, json.dumps(arguments, ensure_ascii=False))]

llama-factory/src/llamafactory/eval/evaluator.py

@@ -0,0 +1,154 @@
+# Copyright 2024 the LlamaFactory team.
+#
+# This code is inspired by the Dan's test library.
+# https://github.com/hendrycks/test/blob/master/evaluate_flan.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.
+#
+# MIT License
+#
+# Copyright (c) 2020 Dan Hendrycks
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+# copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in all
+# copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+# SOFTWARE.
+
+import json
+import os
+from typing import Any, Dict, List, Optional
+
+import numpy as np
+import torch
+from datasets import load_dataset
+from tqdm import tqdm, trange
+from transformers.utils import cached_file
+
+from ..data import get_template_and_fix_tokenizer
+from ..extras.constants import CHOICES, SUBJECTS
+from ..hparams import get_eval_args
+from ..model import load_model, load_tokenizer
+from .template import get_eval_template
+
+
+class Evaluator:
+    def __init__(self, args: Optional[Dict[str, Any]] = None) -> None:
+        self.model_args, self.data_args, self.eval_args, finetuning_args = get_eval_args(args)
+        self.tokenizer = load_tokenizer(self.model_args)["tokenizer"]
+        self.tokenizer.padding_side = "right"  # avoid overflow issue in batched inference for llama2
+        self.template = get_template_and_fix_tokenizer(self.tokenizer, self.data_args.template)
+        self.model = load_model(self.tokenizer, self.model_args, finetuning_args)
+        self.eval_template = get_eval_template(self.eval_args.lang)
+        self.choice_inputs = [self.tokenizer.encode(ch, add_special_tokens=False)[-1] for ch in CHOICES]
+
+    @torch.inference_mode()
+    def batch_inference(self, batch_input: Dict[str, torch.Tensor]) -> List[str]:
+        logits = self.model(**batch_input).logits
+        lengths = torch.sum(batch_input["attention_mask"], dim=-1)
+        word_probs = torch.stack([logits[i, lengths[i] - 1] for i in range(len(lengths))], dim=0)
+        choice_probs = torch.nn.functional.softmax(word_probs[:, self.choice_inputs], dim=-1).detach()
+        return [chr(ord("A") + offset.item()) for offset in torch.argmax(choice_probs, dim=-1)]
+
+    def eval(self) -> None:
+        eval_task = self.eval_args.task.split("_")[0]
+        eval_split = self.eval_args.task.split("_")[1]
+
+        mapping = cached_file(
+            path_or_repo_id=os.path.join(self.eval_args.task_dir, eval_task),
+            filename="mapping.json",
+            cache_dir=self.model_args.cache_dir,
+            token=self.model_args.hf_hub_token,
+        )
+
+        with open(mapping, "r", encoding="utf-8") as f:
+            categorys: Dict[str, Dict[str, str]] = json.load(f)
+
+        category_corrects = {subj: np.array([], dtype="bool") for subj in SUBJECTS}
+        pbar = tqdm(categorys.keys(), desc="Processing subjects", position=0)
+        results = {}
+        for subject in pbar:
+            dataset = load_dataset(
+                path=os.path.join(self.eval_args.task_dir, eval_task),
+                name=subject,
+                cache_dir=self.model_args.cache_dir,
+                download_mode=self.eval_args.download_mode,
+                token=self.model_args.hf_hub_token,
+                trust_remote_code=True,
+            )
+            pbar.set_postfix_str(categorys[subject]["name"])
+            inputs, outputs, labels = [], [], []
+            for i in trange(len(dataset[eval_split]), desc="Formatting batches", position=1, leave=False):
+                support_set = (
+                    dataset["train"].shuffle().select(range(min(self.eval_args.n_shot, len(dataset["train"]))))
+                )
+                messages = self.eval_template.format_example(
+                    target_data=dataset[eval_split][i],
+                    support_set=support_set,
+                    subject_name=categorys[subject]["name"],
+                )
+
+                input_ids, _ = self.template.encode_oneturn(tokenizer=self.tokenizer, messages=messages)
+                inputs.append({"input_ids": input_ids, "attention_mask": [1] * len(input_ids)})
+                labels.append(messages[-1]["content"])
+
+            for i in trange(
+                0, len(inputs), self.eval_args.batch_size, desc="Predicting batches", position=1, leave=False
+            ):
+                batch_input = self.tokenizer.pad(
+                    inputs[i : i + self.eval_args.batch_size], return_attention_mask=True, return_tensors="pt"
+                ).to(self.model.device)
+                preds = self.batch_inference(batch_input)
+                outputs += preds
+
+            corrects = np.array(outputs) == np.array(labels)
+            category_name = categorys[subject]["category"]
+            category_corrects[category_name] = np.concatenate([category_corrects[category_name], corrects], axis=0)
+            category_corrects["Average"] = np.concatenate([category_corrects["Average"], corrects], axis=0)
+            results[subject] = {str(i): outputs[i] for i in range(len(outputs))}
+
+        pbar.close()
+        self._save_results(category_corrects, results)
+
+    def _save_results(self, category_corrects: Dict[str, np.ndarray], results: Dict[str, Dict[int, str]]) -> None:
+        score_info = "\n".join(
+            [
+                "{:>15}: {:.2f}".format(category_name, 100 * np.mean(category_correct))
+                for category_name, category_correct in category_corrects.items()
+                if len(category_correct)
+            ]
+        )
+        print(score_info)
+        if self.eval_args.save_dir is not None:
+            os.makedirs(self.eval_args.save_dir, exist_ok=False)
+            with open(os.path.join(self.eval_args.save_dir, "results.json"), "w", encoding="utf-8", newline="\n") as f:
+                json.dump(results, f, indent=2)
+
+            with open(os.path.join(self.eval_args.save_dir, "results.log"), "w", encoding="utf-8", newline="\n") as f:
+                f.write(score_info)
+
+
+def run_eval() -> None:
+    Evaluator().eval()

llama-factory/src/llamafactory/eval/template.py

@@ -0,0 +1,81 @@
+# 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 dataclasses import dataclass
+from typing import Dict, List, Sequence, Tuple
+
+from ..data import Role
+from ..extras.constants import CHOICES
+
+
+@dataclass
+class EvalTemplate:
+    system: str
+    choice: str
+    answer: str
+
+    def _parse_example(self, example: Dict[str, str]) -> Tuple[str, str]:
+        r"""
+        input: a dict with keys {"question", "A", "B", "C", "D", "answer"}
+        output: a tuple of (prompt, response)
+        """
+        candidates = [self.choice.format(choice=ch, content=example[ch]) for ch in CHOICES if ch in example]
+        return "".join([example["question"]] + candidates + [self.answer]), example["answer"]
+
+    def format_example(
+        self, target_data: Dict[str, str], support_set: Sequence[Dict[str, str]], subject_name: str
+    ) -> List[Dict[str, str]]:
+        r"""
+        Converts dataset examples to messages.
+        """
+        messages = []
+        for k in range(len(support_set)):
+            prompt, response = self._parse_example(support_set[k])
+            messages.append({"role": Role.USER.value, "content": prompt})
+            messages.append({"role": Role.ASSISTANT.value, "content": response})
+
+        prompt, response = self._parse_example(target_data)
+        messages.append({"role": Role.USER.value, "content": prompt})
+        messages.append({"role": Role.ASSISTANT.value, "content": response})
+        messages[0]["content"] = self.system.format(subject=subject_name) + messages[0]["content"]
+        return messages
+
+
+eval_templates: Dict[str, "EvalTemplate"] = {}
+
+
+def _register_eval_template(name: str, system: str, choice: str, answer: str) -> None:
+    eval_templates[name] = EvalTemplate(system=system, choice=choice, answer=answer)
+
+
+def get_eval_template(name: str) -> "EvalTemplate":
+    eval_template = eval_templates.get(name, None)
+    assert eval_template is not None, "Template {} does not exist.".format(name)
+    return eval_template
+
+
+_register_eval_template(
+    name="en",
+    system="The following are multiple choice questions (with answers) about {subject}.\n\n",
+    choice="\n{choice}. {content}",
+    answer="\nAnswer:",
+)
+
+
+_register_eval_template(
+    name="zh",
+    system="以下是中国关于{subject}考试的单项选择题，请选出其中的正确答案。\n\n",
+    choice="\n{choice}. {content}",
+    answer="\n答案：",
+)

llama-factory/src/llamafactory/extras/constants.py

@@ -0,0 +1,1590 @@
+# 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 collections import OrderedDict, defaultdict
+from enum import Enum
+from typing import Dict, Optional
+
+from peft.utils import SAFETENSORS_WEIGHTS_NAME as SAFE_ADAPTER_WEIGHTS_NAME
+from peft.utils import WEIGHTS_NAME as ADAPTER_WEIGHTS_NAME
+from transformers.utils import SAFE_WEIGHTS_INDEX_NAME, SAFE_WEIGHTS_NAME, WEIGHTS_INDEX_NAME, WEIGHTS_NAME
+
+
+CHECKPOINT_NAMES = {
+    SAFE_ADAPTER_WEIGHTS_NAME,
+    ADAPTER_WEIGHTS_NAME,
+    SAFE_WEIGHTS_INDEX_NAME,
+    SAFE_WEIGHTS_NAME,
+    WEIGHTS_INDEX_NAME,
+    WEIGHTS_NAME,
+}
+
+CHOICES = ["A", "B", "C", "D"]
+
+DATA_CONFIG = "dataset_info.json"
+
+DEFAULT_TEMPLATE = defaultdict(str)
+
+FILEEXT2TYPE = {
+    "arrow": "arrow",
+    "csv": "csv",
+    "json": "json",
+    "jsonl": "json",
+    "parquet": "parquet",
+    "txt": "text",
+}
+
+IGNORE_INDEX = -100
+
+LAYERNORM_NAMES = {"norm", "ln"}
+
+LLAMABOARD_CONFIG = "llamaboard_config.yaml"
+
+METHODS = ["full", "freeze", "lora"]
+
+MOD_SUPPORTED_MODELS = {"bloom", "falcon", "gemma", "llama", "mistral", "mixtral", "phi", "starcoder2"}
+
+PEFT_METHODS = {"lora"}
+
+RUNNING_LOG = "running_log.txt"
+
+SUBJECTS = ["Average", "STEM", "Social Sciences", "Humanities", "Other"]
+
+SUPPORTED_MODELS = OrderedDict()
+
+TRAINER_LOG = "trainer_log.jsonl"
+
+TRAINING_ARGS = "training_args.yaml"
+
+TRAINING_STAGES = {
+    "Supervised Fine-Tuning": "sft",
+    "Reward Modeling": "rm",
+    "PPO": "ppo",
+    "DPO": "dpo",
+    "KTO": "kto",
+    "Pre-Training": "pt",
+}
+
+STAGES_USE_PAIR_DATA = {"rm", "dpo"}
+
+SUPPORTED_CLASS_FOR_BLOCK_DIAG_ATTN = {
+    "cohere",
+    "falcon",
+    "gemma",
+    "gemma2",
+    "llama",
+    "mistral",
+    "phi",
+    "phi3",
+    "qwen2",
+    "starcoder2",
+}
+
+SUPPORTED_CLASS_FOR_S2ATTN = {"llama"}
+
+V_HEAD_WEIGHTS_NAME = "value_head.bin"
+
+V_HEAD_SAFE_WEIGHTS_NAME = "value_head.safetensors"
+
+VISION_MODELS = set()
+
+
+class DownloadSource(str, Enum):
+    DEFAULT = "hf"
+    MODELSCOPE = "ms"
+
+
+def register_model_group(
+    models: Dict[str, Dict[DownloadSource, str]],
+    template: Optional[str] = None,
+    vision: bool = False,
+) -> None:
+    prefix = None
+    for name, path in models.items():
+        if prefix is None:
+            prefix = name.split("-")[0]
+        else:
+            assert prefix == name.split("-")[0], "prefix should be identical."
+        SUPPORTED_MODELS[name] = path
+    if template is not None:
+        DEFAULT_TEMPLATE[prefix] = template
+    if vision:
+        VISION_MODELS.add(prefix)
+
+
+register_model_group(
+    models={
+        "Aya-23-8B-Chat": {
+            DownloadSource.DEFAULT: "CohereForAI/aya-23-8B",
+        },
+        "Aya-23-35B-Chat": {
+            DownloadSource.DEFAULT: "CohereForAI/aya-23-35B",
+        },
+    },
+    template="cohere",
+)
+
+
+register_model_group(
+    models={
+        "Baichuan-7B-Base": {
+            DownloadSource.DEFAULT: "baichuan-inc/Baichuan-7B",
+            DownloadSource.MODELSCOPE: "baichuan-inc/baichuan-7B",
+        },
+        "Baichuan-13B-Base": {
+            DownloadSource.DEFAULT: "baichuan-inc/Baichuan-13B-Base",
+            DownloadSource.MODELSCOPE: "baichuan-inc/Baichuan-13B-Base",
+        },
+        "Baichuan-13B-Chat": {
+            DownloadSource.DEFAULT: "baichuan-inc/Baichuan-13B-Chat",
+            DownloadSource.MODELSCOPE: "baichuan-inc/Baichuan-13B-Chat",
+        },
+    },
+    template="baichuan",
+)
+
+
+register_model_group(
+    models={
+        "Baichuan2-7B-Base": {
+            DownloadSource.DEFAULT: "baichuan-inc/Baichuan2-7B-Base",
+            DownloadSource.MODELSCOPE: "baichuan-inc/Baichuan2-7B-Base",
+        },
+        "Baichuan2-13B-Base": {
+            DownloadSource.DEFAULT: "baichuan-inc/Baichuan2-13B-Base",
+            DownloadSource.MODELSCOPE: "baichuan-inc/Baichuan2-13B-Base",
+        },
+        "Baichuan2-7B-Chat": {
+            DownloadSource.DEFAULT: "baichuan-inc/Baichuan2-7B-Chat",
+            DownloadSource.MODELSCOPE: "baichuan-inc/Baichuan2-7B-Chat",
+        },
+        "Baichuan2-13B-Chat": {
+            DownloadSource.DEFAULT: "baichuan-inc/Baichuan2-13B-Chat",
+            DownloadSource.MODELSCOPE: "baichuan-inc/Baichuan2-13B-Chat",
+        },
+    },
+    template="baichuan2",
+)
+
+
+register_model_group(
+    models={
+        "BLOOM-560M": {
+            DownloadSource.DEFAULT: "bigscience/bloom-560m",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/bloom-560m",
+        },
+        "BLOOM-3B": {
+            DownloadSource.DEFAULT: "bigscience/bloom-3b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/bloom-3b",
+        },
+        "BLOOM-7B1": {
+            DownloadSource.DEFAULT: "bigscience/bloom-7b1",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/bloom-7b1",
+        },
+    },
+)
+
+
+register_model_group(
+    models={
+        "BLOOMZ-560M": {
+            DownloadSource.DEFAULT: "bigscience/bloomz-560m",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/bloomz-560m",
+        },
+        "BLOOMZ-3B": {
+            DownloadSource.DEFAULT: "bigscience/bloomz-3b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/bloomz-3b",
+        },
+        "BLOOMZ-7B1-mt": {
+            DownloadSource.DEFAULT: "bigscience/bloomz-7b1-mt",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/bloomz-7b1-mt",
+        },
+    },
+)
+
+
+register_model_group(
+    models={
+        "BlueLM-7B-Base": {
+            DownloadSource.DEFAULT: "vivo-ai/BlueLM-7B-Base",
+            DownloadSource.MODELSCOPE: "vivo-ai/BlueLM-7B-Base",
+        },
+        "BlueLM-7B-Chat": {
+            DownloadSource.DEFAULT: "vivo-ai/BlueLM-7B-Chat",
+            DownloadSource.MODELSCOPE: "vivo-ai/BlueLM-7B-Chat",
+        },
+    },
+    template="bluelm",
+)
+
+
+register_model_group(
+    models={
+        "Breeze-7B": {
+            DownloadSource.DEFAULT: "MediaTek-Research/Breeze-7B-Base-v1_0",
+        },
+        "Breeze-7B-Chat": {
+            DownloadSource.DEFAULT: "MediaTek-Research/Breeze-7B-Instruct-v1_0",
+        },
+    },
+    template="breeze",
+)
+
+
+register_model_group(
+    models={
+        "ChatGLM2-6B-Chat": {
+            DownloadSource.DEFAULT: "THUDM/chatglm2-6b",
+            DownloadSource.MODELSCOPE: "ZhipuAI/chatglm2-6b",
+        }
+    },
+    template="chatglm2",
+)
+
+
+register_model_group(
+    models={
+        "ChatGLM3-6B-Base": {
+            DownloadSource.DEFAULT: "THUDM/chatglm3-6b-base",
+            DownloadSource.MODELSCOPE: "ZhipuAI/chatglm3-6b-base",
+        },
+        "ChatGLM3-6B-Chat": {
+            DownloadSource.DEFAULT: "THUDM/chatglm3-6b",
+            DownloadSource.MODELSCOPE: "ZhipuAI/chatglm3-6b",
+        },
+    },
+    template="chatglm3",
+)
+
+
+register_model_group(
+    models={
+        "ChineseLLaMA2-1.3B": {
+            DownloadSource.DEFAULT: "hfl/chinese-llama-2-1.3b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/chinese-llama-2-1.3b",
+        },
+        "ChineseLLaMA2-7B": {
+            DownloadSource.DEFAULT: "hfl/chinese-llama-2-7b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/chinese-llama-2-7b",
+        },
+        "ChineseLLaMA2-13B": {
+            DownloadSource.DEFAULT: "hfl/chinese-llama-2-13b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/chinese-llama-2-13b",
+        },
+        "ChineseLLaMA2-1.3B-Chat": {
+            DownloadSource.DEFAULT: "hfl/chinese-alpaca-2-1.3b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/chinese-alpaca-2-1.3b",
+        },
+        "ChineseLLaMA2-7B-Chat": {
+            DownloadSource.DEFAULT: "hfl/chinese-alpaca-2-7b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/chinese-alpaca-2-7b",
+        },
+        "ChineseLLaMA2-13B-Chat": {
+            DownloadSource.DEFAULT: "hfl/chinese-alpaca-2-13b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/chinese-alpaca-2-13b",
+        },
+    },
+    template="llama2_zh",
+)
+
+
+register_model_group(
+    models={
+        "CodeGeeX4-9B-Chat": {
+            DownloadSource.DEFAULT: "THUDM/codegeex4-all-9b",
+            DownloadSource.MODELSCOPE: "ZhipuAI/codegeex4-all-9b",
+        },
+    },
+    template="codegeex4",
+)
+
+
+register_model_group(
+    models={
+        "CodeGemma-7B": {
+            DownloadSource.DEFAULT: "google/codegemma-7b",
+        },
+        "CodeGemma-7B-Chat": {
+            DownloadSource.DEFAULT: "google/codegemma-7b-it",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/codegemma-7b-it",
+        },
+        "CodeGemma-1.1-2B": {
+            DownloadSource.DEFAULT: "google/codegemma-1.1-2b",
+        },
+        "CodeGemma-1.1-7B-Chat": {
+            DownloadSource.DEFAULT: "google/codegemma-1.1-7b-it",
+        },
+    },
+    template="gemma",
+)
+
+
+register_model_group(
+    models={
+        "Codestral-22B-v0.1-Chat": {
+            DownloadSource.DEFAULT: "mistralai/Codestral-22B-v0.1",
+        },
+    },
+    template="mistral",
+)
+
+
+register_model_group(
+    models={
+        "CommandR-35B-Chat": {
+            DownloadSource.DEFAULT: "CohereForAI/c4ai-command-r-v01",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/c4ai-command-r-v01",
+        },
+        "CommandR-Plus-104B-Chat": {
+            DownloadSource.DEFAULT: "CohereForAI/c4ai-command-r-plus",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/c4ai-command-r-plus",
+        },
+        "CommandR-35B-4bit-Chat": {
+            DownloadSource.DEFAULT: "CohereForAI/c4ai-command-r-v01-4bit",
+            DownloadSource.MODELSCOPE: "mirror013/c4ai-command-r-v01-4bit",
+        },
+        "CommandR-Plus-104B-4bit-Chat": {
+            DownloadSource.DEFAULT: "CohereForAI/c4ai-command-r-plus-4bit",
+        },
+    },
+    template="cohere",
+)
+
+
+register_model_group(
+    models={
+        "DBRX-132B-Base": {
+            DownloadSource.DEFAULT: "databricks/dbrx-base",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/dbrx-base",
+        },
+        "DBRX-132B-Chat": {
+            DownloadSource.DEFAULT: "databricks/dbrx-instruct",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/dbrx-instruct",
+        },
+    },
+    template="dbrx",
+)
+
+
+register_model_group(
+    models={
+        "DeepSeek-LLM-7B-Base": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-llm-7b-base",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-llm-7b-base",
+        },
+        "DeepSeek-LLM-67B-Base": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-llm-67b-base",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-llm-67b-base",
+        },
+        "DeepSeek-LLM-7B-Chat": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-llm-7b-chat",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-llm-7b-chat",
+        },
+        "DeepSeek-LLM-67B-Chat": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-llm-67b-chat",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-llm-67b-chat",
+        },
+        "DeepSeek-Math-7B-Base": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-math-7b-base",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-math-7b-base",
+        },
+        "DeepSeek-Math-7B-Chat": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-math-7b-instruct",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-math-7b-instruct",
+        },
+        "DeepSeek-MoE-16B-Base": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-moe-16b-base",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-moe-16b-base",
+        },
+        "DeepSeek-MoE-16B-v2-Base": {
+            DownloadSource.DEFAULT: "deepseek-ai/DeepSeek-V2-Lite",
+            DownloadSource.MODELSCOPE: "deepseek-ai/DeepSeek-V2-Lite",
+        },
+        "DeepSeek-MoE-236B-Base": {
+            DownloadSource.DEFAULT: "deepseek-ai/DeepSeek-V2",
+            DownloadSource.MODELSCOPE: "deepseek-ai/DeepSeek-V2",
+        },
+        "DeepSeek-MoE-16B-Chat": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-moe-16b-chat",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-moe-16b-chat",
+        },
+        "DeepSeek-MoE-16B-v2-Chat": {
+            DownloadSource.DEFAULT: "deepseek-ai/DeepSeek-V2-Lite-Chat",
+            DownloadSource.MODELSCOPE: "deepseek-ai/DeepSeek-V2-Lite-Chat",
+        },
+        "DeepSeek-MoE-236B-Chat": {
+            DownloadSource.DEFAULT: "deepseek-ai/DeepSeek-V2-Chat",
+            DownloadSource.MODELSCOPE: "deepseek-ai/DeepSeek-V2-Chat",
+        },
+        "DeepSeek-MoE-Coder-16B-Base": {
+            DownloadSource.DEFAULT: "deepseek-ai/DeepSeek-Coder-V2-Lite-Base",
+        },
+        "DeepSeek-MoE-Coder-236B-Base": {
+            DownloadSource.DEFAULT: "deepseek-ai/DeepSeek-Coder-V2-Base",
+        },
+        "DeepSeek-MoE-Coder-16B-Chat": {
+            DownloadSource.DEFAULT: "deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct",
+        },
+        "DeepSeek-MoE-Coder-236B-Chat": {
+            DownloadSource.DEFAULT: "deepseek-ai/DeepSeek-Coder-V2-Instruct",
+        },
+    },
+    template="deepseek",
+)
+
+
+register_model_group(
+    models={
+        "DeepSeekCoder-6.7B-Base": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-coder-6.7b-base",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-coder-6.7b-base",
+        },
+        "DeepSeekCoder-7B-Base": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-coder-7b-base-v1.5",
+        },
+        "DeepSeekCoder-33B-Base": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-coder-33b-base",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-coder-33b-base",
+        },
+        "DeepSeekCoder-6.7B-Chat": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-coder-6.7b-instruct",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-coder-6.7b-instruct",
+        },
+        "DeepSeekCoder-7B-Chat": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-coder-7b-instruct-v1.5",
+        },
+        "DeepSeekCoder-33B-Chat": {
+            DownloadSource.DEFAULT: "deepseek-ai/deepseek-coder-33b-instruct",
+            DownloadSource.MODELSCOPE: "deepseek-ai/deepseek-coder-33b-instruct",
+        },
+    },
+    template="deepseekcoder",
+)
+
+
+register_model_group(
+    models={
+        "Falcon-7B": {
+            DownloadSource.DEFAULT: "tiiuae/falcon-7b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/falcon-7b",
+        },
+        "Falcon-11B": {
+            DownloadSource.DEFAULT: "tiiuae/falcon-11B",
+        },
+        "Falcon-40B": {
+            DownloadSource.DEFAULT: "tiiuae/falcon-40b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/falcon-40b",
+        },
+        "Falcon-180B": {
+            DownloadSource.DEFAULT: "tiiuae/falcon-180b",
+            DownloadSource.MODELSCOPE: "modelscope/falcon-180B",
+        },
+        "Falcon-7B-Chat": {
+            DownloadSource.DEFAULT: "tiiuae/falcon-7b-instruct",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/falcon-7b-instruct",
+        },
+        "Falcon-40B-Chat": {
+            DownloadSource.DEFAULT: "tiiuae/falcon-40b-instruct",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/falcon-40b-instruct",
+        },
+        "Falcon-180B-Chat": {
+            DownloadSource.DEFAULT: "tiiuae/falcon-180b-chat",
+            DownloadSource.MODELSCOPE: "modelscope/falcon-180B-chat",
+        },
+    },
+    template="falcon",
+)
+
+
+register_model_group(
+    models={
+        "Gemma-2B": {
+            DownloadSource.DEFAULT: "google/gemma-2b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/gemma-2b",
+        },
+        "Gemma-7B": {
+            DownloadSource.DEFAULT: "google/gemma-7b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/gemma-2b-it",
+        },
+        "Gemma-2B-Chat": {
+            DownloadSource.DEFAULT: "google/gemma-2b-it",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/gemma-7b",
+        },
+        "Gemma-7B-Chat": {
+            DownloadSource.DEFAULT: "google/gemma-7b-it",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/gemma-7b-it",
+        },
+        "Gemma-1.1-2B-Chat": {
+            DownloadSource.DEFAULT: "google/gemma-1.1-2b-it",
+        },
+        "Gemma-1.1-7B-Chat": {
+            DownloadSource.DEFAULT: "google/gemma-1.1-7b-it",
+        },
+        "Gemma-2-9B": {
+            DownloadSource.DEFAULT: "google/gemma-2-9b",
+            DownloadSource.MODELSCOPE: "LLM-Research/gemma-2-9b",
+        },
+        "Gemma-2-27B": {
+            DownloadSource.DEFAULT: "google/gemma-2-27b",
+            DownloadSource.MODELSCOPE: "LLM-Research/gemma-2-27b",
+        },
+        "Gemma-2-9B-Chat": {
+            DownloadSource.DEFAULT: "google/gemma-2-9b-it",
+            DownloadSource.MODELSCOPE: "LLM-Research/gemma-2-9b-it",
+        },
+        "Gemma-2-27B-Chat": {
+            DownloadSource.DEFAULT: "google/gemma-2-27b-it",
+            DownloadSource.MODELSCOPE: "LLM-Research/gemma-2-27b-it",
+        },
+    },
+    template="gemma",
+)
+
+
+register_model_group(
+    models={
+        "GLM-4-9B": {
+            DownloadSource.DEFAULT: "THUDM/glm-4-9b",
+            DownloadSource.MODELSCOPE: "ZhipuAI/glm-4-9b",
+        },
+        "GLM-4-9B-Chat": {
+            DownloadSource.DEFAULT: "THUDM/glm-4-9b-chat",
+            DownloadSource.MODELSCOPE: "ZhipuAI/glm-4-9b-chat",
+        },
+        "GLM-4-9B-1M-Chat": {
+            DownloadSource.DEFAULT: "THUDM/glm-4-9b-chat-1m",
+            DownloadSource.MODELSCOPE: "ZhipuAI/glm-4-9b-chat-1m",
+        },
+    },
+    template="glm4",
+)
+
+
+register_model_group(
+    models={
+        "InternLM-7B": {
+            DownloadSource.DEFAULT: "internlm/internlm-7b",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm-7b",
+        },
+        "InternLM-20B": {
+            DownloadSource.DEFAULT: "internlm/internlm-20b",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm-20b",
+        },
+        "InternLM-7B-Chat": {
+            DownloadSource.DEFAULT: "internlm/internlm-chat-7b",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm-chat-7b",
+        },
+        "InternLM-20B-Chat": {
+            DownloadSource.DEFAULT: "internlm/internlm-chat-20b",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm-chat-20b",
+        },
+    },
+    template="intern",
+)
+
+
+register_model_group(
+    models={
+        "InternLM2-7B": {
+            DownloadSource.DEFAULT: "internlm/internlm2-7b",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm2-7b",
+        },
+        "InternLM2-20B": {
+            DownloadSource.DEFAULT: "internlm/internlm2-20b",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm2-20b",
+        },
+        "InternLM2-7B-Chat": {
+            DownloadSource.DEFAULT: "internlm/internlm2-chat-7b",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm2-chat-7b",
+        },
+        "InternLM2-20B-Chat": {
+            DownloadSource.DEFAULT: "internlm/internlm2-chat-20b",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm2-chat-20b",
+        },
+    },
+    template="intern2",
+)
+
+
+register_model_group(
+    models={
+        "InternLM2.5-7B": {
+            DownloadSource.DEFAULT: "internlm/internlm2_5-7b",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm2_5-7b",
+        },
+        "InternLM2.5-7B-Chat": {
+            DownloadSource.DEFAULT: "internlm/internlm2_5-7b-chat",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm2_5-7b-chat",
+        },
+        "InternLM2.5-7B-1M-Chat": {
+            DownloadSource.DEFAULT: "internlm/internlm2_5-7b-chat-1m",
+            DownloadSource.MODELSCOPE: "Shanghai_AI_Laboratory/internlm2_5-7b-chat-1m",
+        },
+    },
+    template="intern2",
+)
+
+
+register_model_group(
+    models={
+        "Jamba-v0.1": {
+            DownloadSource.DEFAULT: "ai21labs/Jamba-v0.1",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/Jamba-v0.1",
+        }
+    },
+)
+
+
+register_model_group(
+    models={
+        "LingoWhale-8B": {
+            DownloadSource.DEFAULT: "deeplang-ai/LingoWhale-8B",
+            DownloadSource.MODELSCOPE: "DeepLang/LingoWhale-8B",
+        }
+    },
+)
+
+
+register_model_group(
+    models={
+        "LLaMA-7B": {
+            DownloadSource.DEFAULT: "huggyllama/llama-7b",
+            DownloadSource.MODELSCOPE: "skyline2006/llama-7b",
+        },
+        "LLaMA-13B": {
+            DownloadSource.DEFAULT: "huggyllama/llama-13b",
+            DownloadSource.MODELSCOPE: "skyline2006/llama-13b",
+        },
+        "LLaMA-30B": {
+            DownloadSource.DEFAULT: "huggyllama/llama-30b",
+            DownloadSource.MODELSCOPE: "skyline2006/llama-30b",
+        },
+        "LLaMA-65B": {
+            DownloadSource.DEFAULT: "huggyllama/llama-65b",
+            DownloadSource.MODELSCOPE: "skyline2006/llama-65b",
+        },
+    }
+)
+
+
+register_model_group(
+    models={
+        "LLaMA2-7B": {
+            DownloadSource.DEFAULT: "meta-llama/Llama-2-7b-hf",
+            DownloadSource.MODELSCOPE: "modelscope/Llama-2-7b-ms",
+        },
+        "LLaMA2-13B": {
+            DownloadSource.DEFAULT: "meta-llama/Llama-2-13b-hf",
+            DownloadSource.MODELSCOPE: "modelscope/Llama-2-13b-ms",
+        },
+        "LLaMA2-70B": {
+            DownloadSource.DEFAULT: "meta-llama/Llama-2-70b-hf",
+            DownloadSource.MODELSCOPE: "modelscope/Llama-2-70b-ms",
+        },
+        "LLaMA2-7B-Chat": {
+            DownloadSource.DEFAULT: "meta-llama/Llama-2-7b-chat-hf",
+            DownloadSource.MODELSCOPE: "modelscope/Llama-2-7b-chat-ms",
+        },
+        "LLaMA2-13B-Chat": {
+            DownloadSource.DEFAULT: "meta-llama/Llama-2-13b-chat-hf",
+            DownloadSource.MODELSCOPE: "modelscope/Llama-2-13b-chat-ms",
+        },
+        "LLaMA2-70B-Chat": {
+            DownloadSource.DEFAULT: "meta-llama/Llama-2-70b-chat-hf",
+            DownloadSource.MODELSCOPE: "modelscope/Llama-2-70b-chat-ms",
+        },
+    },
+    template="llama2",
+)
+
+
+register_model_group(
+    models={
+        "LLaMA3-8B": {
+            DownloadSource.DEFAULT: "meta-llama/Meta-Llama-3-8B",
+            DownloadSource.MODELSCOPE: "LLM-Research/Meta-Llama-3-8B",
+        },
+        "LLaMA3-70B": {
+            DownloadSource.DEFAULT: "meta-llama/Meta-Llama-3-70B",
+            DownloadSource.MODELSCOPE: "LLM-Research/Meta-Llama-3-70B",
+        },
+        "LLaMA3-8B-Chat": {
+            DownloadSource.DEFAULT: "meta-llama/Meta-Llama-3-8B-Instruct",
+            DownloadSource.MODELSCOPE: "LLM-Research/Meta-Llama-3-8B-Instruct",
+        },
+        "LLaMA3-70B-Chat": {
+            DownloadSource.DEFAULT: "meta-llama/Meta-Llama-3-70B-Instruct",
+            DownloadSource.MODELSCOPE: "LLM-Research/Meta-Llama-3-70B-Instruct",
+        },
+        "LLaMA3-8B-Chinese-Chat": {
+            DownloadSource.DEFAULT: "shenzhi-wang/Llama3-8B-Chinese-Chat",
+            DownloadSource.MODELSCOPE: "LLM-Research/Llama3-8B-Chinese-Chat",
+        },
+        "LLaMA3-70B-Chinese-Chat": {
+            DownloadSource.DEFAULT: "shenzhi-wang/Llama3-70B-Chinese-Chat",
+        },
+    },
+    template="llama3",
+)
+
+
+register_model_group(
+    models={
+        "LLaVA1.5-7B-Chat": {
+            DownloadSource.DEFAULT: "llava-hf/llava-1.5-7b-hf",
+        },
+        "LLaVA1.5-13B-Chat": {
+            DownloadSource.DEFAULT: "llava-hf/llava-1.5-13b-hf",
+        },
+    },
+    template="vicuna",
+    vision=True,
+)
+
+
+register_model_group(
+    models={
+        "MiniCPM-2B-SFT-Chat": {
+            DownloadSource.DEFAULT: "openbmb/MiniCPM-2B-sft-bf16",
+            DownloadSource.MODELSCOPE: "OpenBMB/miniCPM-bf16",
+        },
+        "MiniCPM-2B-DPO-Chat": {
+            DownloadSource.DEFAULT: "openbmb/MiniCPM-2B-dpo-bf16",
+            DownloadSource.MODELSCOPE: "OpenBMB/MiniCPM-2B-dpo-bf16",
+        },
+    },
+    template="cpm",
+)
+
+
+register_model_group(
+    models={
+        "Mistral-7B-v0.1": {
+            DownloadSource.DEFAULT: "mistralai/Mistral-7B-v0.1",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/Mistral-7B-v0.1",
+        },
+        "Mistral-7B-v0.1-Chat": {
+            DownloadSource.DEFAULT: "mistralai/Mistral-7B-Instruct-v0.1",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/Mistral-7B-Instruct-v0.1",
+        },
+        "Mistral-7B-v0.2": {
+            DownloadSource.DEFAULT: "alpindale/Mistral-7B-v0.2-hf",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/Mistral-7B-v0.2-hf",
+        },
+        "Mistral-7B-v0.2-Chat": {
+            DownloadSource.DEFAULT: "mistralai/Mistral-7B-Instruct-v0.2",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/Mistral-7B-Instruct-v0.2",
+        },
+        "Mistral-7B-v0.3": {
+            DownloadSource.DEFAULT: "mistralai/Mistral-7B-v0.3",
+        },
+        "Mistral-7B-v0.3-Chat": {
+            DownloadSource.DEFAULT: "mistralai/Mistral-7B-Instruct-v0.3",
+        },
+    },
+    template="mistral",
+)
+
+
+register_model_group(
+    models={
+        "Mixtral-8x7B-v0.1": {
+            DownloadSource.DEFAULT: "mistralai/Mixtral-8x7B-v0.1",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/Mixtral-8x7B-v0.1",
+        },
+        "Mixtral-8x7B-v0.1-Chat": {
+            DownloadSource.DEFAULT: "mistralai/Mixtral-8x7B-Instruct-v0.1",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/Mixtral-8x7B-Instruct-v0.1",
+        },
+        "Mixtral-8x22B-v0.1": {
+            DownloadSource.DEFAULT: "mistralai/Mixtral-8x22B-v0.1",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/Mixtral-8x22B-v0.1",
+        },
+        "Mixtral-8x22B-v0.1-Chat": {
+            DownloadSource.DEFAULT: "mistralai/Mixtral-8x22B-Instruct-v0.1",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/Mixtral-8x22B-Instruct-v0.1",
+        },
+    },
+    template="mistral",
+)
+
+
+register_model_group(
+    models={
+        "OLMo-1B": {
+            DownloadSource.DEFAULT: "allenai/OLMo-1B-hf",
+        },
+        "OLMo-7B": {
+            DownloadSource.DEFAULT: "allenai/OLMo-7B-hf",
+        },
+        "OLMo-7B-Chat": {
+            DownloadSource.DEFAULT: "ssec-uw/OLMo-7B-Instruct-hf",
+        },
+        "OLMo-1.7-7B": {
+            DownloadSource.DEFAULT: "allenai/OLMo-1.7-7B-hf",
+        },
+    },
+)
+
+
+register_model_group(
+    models={
+        "OpenChat3.5-7B-Chat": {
+            DownloadSource.DEFAULT: "openchat/openchat-3.5-0106",
+            DownloadSource.MODELSCOPE: "xcwzxcwz/openchat-3.5-0106",
+        }
+    },
+    template="openchat",
+)
+
+
+register_model_group(
+    models={
+        "OpenChat3.6-8B-Chat": {
+            DownloadSource.DEFAULT: "openchat/openchat-3.6-8b-20240522",
+        }
+    },
+    template="openchat-3.6",
+)
+
+
+register_model_group(
+    models={
+        "Orion-14B-Base": {
+            DownloadSource.DEFAULT: "OrionStarAI/Orion-14B-Base",
+            DownloadSource.MODELSCOPE: "OrionStarAI/Orion-14B-Base",
+        },
+        "Orion-14B-Chat": {
+            DownloadSource.DEFAULT: "OrionStarAI/Orion-14B-Chat",
+            DownloadSource.MODELSCOPE: "OrionStarAI/Orion-14B-Chat",
+        },
+        "Orion-14B-Long-Chat": {
+            DownloadSource.DEFAULT: "OrionStarAI/Orion-14B-LongChat",
+            DownloadSource.MODELSCOPE: "OrionStarAI/Orion-14B-LongChat",
+        },
+        "Orion-14B-RAG-Chat": {
+            DownloadSource.DEFAULT: "OrionStarAI/Orion-14B-Chat-RAG",
+            DownloadSource.MODELSCOPE: "OrionStarAI/Orion-14B-Chat-RAG",
+        },
+        "Orion-14B-Plugin-Chat": {
+            DownloadSource.DEFAULT: "OrionStarAI/Orion-14B-Chat-Plugin",
+            DownloadSource.MODELSCOPE: "OrionStarAI/Orion-14B-Chat-Plugin",
+        },
+    },
+    template="orion",
+)
+
+
+register_model_group(
+    models={
+        "PaliGemma-3B-pt-224": {
+            DownloadSource.DEFAULT: "google/paligemma-3b-pt-224",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/paligemma-3b-pt-224",
+        },
+        "PaliGemma-3B-pt-448": {
+            DownloadSource.DEFAULT: "google/paligemma-3b-pt-448",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/paligemma-3b-pt-448",
+        },
+        "PaliGemma-3B-pt-896": {
+            DownloadSource.DEFAULT: "google/paligemma-3b-pt-896",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/paligemma-3b-pt-896",
+        },
+        "PaliGemma-3B-mix-224": {
+            DownloadSource.DEFAULT: "google/paligemma-3b-mix-224",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/paligemma-3b-mix-224",
+        },
+        "PaliGemma-3B-mix-448": {
+            DownloadSource.DEFAULT: "google/paligemma-3b-mix-448",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/paligemma-3b-mix-448",
+        },
+    },
+    vision=True,
+)
+
+
+register_model_group(
+    models={
+        "Phi-1.5-1.3B": {
+            DownloadSource.DEFAULT: "microsoft/phi-1_5",
+            DownloadSource.MODELSCOPE: "allspace/PHI_1-5",
+        },
+        "Phi-2-2.7B": {
+            DownloadSource.DEFAULT: "microsoft/phi-2",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/phi-2",
+        },
+    }
+)
+
+
+register_model_group(
+    models={
+        "Phi3-4B-4k-Chat": {
+            DownloadSource.DEFAULT: "microsoft/Phi-3-mini-4k-instruct",
+            DownloadSource.MODELSCOPE: "LLM-Research/Phi-3-mini-4k-instruct",
+        },
+        "Phi3-4B-128k-Chat": {
+            DownloadSource.DEFAULT: "microsoft/Phi-3-mini-128k-instruct",
+            DownloadSource.MODELSCOPE: "LLM-Research/Phi-3-mini-128k-instruct",
+        },
+        "Phi3-7B-8k-Chat": {
+            DownloadSource.DEFAULT: "microsoft/Phi-3-small-8k-instruct",
+            DownloadSource.MODELSCOPE: "LLM-Research/Phi-3-small-8k-instruct",
+        },
+        "Phi3-7B-128k-Chat": {
+            DownloadSource.DEFAULT: "microsoft/Phi-3-small-128k-instruct",
+            DownloadSource.MODELSCOPE: "LLM-Research/Phi-3-small-128k-instruct",
+        },
+        "Phi3-14B-8k-Chat": {
+            DownloadSource.DEFAULT: "microsoft/Phi-3-medium-4k-instruct",
+            DownloadSource.MODELSCOPE: "LLM-Research/Phi-3-medium-4k-instruct",
+        },
+        "Phi3-14B-128k-Chat": {
+            DownloadSource.DEFAULT: "microsoft/Phi-3-medium-128k-instruct",
+            DownloadSource.MODELSCOPE: "LLM-Research/Phi-3-medium-128k-instruct",
+        },
+    },
+    template="phi",
+)
+
+
+register_model_group(
+    models={
+        "Qwen-1.8B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-1_8B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-1_8B",
+        },
+        "Qwen-7B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-7B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-7B",
+        },
+        "Qwen-14B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-14B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-14B",
+        },
+        "Qwen-72B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-72B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-72B",
+        },
+        "Qwen-1.8B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-1_8B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-1_8B-Chat",
+        },
+        "Qwen-7B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-7B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-7B-Chat",
+        },
+        "Qwen-14B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-14B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-14B-Chat",
+        },
+        "Qwen-72B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-72B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-72B-Chat",
+        },
+        "Qwen-1.8B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-1_8B-Chat-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-1_8B-Chat-Int8",
+        },
+        "Qwen-1.8B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-1_8B-Chat-Int4",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-1_8B-Chat-Int4",
+        },
+        "Qwen-7B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-7B-Chat-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-7B-Chat-Int8",
+        },
+        "Qwen-7B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-7B-Chat-Int4",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-7B-Chat-Int4",
+        },
+        "Qwen-14B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-14B-Chat-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-14B-Chat-Int8",
+        },
+        "Qwen-14B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-14B-Chat-Int4",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-14B-Chat-Int4",
+        },
+        "Qwen-72B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-72B-Chat-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-72B-Chat-Int8",
+        },
+        "Qwen-72B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-72B-Chat-Int4",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-72B-Chat-Int4",
+        },
+    },
+    template="qwen",
+)
+
+
+register_model_group(
+    models={
+        "Qwen1.5-0.5B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-0.5B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-0.5B",
+        },
+        "Qwen1.5-1.8B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-1.8B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-1.8B",
+        },
+        "Qwen1.5-4B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-4B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-4B",
+        },
+        "Qwen1.5-7B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-7B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-7B",
+        },
+        "Qwen1.5-14B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-14B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-14B",
+        },
+        "Qwen1.5-32B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-32B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-32B",
+        },
+        "Qwen1.5-72B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-72B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-72B",
+        },
+        "Qwen1.5-110B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-110B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-110B",
+        },
+        "Qwen1.5-MoE-A2.7B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-MoE-A2.7B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-MoE-A2.7B",
+        },
+        "Qwen1.5-Code-7B": {
+            DownloadSource.DEFAULT: "Qwen/CodeQwen1.5-7B",
+            DownloadSource.MODELSCOPE: "qwen/CodeQwen1.5-7B",
+        },
+        "Qwen1.5-0.5B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-0.5B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-0.5B-Chat",
+        },
+        "Qwen1.5-1.8B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-1.8B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-1.8B-Chat",
+        },
+        "Qwen1.5-4B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-4B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-4B-Chat",
+        },
+        "Qwen1.5-7B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-7B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-7B-Chat",
+        },
+        "Qwen1.5-14B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-14B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-14B-Chat",
+        },
+        "Qwen1.5-32B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-32B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-32B-Chat",
+        },
+        "Qwen1.5-72B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-72B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-72B-Chat",
+        },
+        "Qwen1.5-110B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-110B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-110B-Chat",
+        },
+        "Qwen1.5-MoE-A2.7B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-MoE-A2.7B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-MoE-A2.7B-Chat",
+        },
+        "Qwen1.5-Code-7B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/CodeQwen1.5-7B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/CodeQwen1.5-7B-Chat",
+        },
+        "Qwen1.5-0.5B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-0.5B-Chat-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-0.5B-Chat-GPTQ-Int8",
+        },
+        "Qwen1.5-0.5B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-0.5B-Chat-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-0.5B-Chat-AWQ",
+        },
+        "Qwen1.5-1.8B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-1.8B-Chat-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-1.8B-Chat-GPTQ-Int8",
+        },
+        "Qwen1.5-1.8B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-1.8B-Chat-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-1.8B-Chat-AWQ",
+        },
+        "Qwen1.5-4B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-4B-Chat-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-4B-Chat-GPTQ-Int8",
+        },
+        "Qwen1.5-4B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-4B-Chat-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-4B-Chat-AWQ",
+        },
+        "Qwen1.5-7B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-7B-Chat-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-7B-Chat-GPTQ-Int8",
+        },
+        "Qwen1.5-7B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-7B-Chat-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-7B-Chat-AWQ",
+        },
+        "Qwen1.5-14B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-14B-Chat-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-14B-Chat-GPTQ-Int8",
+        },
+        "Qwen1.5-14B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-14B-Chat-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-14B-Chat-AWQ",
+        },
+        "Qwen1.5-32B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-32B-Chat-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-32B-Chat-AWQ",
+        },
+        "Qwen1.5-72B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-72B-Chat-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-72B-Chat-GPTQ-Int8",
+        },
+        "Qwen1.5-72B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-72B-Chat-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-72B-Chat-AWQ",
+        },
+        "Qwen1.5-110B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-110B-Chat-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-110B-Chat-AWQ",
+        },
+        "Qwen1.5-MoE-A2.7B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen1.5-MoE-A2.7B-Chat-GPTQ-Int4",
+            DownloadSource.MODELSCOPE: "qwen/Qwen1.5-MoE-A2.7B-Chat-GPTQ-Int4",
+        },
+        "Qwen1.5-Code-7B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/CodeQwen1.5-7B-Chat-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/CodeQwen1.5-7B-Chat-AWQ",
+        },
+    },
+    template="qwen",
+)
+
+
+register_model_group(
+    models={
+        "Qwen2-0.5B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-0.5B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-0.5B",
+        },
+        "Qwen2-1.5B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-1.5B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-1.5B",
+        },
+        "Qwen2-7B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-7B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-7B",
+        },
+        "Qwen2-72B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-72B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-72B",
+        },
+        "Qwen2-MoE-57B": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-57B-A14B",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-57B-A14B",
+        },
+        "Qwen2-0.5B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-0.5B-Instruct",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-0.5B-Instruct",
+        },
+        "Qwen2-1.5B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-1.5B-Instruct",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-1.5B-Instruct",
+        },
+        "Qwen2-7B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-7B-Instruct",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-7B-Instruct",
+        },
+        "Qwen2-72B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-72B-Instruct",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-72B-Instruct",
+        },
+        "Qwen2-MoE-57B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-57B-A14B-Instruct",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-57B-A14B-Instruct",
+        },
+        "Qwen2-0.5B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-0.5B-Instruct-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-0.5B-Instruct-GPTQ-Int8",
+        },
+        "Qwen2-0.5B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-0.5B-Instruct-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-0.5B-Instruct-AWQ",
+        },
+        "Qwen2-1.5B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-1.5B-Instruct-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-1.5B-Instruct-GPTQ-Int8",
+        },
+        "Qwen2-1.5B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-1.5B-Instruct-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-1.5B-Instruct-AWQ",
+        },
+        "Qwen2-7B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-7B-Instruct-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-7B-Instruct-GPTQ-Int8",
+        },
+        "Qwen2-7B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-7B-Instruct-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-7B-Instruct-AWQ",
+        },
+        "Qwen2-72B-int8-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-72B-Instruct-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-72B-Instruct-GPTQ-Int8",
+        },
+        "Qwen2-72B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-72B-Instruct-AWQ",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-72B-Instruct-AWQ",
+        },
+        "Qwen2-MoE-57B-int4-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2-57B-A14B-Instruct-GPTQ-Int4",
+            DownloadSource.MODELSCOPE: "qwen/Qwen2-57B-A14B-Instruct-GPTQ-Int4",
+        },
+    },
+    template="qwen",
+)
+
+
+register_model_group(
+    models={
+        "SOLAR-10.7B": {
+            DownloadSource.DEFAULT: "upstage/SOLAR-10.7B-v1.0",
+        },
+        "SOLAR-10.7B-Chat": {
+            DownloadSource.DEFAULT: "upstage/SOLAR-10.7B-Instruct-v1.0",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/SOLAR-10.7B-Instruct-v1.0",
+        },
+    },
+    template="solar",
+)
+
+
+register_model_group(
+    models={
+        "Skywork-13B-Base": {
+            DownloadSource.DEFAULT: "Skywork/Skywork-13B-base",
+            DownloadSource.MODELSCOPE: "skywork/Skywork-13B-base",
+        }
+    }
+)
+
+
+register_model_group(
+    models={
+        "StarCoder2-3B": {
+            DownloadSource.DEFAULT: "bigcode/starcoder2-3b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/starcoder2-3b",
+        },
+        "StarCoder2-7B": {
+            DownloadSource.DEFAULT: "bigcode/starcoder2-7b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/starcoder2-7b",
+        },
+        "StarCoder2-15B": {
+            DownloadSource.DEFAULT: "bigcode/starcoder2-15b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/starcoder2-15b",
+        },
+    }
+)
+
+
+register_model_group(
+    models={
+        "TeleChat-1B-Chat": {
+            DownloadSource.DEFAULT: "Tele-AI/TeleChat-1B",
+            DownloadSource.MODELSCOPE: "TeleAI/TeleChat-1B",
+        },
+        "TeleChat-7B-Chat": {
+            DownloadSource.DEFAULT: "Tele-AI/telechat-7B",
+            DownloadSource.MODELSCOPE: "TeleAI/telechat-7B",
+        },
+        "TeleChat-12B-Chat": {
+            DownloadSource.DEFAULT: "Tele-AI/TeleChat-12B",
+            DownloadSource.MODELSCOPE: "TeleAI/TeleChat-12B",
+        },
+        "TeleChat-12B-v2-Chat": {
+            DownloadSource.DEFAULT: "Tele-AI/TeleChat-12B-v2",
+            DownloadSource.MODELSCOPE: "TeleAI/TeleChat-12B-v2",
+        },
+    },
+    template="telechat",
+)
+
+
+register_model_group(
+    models={
+        "Vicuna1.5-7B-Chat": {
+            DownloadSource.DEFAULT: "lmsys/vicuna-7b-v1.5",
+            DownloadSource.MODELSCOPE: "Xorbits/vicuna-7b-v1.5",
+        },
+        "Vicuna1.5-13B-Chat": {
+            DownloadSource.DEFAULT: "lmsys/vicuna-13b-v1.5",
+            DownloadSource.MODELSCOPE: "Xorbits/vicuna-13b-v1.5",
+        },
+    },
+    template="vicuna",
+)
+
+
+register_model_group(
+    models={
+        "XuanYuan-6B": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan-6B",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan-6B",
+        },
+        "XuanYuan-70B": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan-70B",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan-70B",
+        },
+        "XuanYuan-2-70B": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan2-70B",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan2-70B",
+        },
+        "XuanYuan-6B-Chat": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan-6B-Chat",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan-6B-Chat",
+        },
+        "XuanYuan-70B-Chat": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan-70B-Chat",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan-70B-Chat",
+        },
+        "XuanYuan-2-70B-Chat": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan2-70B-Chat",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan2-70B-Chat",
+        },
+        "XuanYuan-6B-int8-Chat": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan-6B-Chat-8bit",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan-6B-Chat-8bit",
+        },
+        "XuanYuan-6B-int4-Chat": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan-6B-Chat-4bit",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan-6B-Chat-4bit",
+        },
+        "XuanYuan-70B-int8-Chat": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan-70B-Chat-8bit",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan-70B-Chat-8bit",
+        },
+        "XuanYuan-70B-int4-Chat": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan-70B-Chat-4bit",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan-70B-Chat-4bit",
+        },
+        "XuanYuan-2-70B-int8-Chat": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan2-70B-Chat-8bit",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan2-70B-Chat-8bit",
+        },
+        "XuanYuan-2-70B-int4-Chat": {
+            DownloadSource.DEFAULT: "Duxiaoman-DI/XuanYuan2-70B-Chat-4bit",
+            DownloadSource.MODELSCOPE: "Duxiaoman-DI/XuanYuan2-70B-Chat-4bit",
+        },
+    },
+    template="xuanyuan",
+)
+
+
+register_model_group(
+    models={
+        "XVERSE-7B": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-7B",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-7B",
+        },
+        "XVERSE-13B": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-13B",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-13B",
+        },
+        "XVERSE-65B": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-65B",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-65B",
+        },
+        "XVERSE-65B-2": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-65B-2",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-65B-2",
+        },
+        "XVERSE-7B-Chat": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-7B-Chat",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-7B-Chat",
+        },
+        "XVERSE-13B-Chat": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-13B-Chat",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-13B-Chat",
+        },
+        "XVERSE-65B-Chat": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-65B-Chat",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-65B-Chat",
+        },
+        "XVERSE-MoE-A4.2B": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-MoE-A4.2B",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-MoE-A4.2B",
+        },
+        "XVERSE-7B-int8-Chat": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-7B-Chat-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-7B-Chat-GPTQ-Int8",
+        },
+        "XVERSE-7B-int4-Chat": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-7B-Chat-GPTQ-Int4",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-7B-Chat-GPTQ-Int4",
+        },
+        "XVERSE-13B-int8-Chat": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-13B-Chat-GPTQ-Int8",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-13B-Chat-GPTQ-Int8",
+        },
+        "XVERSE-13B-int4-Chat": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-13B-Chat-GPTQ-Int4",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-13B-Chat-GPTQ-Int4",
+        },
+        "XVERSE-65B-int4-Chat": {
+            DownloadSource.DEFAULT: "xverse/XVERSE-65B-Chat-GPTQ-Int4",
+            DownloadSource.MODELSCOPE: "xverse/XVERSE-65B-Chat-GPTQ-Int4",
+        },
+    },
+    template="xverse",
+)
+
+
+register_model_group(
+    models={
+        "Yayi-7B": {
+            DownloadSource.DEFAULT: "wenge-research/yayi-7b-llama2",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/yayi-7b-llama2",
+        },
+        "Yayi-13B": {
+            DownloadSource.DEFAULT: "wenge-research/yayi-13b-llama2",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/yayi-13b-llama2",
+        },
+    },
+    template="yayi",
+)
+
+
+register_model_group(
+    models={
+        "Yi-6B": {
+            DownloadSource.DEFAULT: "01-ai/Yi-6B",
+            DownloadSource.MODELSCOPE: "01ai/Yi-6B",
+        },
+        "Yi-9B": {
+            DownloadSource.DEFAULT: "01-ai/Yi-9B",
+            DownloadSource.MODELSCOPE: "01ai/Yi-9B",
+        },
+        "Yi-34B": {
+            DownloadSource.DEFAULT: "01-ai/Yi-34B",
+            DownloadSource.MODELSCOPE: "01ai/Yi-34B",
+        },
+        "Yi-6B-Chat": {
+            DownloadSource.DEFAULT: "01-ai/Yi-6B-Chat",
+            DownloadSource.MODELSCOPE: "01ai/Yi-6B-Chat",
+        },
+        "Yi-34B-Chat": {
+            DownloadSource.DEFAULT: "01-ai/Yi-34B-Chat",
+            DownloadSource.MODELSCOPE: "01ai/Yi-34B-Chat",
+        },
+        "Yi-6B-int8-Chat": {
+            DownloadSource.DEFAULT: "01-ai/Yi-6B-Chat-8bits",
+            DownloadSource.MODELSCOPE: "01ai/Yi-6B-Chat-8bits",
+        },
+        "Yi-6B-int4-Chat": {
+            DownloadSource.DEFAULT: "01-ai/Yi-6B-Chat-4bits",
+            DownloadSource.MODELSCOPE: "01ai/Yi-6B-Chat-4bits",
+        },
+        "Yi-34B-int8-Chat": {
+            DownloadSource.DEFAULT: "01-ai/Yi-34B-Chat-8bits",
+            DownloadSource.MODELSCOPE: "01ai/Yi-34B-Chat-8bits",
+        },
+        "Yi-34B-int4-Chat": {
+            DownloadSource.DEFAULT: "01-ai/Yi-34B-Chat-4bits",
+            DownloadSource.MODELSCOPE: "01ai/Yi-34B-Chat-4bits",
+        },
+        "Yi-1.5-6B": {
+            DownloadSource.DEFAULT: "01-ai/Yi-1.5-6B",
+            DownloadSource.MODELSCOPE: "01ai/Yi-1.5-6B",
+        },
+        "Yi-1.5-9B": {
+            DownloadSource.DEFAULT: "01-ai/Yi-1.5-9B",
+            DownloadSource.MODELSCOPE: "01ai/Yi-1.5-9B",
+        },
+        "Yi-1.5-34B": {
+            DownloadSource.DEFAULT: "01-ai/Yi-1.5-34B",
+            DownloadSource.MODELSCOPE: "01ai/Yi-1.5-34B",
+        },
+        "Yi-1.5-6B-Chat": {
+            DownloadSource.DEFAULT: "01-ai/Yi-1.5-6B-Chat",
+            DownloadSource.MODELSCOPE: "01ai/Yi-1.5-6B-Chat",
+        },
+        "Yi-1.5-9B-Chat": {
+            DownloadSource.DEFAULT: "01-ai/Yi-1.5-9B-Chat",
+            DownloadSource.MODELSCOPE: "01ai/Yi-1.5-9B-Chat",
+        },
+        "Yi-1.5-34B-Chat": {
+            DownloadSource.DEFAULT: "01-ai/Yi-1.5-34B-Chat",
+            DownloadSource.MODELSCOPE: "01ai/Yi-1.5-34B-Chat",
+        },
+    },
+    template="yi",
+)
+
+
+register_model_group(
+    models={
+        "YiVL-6B-Chat": {
+            DownloadSource.DEFAULT: "BUAADreamer/Yi-VL-6B-hf",
+        },
+        "YiVL-34B-Chat": {
+            DownloadSource.DEFAULT: "BUAADreamer/Yi-VL-34B-hf",
+        },
+    },
+    template="yi_vl",
+    vision=True,
+)
+
+
+register_model_group(
+    models={
+        "Yuan2-2B-Chat": {
+            DownloadSource.DEFAULT: "IEITYuan/Yuan2-2B-hf",
+            DownloadSource.MODELSCOPE: "YuanLLM/Yuan2.0-2B-hf",
+        },
+        "Yuan2-51B-Chat": {
+            DownloadSource.DEFAULT: "IEITYuan/Yuan2-51B-hf",
+            DownloadSource.MODELSCOPE: "YuanLLM/Yuan2.0-51B-hf",
+        },
+        "Yuan2-102B-Chat": {
+            DownloadSource.DEFAULT: "IEITYuan/Yuan2-102B-hf",
+            DownloadSource.MODELSCOPE: "YuanLLM/Yuan2.0-102B-hf",
+        },
+    },
+    template="yuan",
+)
+
+
+register_model_group(
+    models={
+        "Zephyr-7B-Alpha-Chat": {
+            DownloadSource.DEFAULT: "HuggingFaceH4/zephyr-7b-alpha",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/zephyr-7b-alpha",
+        },
+        "Zephyr-7B-Beta-Chat": {
+            DownloadSource.DEFAULT: "HuggingFaceH4/zephyr-7b-beta",
+            DownloadSource.MODELSCOPE: "modelscope/zephyr-7b-beta",
+        },
+        "Zephyr-141B-ORPO-Chat": {
+            DownloadSource.DEFAULT: "HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1",
+        },
+    },
+    template="zephyr",
+)

llama-factory/src/llamafactory/extras/env.py

@@ -0,0 +1,75 @@
+# Copyright 2024 HuggingFace Inc. and the LlamaFactory team.
+#
+# This code is inspired by the HuggingFace's transformers library.
+# https://github.com/huggingface/transformers/blob/v4.40.0/src/transformers/commands/env.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 platform
+
+import accelerate
+import datasets
+import peft
+import torch
+import transformers
+import trl
+from transformers.utils import is_torch_cuda_available, is_torch_npu_available
+
+
+VERSION = "0.8.4.dev0"
+
+
+def print_env() -> None:
+    info = {
+        "`llamafactory` version": VERSION,
+        "Platform": platform.platform(),
+        "Python version": platform.python_version(),
+        "PyTorch version": torch.__version__,
+        "Transformers version": transformers.__version__,
+        "Datasets version": datasets.__version__,
+        "Accelerate version": accelerate.__version__,
+        "PEFT version": peft.__version__,
+        "TRL version": trl.__version__,
+    }
+
+    if is_torch_cuda_available():
+        info["PyTorch version"] += " (GPU)"
+        info["GPU type"] = torch.cuda.get_device_name()
+
+    if is_torch_npu_available():
+        info["PyTorch version"] += " (NPU)"
+        info["NPU type"] = torch.npu.get_device_name()
+        info["CANN version"] = torch.version.cann
+
+    try:
+        import deepspeed  # type: ignore
+
+        info["DeepSpeed version"] = deepspeed.__version__
+    except Exception:
+        pass
+
+    try:
+        import bitsandbytes
+
+        info["Bitsandbytes version"] = bitsandbytes.__version__
+    except Exception:
+        pass
+
+    try:
+        import vllm
+
+        info["vLLM version"] = vllm.__version__
+    except Exception:
+        pass
+
+    print("\n" + "\n".join(["- {}: {}".format(key, value) for key, value in info.items()]) + "\n")

llama-factory/src/llamafactory/extras/logging.py

@@ -0,0 +1,82 @@
+# 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 logging
+import os
+import sys
+from concurrent.futures import ThreadPoolExecutor
+
+from .constants import RUNNING_LOG
+
+
+class LoggerHandler(logging.Handler):
+    r"""
+    Logger handler used in Web UI.
+    """
+
+    def __init__(self, output_dir: str) -> None:
+        super().__init__()
+        formatter = logging.Formatter(
+            fmt="%(asctime)s - %(levelname)s - %(name)s - %(message)s", datefmt="%m/%d/%Y %H:%M:%S"
+        )
+        self.setLevel(logging.INFO)
+        self.setFormatter(formatter)
+
+        os.makedirs(output_dir, exist_ok=True)
+        self.running_log = os.path.join(output_dir, RUNNING_LOG)
+        if os.path.exists(self.running_log):
+            os.remove(self.running_log)
+
+        self.thread_pool = ThreadPoolExecutor(max_workers=1)
+
+    def _write_log(self, log_entry: str) -> None:
+        with open(self.running_log, "a", encoding="utf-8") as f:
+            f.write(log_entry + "\n\n")
+
+    def emit(self, record) -> None:
+        if record.name == "httpx":
+            return
+
+        log_entry = self.format(record)
+        self.thread_pool.submit(self._write_log, log_entry)
+
+    def close(self) -> None:
+        self.thread_pool.shutdown(wait=True)
+        return super().close()
+
+
+def get_logger(name: str) -> logging.Logger:
+    r"""
+    Gets a standard logger with a stream hander to stdout.
+    """
+    formatter = logging.Formatter(
+        fmt="%(asctime)s - %(levelname)s - %(name)s - %(message)s", datefmt="%m/%d/%Y %H:%M:%S"
+    )
+    handler = logging.StreamHandler(sys.stdout)
+    handler.setFormatter(formatter)
+
+    logger = logging.getLogger(name)
+    logger.setLevel(logging.INFO)
+    logger.addHandler(handler)
+
+    return logger
+
+
+def reset_logging() -> None:
+    r"""
+    Removes basic config of root logger. (unused in script)
+    """
+    root = logging.getLogger()
+    list(map(root.removeHandler, root.handlers))
+    list(map(root.removeFilter, root.filters))

llama-factory/src/llamafactory/extras/misc.py

@@ -0,0 +1,228 @@
+# Copyright 2024 HuggingFace Inc. and the LlamaFactory team.
+#
+# This code is inspired by the HuggingFace's PEFT library.
+# https://github.com/huggingface/peft/blob/v0.10.0/src/peft/peft_model.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 gc
+import os
+from typing import TYPE_CHECKING, Tuple, Union
+
+import torch
+import transformers.dynamic_module_utils
+from transformers import InfNanRemoveLogitsProcessor, LogitsProcessorList
+from transformers.dynamic_module_utils import get_relative_imports
+from transformers.utils import (
+    is_torch_bf16_gpu_available,
+    is_torch_cuda_available,
+    is_torch_mps_available,
+    is_torch_npu_available,
+    is_torch_xpu_available,
+)
+from transformers.utils.versions import require_version
+
+from .logging import get_logger
+
+
+_is_fp16_available = is_torch_npu_available() or is_torch_cuda_available()
+try:
+    _is_bf16_available = is_torch_bf16_gpu_available()
+except Exception:
+    _is_bf16_available = False
+
+
+if TYPE_CHECKING:
+    from numpy.typing import NDArray
+
+    from ..hparams import ModelArguments
+
+
+logger = get_logger(__name__)
+
+
+class AverageMeter:
+    r"""
+    Computes and stores the average and current value.
+    """
+
+    def __init__(self):
+        self.reset()
+
+    def reset(self):
+        self.val = 0
+        self.avg = 0
+        self.sum = 0
+        self.count = 0
+
+    def update(self, val, n=1):
+        self.val = val
+        self.sum += val * n
+        self.count += n
+        self.avg = self.sum / self.count
+
+
+def check_dependencies() -> None:
+    r"""
+    Checks the version of the required packages.
+    """
+    if os.environ.get("DISABLE_VERSION_CHECK", "0").lower() in ["true", "1"]:
+        logger.warning("Version checking has been disabled, may lead to unexpected behaviors.")
+    else:
+        require_version("transformers>=4.41.2", "To fix: pip install transformers>=4.41.2")
+        require_version("datasets>=2.16.0", "To fix: pip install datasets>=2.16.0")
+        require_version("accelerate>=0.30.1", "To fix: pip install accelerate>=0.30.1")
+        require_version("peft>=0.11.1", "To fix: pip install peft>=0.11.1")
+        require_version("trl>=0.8.6", "To fix: pip install trl>=0.8.6")
+
+
+def count_parameters(model: "torch.nn.Module") -> Tuple[int, int]:
+    r"""
+    Returns the number of trainable parameters and number of all parameters in the model.
+    """
+    trainable_params, all_param = 0, 0
+    for param in model.parameters():
+        num_params = param.numel()
+        # if using DS Zero 3 and the weights are initialized empty
+        if num_params == 0 and hasattr(param, "ds_numel"):
+            num_params = param.ds_numel
+
+        # Due to the design of 4bit linear layers from bitsandbytes, multiply the number of parameters by itemsize
+        if param.__class__.__name__ == "Params4bit":
+            if hasattr(param, "quant_storage") and hasattr(param.quant_storage, "itemsize"):
+                num_bytes = param.quant_storage.itemsize
+            elif hasattr(param, "element_size"):  # for older pytorch version
+                num_bytes = param.element_size()
+            else:
+                num_bytes = 1
+
+            num_params = num_params * 2 * num_bytes
+
+        all_param += num_params
+        if param.requires_grad:
+            trainable_params += num_params
+
+    return trainable_params, all_param
+
+
+def get_current_device() -> "torch.device":
+    r"""
+    Gets the current available device.
+    """
+    if is_torch_xpu_available():
+        device = "xpu:{}".format(os.environ.get("LOCAL_RANK", "0"))
+    elif is_torch_npu_available():
+        device = "npu:{}".format(os.environ.get("LOCAL_RANK", "0"))
+    elif is_torch_mps_available():
+        device = "mps:{}".format(os.environ.get("LOCAL_RANK", "0"))
+    elif is_torch_cuda_available():
+        device = "cuda:{}".format(os.environ.get("LOCAL_RANK", "0"))
+    else:
+        device = "cpu"
+
+    return torch.device(device)
+
+
+def get_device_count() -> int:
+    r"""
+    Gets the number of available GPU or NPU devices.
+    """
+    if is_torch_npu_available():
+        return torch.npu.device_count()
+    elif is_torch_cuda_available():
+        return torch.cuda.device_count()
+    else:
+        return 0
+
+
+def get_logits_processor() -> "LogitsProcessorList":
+    r"""
+    Gets logits processor that removes NaN and Inf logits.
+    """
+    logits_processor = LogitsProcessorList()
+    logits_processor.append(InfNanRemoveLogitsProcessor())
+    return logits_processor
+
+
+def has_tokenized_data(path: "os.PathLike") -> bool:
+    r"""
+    Checks if the path has a tokenized dataset.
+    """
+    return os.path.isdir(path) and len(os.listdir(path)) > 0
+
+
+def infer_optim_dtype(model_dtype: "torch.dtype") -> "torch.dtype":
+    r"""
+    Infers the optimal dtype according to the model_dtype and device compatibility.
+    """
+    if _is_bf16_available and model_dtype == torch.bfloat16:
+        return torch.bfloat16
+    elif _is_fp16_available:
+        return torch.float16
+    else:
+        return torch.float32
+
+
+def is_gpu_or_npu_available() -> bool:
+    r"""
+    Checks if the GPU or NPU is available.
+    """
+    return is_torch_npu_available() or is_torch_cuda_available()
+
+
+def numpify(inputs: Union["NDArray", "torch.Tensor"]) -> "NDArray":
+    if isinstance(inputs, torch.Tensor):
+        inputs = inputs.cpu()
+        if inputs.dtype == torch.bfloat16:  # numpy does not support bfloat16 until 1.21.4
+            inputs = inputs.to(torch.float32)
+
+        inputs = inputs.numpy()
+
+    return inputs
+
+
+def skip_check_imports() -> None:
+    if os.environ.get("FORCE_CHECK_IMPORTS", "0").lower() not in ["true", "1"]:
+        transformers.dynamic_module_utils.check_imports = get_relative_imports
+
+
+def torch_gc() -> None:
+    r"""
+    Collects GPU or NPU memory.
+    """
+    gc.collect()
+    if is_torch_xpu_available():
+        torch.xpu.empty_cache()
+    elif is_torch_npu_available():
+        torch.npu.empty_cache()
+    elif is_torch_mps_available():
+        torch.mps.empty_cache()
+    elif is_torch_cuda_available():
+        torch.cuda.empty_cache()
+
+
+def try_download_model_from_ms(model_args: "ModelArguments") -> str:
+    if not use_modelscope() or os.path.exists(model_args.model_name_or_path):
+        return model_args.model_name_or_path
+
+    try:
+        from modelscope import snapshot_download
+
+        revision = "master" if model_args.model_revision == "main" else model_args.model_revision
+        return snapshot_download(model_args.model_name_or_path, revision=revision, cache_dir=model_args.cache_dir)
+    except ImportError:
+        raise ImportError("Please install modelscope via `pip install modelscope -U`")
+
+
+def use_modelscope() -> bool:
+    return os.environ.get("USE_MODELSCOPE_HUB", "0").lower() in ["true", "1"]

llama-factory/src/llamafactory/extras/packages.py

@@ -0,0 +1,88 @@
+# Copyright 2024 HuggingFace Inc. and the LlamaFactory team.
+#
+# This code is inspired by the HuggingFace's transformers library.
+# https://github.com/huggingface/transformers/blob/v4.40.0/src/transformers/utils/import_utils.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 importlib.metadata
+import importlib.util
+from functools import lru_cache
+from typing import TYPE_CHECKING
+
+from packaging import version
+
+
+if TYPE_CHECKING:
+    from packaging.version import Version
+
+
+def _is_package_available(name: str) -> bool:
+    return importlib.util.find_spec(name) is not None
+
+
+def _get_package_version(name: str) -> "Version":
+    try:
+        return version.parse(importlib.metadata.version(name))
+    except Exception:
+        return version.parse("0.0.0")
+
+
+def is_fastapi_available():
+    return _is_package_available("fastapi")
+
+
+def is_galore_available():
+    return _is_package_available("galore_torch")
+
+
+def is_gradio_available():
+    return _is_package_available("gradio")
+
+
+def is_matplotlib_available():
+    return _is_package_available("matplotlib")
+
+
+def is_pillow_available():
+    return _is_package_available("PIL")
+
+
+def is_requests_available():
+    return _is_package_available("requests")
+
+
+def is_rouge_available():
+    return _is_package_available("rouge_chinese")
+
+
+def is_starlette_available():
+    return _is_package_available("sse_starlette")
+
+
+def is_uvicorn_available():
+    return _is_package_available("uvicorn")
+
+
+def is_vllm_available():
+    return _is_package_available("vllm")
+
+
+@lru_cache
+def is_vllm_version_greater_than_0_5():
+    return _get_package_version("vllm") >= version.parse("0.5.0")
+
+
+@lru_cache
+def is_vllm_version_greater_than_0_5_1():
+    return _get_package_version("vllm") >= version.parse("0.5.1")

llama-factory/src/llamafactory/extras/ploting.py

@@ -0,0 +1,101 @@
+# 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 math
+import os
+from typing import Any, Dict, List
+
+from transformers.trainer import TRAINER_STATE_NAME
+
+from .logging import get_logger
+from .packages import is_matplotlib_available
+
+
+if is_matplotlib_available():
+    import matplotlib.figure
+    import matplotlib.pyplot as plt
+
+
+logger = get_logger(__name__)
+
+
+def smooth(scalars: List[float]) -> List[float]:
+    r"""
+    EMA implementation according to TensorBoard.
+    """
+    if len(scalars) == 0:
+        return []
+
+    last = scalars[0]
+    smoothed = []
+    weight = 1.8 * (1 / (1 + math.exp(-0.05 * len(scalars))) - 0.5)  # a sigmoid function
+    for next_val in scalars:
+        smoothed_val = last * weight + (1 - weight) * next_val
+        smoothed.append(smoothed_val)
+        last = smoothed_val
+    return smoothed
+
+
+def gen_loss_plot(trainer_log: List[Dict[str, Any]]) -> "matplotlib.figure.Figure":
+    r"""
+    Plots loss curves in LlamaBoard.
+    """
+    plt.close("all")
+    plt.switch_backend("agg")
+    fig = plt.figure()
+    ax = fig.add_subplot(111)
+    steps, losses = [], []
+    for log in trainer_log:
+        if log.get("loss", None):
+            steps.append(log["current_steps"])
+            losses.append(log["loss"])
+
+    ax.plot(steps, losses, color="#1f77b4", alpha=0.4, label="original")
+    ax.plot(steps, smooth(losses), color="#1f77b4", label="smoothed")
+    ax.legend()
+    ax.set_xlabel("step")
+    ax.set_ylabel("loss")
+    return fig
+
+
+def plot_loss(save_dictionary: os.PathLike, keys: List[str] = ["loss"]) -> None:
+    r"""
+    Plots loss curves and saves the image.
+    """
+    plt.switch_backend("agg")
+    with open(os.path.join(save_dictionary, TRAINER_STATE_NAME), "r", encoding="utf-8") as f:
+        data = json.load(f)
+
+    for key in keys:
+        steps, metrics = [], []
+        for i in range(len(data["log_history"])):
+            if key in data["log_history"][i]:
+                steps.append(data["log_history"][i]["step"])
+                metrics.append(data["log_history"][i][key])
+
+        if len(metrics) == 0:
+            logger.warning(f"No metric {key} to plot.")
+            continue
+
+        plt.figure()
+        plt.plot(steps, metrics, color="#1f77b4", alpha=0.4, label="original")
+        plt.plot(steps, smooth(metrics), color="#1f77b4", label="smoothed")
+        plt.title("training {} of {}".format(key, save_dictionary))
+        plt.xlabel("step")
+        plt.ylabel(key)
+        plt.legend()
+        figure_path = os.path.join(save_dictionary, "training_{}.png".format(key.replace("/", "_")))
+        plt.savefig(figure_path, format="png", dpi=100)
+        print("Figure saved at:", figure_path)

llama-factory/src/llamafactory/hparams/__init__.py

@@ -0,0 +1,32 @@
+# 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 .data_args import DataArguments
+from .evaluation_args import EvaluationArguments
+from .finetuning_args import FinetuningArguments
+from .generating_args import GeneratingArguments
+from .model_args import ModelArguments
+from .parser import get_eval_args, get_infer_args, get_train_args
+
+
+__all__ = [
+    "DataArguments",
+    "EvaluationArguments",
+    "FinetuningArguments",
+    "GeneratingArguments",
+    "ModelArguments",
+    "get_eval_args",
+    "get_infer_args",
+    "get_train_args",
+]

llama-factory/src/llamafactory/hparams/data_args.py

@@ -0,0 +1,143 @@
+# Copyright 2024 HuggingFace Inc. and the LlamaFactory team.
+#
+# This code is inspired by the HuggingFace's transformers library.
+# https://github.com/huggingface/transformers/blob/v4.40.0/examples/pytorch/language-modeling/run_clm.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.
+
+from dataclasses import dataclass, field
+from typing import Literal, Optional
+
+
+@dataclass
+class DataArguments:
+    r"""
+    Arguments pertaining to what data we are going to input our model for training and evaluation.
+    """
+
+    template: Optional[str] = field(
+        default=None,
+        metadata={"help": "Which template to use for constructing prompts in training and inference."},
+    )
+    dataset: Optional[str] = field(
+        default=None,
+        metadata={"help": "The name of dataset(s) to use for training. Use commas to separate multiple datasets."},
+    )
+    eval_dataset: Optional[str] = field(
+        default=None,
+        metadata={"help": "The name of dataset(s) to use for evaluation. Use commas to separate multiple datasets."},
+    )
+    dataset_dir: str = field(
+        default="data",
+        metadata={"help": "Path to the folder containing the datasets."},
+    )
+    cutoff_len: int = field(
+        default=1024,
+        metadata={"help": "The cutoff length of the tokenized inputs in the dataset."},
+    )
+    train_on_prompt: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to disable the mask on the prompt."},
+    )
+    mask_history: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to mask the history and train on the last turn only."},
+    )
+    streaming: bool = field(
+        default=False,
+        metadata={"help": "Enable dataset streaming."},
+    )
+    buffer_size: int = field(
+        default=16384,
+        metadata={"help": "Size of the buffer to randomly sample examples from in dataset streaming."},
+    )
+    mix_strategy: Literal["concat", "interleave_under", "interleave_over"] = field(
+        default="concat",
+        metadata={"help": "Strategy to use in dataset mixing (concat/interleave) (undersampling/oversampling)."},
+    )
+    interleave_probs: Optional[str] = field(
+        default=None,
+        metadata={"help": "Probabilities to sample data from datasets. Use commas to separate multiple datasets."},
+    )
+    overwrite_cache: bool = field(
+        default=False,
+        metadata={"help": "Overwrite the cached training and evaluation sets."},
+    )
+    preprocessing_num_workers: Optional[int] = field(
+        default=None,
+        metadata={"help": "The number of processes to use for the pre-processing."},
+    )
+    max_samples: Optional[int] = field(
+        default=None,
+        metadata={"help": "For debugging purposes, truncate the number of examples for each dataset."},
+    )
+    eval_num_beams: Optional[int] = field(
+        default=None,
+        metadata={"help": "Number of beams to use for evaluation. This argument will be passed to `model.generate`"},
+    )
+    ignore_pad_token_for_loss: bool = field(
+        default=True,
+        metadata={"help": "Whether or not to ignore the tokens corresponding to the pad label in loss computation."},
+    )
+    val_size: float = field(
+        default=0.0,
+        metadata={"help": "Size of the development set, should be an integer or a float in range `[0,1)`."},
+    )
+    packing: Optional[bool] = field(
+        default=None,
+        metadata={"help": "Enable sequences packing in training. Will automatically enable in pre-training."},
+    )
+    neat_packing: bool = field(
+        default=False,
+        metadata={"help": "Enable sequence packing without cross-attention."},
+    )
+    tool_format: Optional[str] = field(
+        default=None,
+        metadata={"help": "Tool format to use for constructing function calling examples."},
+    )
+    tokenized_path: Optional[str] = field(
+        default=None,
+        metadata={"help": "Path to save or load the tokenized datasets."},
+    )
+
+    def __post_init__(self):
+        def split_arg(arg):
+            if isinstance(arg, str):
+                return [item.strip() for item in arg.split(",")]
+            return arg
+
+        self.dataset = split_arg(self.dataset)
+        self.eval_dataset = split_arg(self.eval_dataset)
+
+        if self.dataset is None and self.val_size > 1e-6:
+            raise ValueError("Cannot specify `val_size` if `dataset` is None.")
+
+        if self.eval_dataset is not None and self.val_size > 1e-6:
+            raise ValueError("Cannot specify `val_size` if `eval_dataset` is not None.")
+
+        if self.interleave_probs is not None:
+            if self.mix_strategy == "concat":
+                raise ValueError("`interleave_probs` is only valid for interleaved mixing.")
+
+            self.interleave_probs = list(map(float, split_arg(self.interleave_probs)))
+            if self.dataset is not None and len(self.dataset) != len(self.interleave_probs):
+                raise ValueError("The length of dataset and interleave probs should be identical.")
+
+            if self.eval_dataset is not None and len(self.eval_dataset) != len(self.interleave_probs):
+                raise ValueError("The length of eval dataset and interleave probs should be identical.")
+
+        if self.streaming and self.val_size > 1e-6 and self.val_size < 1:
+            raise ValueError("Streaming mode should have an integer val size.")
+
+        if self.streaming and self.max_samples is not None:
+            raise ValueError("`max_samples` is incompatible with `streaming`.")

llama-factory/src/llamafactory/hparams/evaluation_args.py

@@ -0,0 +1,62 @@
+# 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 dataclasses import dataclass, field
+from typing import Literal, Optional
+
+from datasets import DownloadMode
+
+
+@dataclass
+class EvaluationArguments:
+    r"""
+    Arguments pertaining to specify the evaluation parameters.
+    """
+
+    task: str = field(
+        metadata={"help": "Name of the evaluation task."},
+    )
+    task_dir: str = field(
+        default="evaluation",
+        metadata={"help": "Path to the folder containing the evaluation datasets."},
+    )
+    batch_size: int = field(
+        default=4,
+        metadata={"help": "The batch size per GPU for evaluation."},
+    )
+    seed: int = field(
+        default=42,
+        metadata={"help": "Random seed to be used with data loaders."},
+    )
+    lang: Literal["en", "zh"] = field(
+        default="en",
+        metadata={"help": "Language used at evaluation."},
+    )
+    n_shot: int = field(
+        default=5,
+        metadata={"help": "Number of examplars for few-shot learning."},
+    )
+    save_dir: Optional[str] = field(
+        default=None,
+        metadata={"help": "Path to save the evaluation results."},
+    )
+    download_mode: DownloadMode = field(
+        default=DownloadMode.REUSE_DATASET_IF_EXISTS,
+        metadata={"help": "Download mode used for the evaluation datasets."},
+    )
+
+    def __post_init__(self):
+        if self.save_dir is not None and os.path.exists(self.save_dir):
+            raise ValueError("`save_dir` already exists, use another one.")

llama-factory/src/llamafactory/hparams/finetuning_args.py

@@ -0,0 +1,400 @@
+# 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 dataclasses import dataclass, field
+from typing import List, Literal, Optional
+
+
+@dataclass
+class FreezeArguments:
+    r"""
+    Arguments pertaining to the freeze (partial-parameter) training.
+    """
+
+    freeze_trainable_layers: int = field(
+        default=2,
+        metadata={
+            "help": (
+                "The number of trainable layers for freeze (partial-parameter) fine-tuning. "
+                "Positive numbers mean the last n layers are set as trainable, "
+                "negative numbers mean the first n layers are set as trainable."
+            )
+        },
+    )
+    freeze_trainable_modules: str = field(
+        default="all",
+        metadata={
+            "help": (
+                "Name(s) of trainable modules for freeze (partial-parameter) fine-tuning. "
+                "Use commas to separate multiple modules. "
+                "Use `all` to specify all the available modules."
+            )
+        },
+    )
+    freeze_extra_modules: Optional[str] = field(
+        default=None,
+        metadata={
+            "help": (
+                "Name(s) of modules apart from hidden layers to be set as trainable "
+                "for freeze (partial-parameter) fine-tuning. "
+                "Use commas to separate multiple modules."
+            )
+        },
+    )
+
+
+@dataclass
+class LoraArguments:
+    r"""
+    Arguments pertaining to the LoRA training.
+    """
+
+    additional_target: Optional[str] = field(
+        default=None,
+        metadata={
+            "help": (
+                "Name(s) of modules apart from LoRA layers to be set as trainable "
+                "and saved in the final checkpoint. "
+                "Use commas to separate multiple modules."
+            )
+        },
+    )
+    lora_alpha: Optional[int] = field(
+        default=None,
+        metadata={"help": "The scale factor for LoRA fine-tuning (default: lora_rank * 2)."},
+    )
+    lora_dropout: float = field(
+        default=0.0,
+        metadata={"help": "Dropout rate for the LoRA fine-tuning."},
+    )
+    lora_rank: int = field(
+        default=8,
+        metadata={"help": "The intrinsic dimension for LoRA fine-tuning."},
+    )
+    lora_target: str = field(
+        default="all",
+        metadata={
+            "help": (
+                "Name(s) of target modules to apply LoRA. "
+                "Use commas to separate multiple modules. "
+                "Use `all` to specify all the linear modules."
+            )
+        },
+    )
+    loraplus_lr_ratio: Optional[float] = field(
+        default=None,
+        metadata={"help": "LoRA plus learning rate ratio (lr_B / lr_A)."},
+    )
+    loraplus_lr_embedding: float = field(
+        default=1e-6,
+        metadata={"help": "LoRA plus learning rate for lora embedding layers."},
+    )
+    use_rslora: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to use the rank stabilization scaling factor for LoRA layer."},
+    )
+    use_dora: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to use the weight-decomposed lora method (DoRA)."},
+    )
+    pissa_init: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to initialize a PiSSA adapter."},
+    )
+    pissa_iter: int = field(
+        default=16,
+        metadata={"help": "The number of iteration steps performed by FSVD in PiSSA. Use -1 to disable it."},
+    )
+    pissa_convert: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to convert the PiSSA adapter to a normal LoRA adapter."},
+    )
+    create_new_adapter: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to create a new adapter with randomly initialized weight."},
+    )
+
+
+@dataclass
+class RLHFArguments:
+    r"""
+    Arguments pertaining to the PPO, DPO and KTO training.
+    """
+
+    pref_beta: float = field(
+        default=0.1,
+        metadata={"help": "The beta parameter in the preference loss."},
+    )
+    pref_ftx: float = field(
+        default=0.0,
+        metadata={"help": "The supervised fine-tuning loss coefficient in DPO training."},
+    )
+    pref_loss: Literal["sigmoid", "hinge", "ipo", "kto_pair", "orpo", "simpo"] = field(
+        default="sigmoid",
+        metadata={"help": "The type of DPO loss to use."},
+    )
+    dpo_label_smoothing: float = field(
+        default=0.0,
+        metadata={"help": "The robust DPO label smoothing parameter in cDPO that should be between 0 and 0.5."},
+    )
+    kto_chosen_weight: float = field(
+        default=1.0,
+        metadata={"help": "The weight factor of the desirable losses in KTO training."},
+    )
+    kto_rejected_weight: float = field(
+        default=1.0,
+        metadata={"help": "The weight factor of the undesirable losses in KTO training."},
+    )
+    simpo_gamma: float = field(
+        default=0.5,
+        metadata={"help": "The target reward margin term in SimPO loss."},
+    )
+    ppo_buffer_size: int = field(
+        default=1,
+        metadata={"help": "The number of mini-batches to make experience buffer in a PPO optimization step."},
+    )
+    ppo_epochs: int = field(
+        default=4,
+        metadata={"help": "The number of epochs to perform in a PPO optimization step."},
+    )
+    ppo_score_norm: bool = field(
+        default=False,
+        metadata={"help": "Use score normalization in PPO training."},
+    )
+    ppo_target: float = field(
+        default=6.0,
+        metadata={"help": "Target KL value for adaptive KL control in PPO training."},
+    )
+    ppo_whiten_rewards: bool = field(
+        default=False,
+        metadata={"help": "Whiten the rewards before compute advantages in PPO training."},
+    )
+    ref_model: Optional[str] = field(
+        default=None,
+        metadata={"help": "Path to the reference model used for the PPO or DPO training."},
+    )
+    ref_model_adapters: Optional[str] = field(
+        default=None,
+        metadata={"help": "Path to the adapters of the reference model."},
+    )
+    ref_model_quantization_bit: Optional[int] = field(
+        default=None,
+        metadata={"help": "The number of bits to quantize the reference model."},
+    )
+    reward_model: Optional[str] = field(
+        default=None,
+        metadata={"help": "Path to the reward model used for the PPO training."},
+    )
+    reward_model_adapters: Optional[str] = field(
+        default=None,
+        metadata={"help": "Path to the adapters of the reward model."},
+    )
+    reward_model_quantization_bit: Optional[int] = field(
+        default=None,
+        metadata={"help": "The number of bits to quantize the reward model."},
+    )
+    reward_model_type: Literal["lora", "full", "api"] = field(
+        default="lora",
+        metadata={"help": "The type of the reward model in PPO training. Lora model only supports lora training."},
+    )
+
+
+@dataclass
+class GaloreArguments:
+    r"""
+    Arguments pertaining to the GaLore algorithm.
+    """
+
+    use_galore: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to use the gradient low-Rank projection (GaLore)."},
+    )
+    galore_target: str = field(
+        default="all",
+        metadata={
+            "help": (
+                "Name(s) of modules to apply GaLore. Use commas to separate multiple modules. "
+                "Use `all` to specify all the linear modules."
+            )
+        },
+    )
+    galore_rank: int = field(
+        default=16,
+        metadata={"help": "The rank of GaLore gradients."},
+    )
+    galore_update_interval: int = field(
+        default=200,
+        metadata={"help": "Number of steps to update the GaLore projection."},
+    )
+    galore_scale: float = field(
+        default=0.25,
+        metadata={"help": "GaLore scaling coefficient."},
+    )
+    galore_proj_type: Literal["std", "reverse_std", "right", "left", "full"] = field(
+        default="std",
+        metadata={"help": "Type of GaLore projection."},
+    )
+    galore_layerwise: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to enable layer-wise update to further save memory."},
+    )
+
+
+@dataclass
+class BAdamArgument:
+    r"""
+    Arguments pertaining to the BAdam optimizer.
+    """
+
+    use_badam: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to use the BAdam optimizer."},
+    )
+    badam_mode: Literal["layer", "ratio"] = field(
+        default="layer",
+        metadata={"help": "Whether to use layer-wise or ratio-wise BAdam optimizer."},
+    )
+    badam_start_block: Optional[int] = field(
+        default=None,
+        metadata={"help": "The starting block index for layer-wise BAdam."},
+    )
+    badam_switch_mode: Optional[Literal["ascending", "descending", "random", "fixed"]] = field(
+        default="ascending",
+        metadata={"help": "the strategy of picking block to update for layer-wise BAdam."},
+    )
+    badam_switch_interval: Optional[int] = field(
+        default=50,
+        metadata={
+            "help": "Number of steps to update the block for layer-wise BAdam. Use -1 to disable the block update."
+        },
+    )
+    badam_update_ratio: float = field(
+        default=0.05,
+        metadata={"help": "The ratio of the update for ratio-wise BAdam."},
+    )
+    badam_mask_mode: Literal["adjacent", "scatter"] = field(
+        default="adjacent",
+        metadata={
+            "help": (
+                "The mode of the mask for BAdam optimizer. "
+                "`adjacent` means that the trainable parameters are adjacent to each other, "
+                "`scatter` means that trainable parameters are randomly choosed from the weight."
+            )
+        },
+    )
+    badam_verbose: int = field(
+        default=0,
+        metadata={
+            "help": (
+                "The verbosity level of BAdam optimizer. "
+                "0 for no print, 1 for print the block prefix, 2 for print trainable parameters."
+            )
+        },
+    )
+
+
+@dataclass
+class FinetuningArguments(FreezeArguments, LoraArguments, RLHFArguments, GaloreArguments, BAdamArgument):
+    r"""
+    Arguments pertaining to which techniques we are going to fine-tuning with.
+    """
+
+    pure_bf16: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to train model in purely bf16 precision (without AMP)."},
+    )
+    stage: Literal["pt", "sft", "rm", "ppo", "dpo", "kto"] = field(
+        default="sft",
+        metadata={"help": "Which stage will be performed in training."},
+    )
+    finetuning_type: Literal["lora", "freeze", "full"] = field(
+        default="lora",
+        metadata={"help": "Which fine-tuning method to use."},
+    )
+    use_llama_pro: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to make only the parameters in the expanded blocks trainable."},
+    )
+    freeze_vision_tower: bool = field(
+        default=True,
+        metadata={"help": "Whether ot not to freeze vision tower in MLLM training."},
+    )
+    train_mm_proj_only: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to train the multimodal projector for MLLM only."},
+    )
+    compute_accuracy: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to compute the token-level accuracy at evaluation."},
+    )
+    plot_loss: bool = field(
+        default=False,
+        metadata={"help": "Whether or not to save the training loss curves."},
+    )
+
+    def __post_init__(self):
+        def split_arg(arg):
+            if isinstance(arg, str):
+                return [item.strip() for item in arg.split(",")]
+            return arg
+
+        self.freeze_trainable_modules: List[str] = split_arg(self.freeze_trainable_modules)
+        self.freeze_extra_modules: Optional[List[str]] = split_arg(self.freeze_extra_modules)
+        self.lora_alpha: int = self.lora_alpha or self.lora_rank * 2
+        self.lora_target: List[str] = split_arg(self.lora_target)
+        self.additional_target: Optional[List[str]] = split_arg(self.additional_target)
+        self.galore_target: List[str] = split_arg(self.galore_target)
+        self.freeze_vision_tower = self.freeze_vision_tower or self.train_mm_proj_only
+        self.use_ref_model = self.stage == "dpo" and self.pref_loss not in ["orpo", "simpo"]
+
+        assert self.finetuning_type in ["lora", "freeze", "full"], "Invalid fine-tuning method."
+        assert self.ref_model_quantization_bit in [None, 8, 4], "We only accept 4-bit or 8-bit quantization."
+        assert self.reward_model_quantization_bit in [None, 8, 4], "We only accept 4-bit or 8-bit quantization."
+
+        if self.stage == "ppo" and self.reward_model is None:
+            raise ValueError("`reward_model` is necessary for PPO training.")
+
+        if self.stage == "ppo" and self.reward_model_type == "lora" and self.finetuning_type != "lora":
+            raise ValueError("`reward_model_type` cannot be lora for Freeze/Full PPO training.")
+
+        if self.stage == "dpo" and self.pref_loss != "sigmoid" and self.dpo_label_smoothing > 1e-6:
+            raise ValueError("`dpo_label_smoothing` is only valid for sigmoid loss function.")
+
+        if self.use_llama_pro and self.finetuning_type == "full":
+            raise ValueError("`use_llama_pro` is only valid for Freeze or LoRA training.")
+
+        if self.finetuning_type == "lora" and (self.use_galore or self.use_badam):
+            raise ValueError("Cannot use LoRA with GaLore or BAdam together.")
+
+        if self.use_galore and self.use_badam:
+            raise ValueError("Cannot use GaLore with BAdam together.")
+
+        if self.pissa_init and (self.stage in ["ppo", "kto"] or self.use_ref_model):
+            raise ValueError("Cannot use PiSSA for current training stage.")
+
+        if self.train_mm_proj_only and self.finetuning_type != "full":
+            raise ValueError("`train_mm_proj_only` is only valid for full training.")
+
+        if self.finetuning_type != "lora":
+            if self.loraplus_lr_ratio is not None:
+                raise ValueError("`loraplus_lr_ratio` is only valid for LoRA training.")
+
+            if self.use_rslora:
+                raise ValueError("`use_rslora` is only valid for LoRA training.")
+
+            if self.use_dora:
+                raise ValueError("`use_dora` is only valid for LoRA training.")
+
+            if self.pissa_init:
+                raise ValueError("`pissa_init` is only valid for LoRA training.")

llama-factory/src/llamafactory/hparams/generating_args.py

@@ -0,0 +1,74 @@
+# 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 dataclasses import asdict, dataclass, field
+from typing import Any, Dict, Optional
+
+
+@dataclass
+class GeneratingArguments:
+    r"""
+    Arguments pertaining to specify the decoding parameters.
+    """
+
+    do_sample: bool = field(
+        default=True,
+        metadata={"help": "Whether or not to use sampling, use greedy decoding otherwise."},
+    )
+    temperature: float = field(
+        default=0.95,
+        metadata={"help": "The value used to modulate the next token probabilities."},
+    )
+    top_p: float = field(
+        default=0.7,
+        metadata={
+            "help": "The smallest set of most probable tokens with probabilities that add up to top_p or higher are kept."
+        },
+    )
+    top_k: int = field(
+        default=50,
+        metadata={"help": "The number of highest probability vocabulary tokens to keep for top-k filtering."},
+    )
+    num_beams: int = field(
+        default=1,
+        metadata={"help": "Number of beams for beam search. 1 means no beam search."},
+    )
+    max_length: int = field(
+        default=1024,
+        metadata={"help": "The maximum length the generated tokens can have. It can be overridden by max_new_tokens."},
+    )
+    max_new_tokens: int = field(
+        default=1024,
+        metadata={"help": "The maximum numbers of tokens to generate, ignoring the number of tokens in the prompt."},
+    )
+    repetition_penalty: float = field(
+        default=1.0,
+        metadata={"help": "The parameter for repetition penalty. 1.0 means no penalty."},
+    )
+    length_penalty: float = field(
+        default=1.0,
+        metadata={"help": "Exponential penalty to the length that is used with beam-based generation."},
+    )
+    default_system: Optional[str] = field(
+        default=None,
+        metadata={"help": "Default system message to use in chat completion."},
+    )
+
+    def to_dict(self) -> Dict[str, Any]:
+        args = asdict(self)
+        if args.get("max_new_tokens", -1) > 0:
+            args.pop("max_length", None)
+        else:
+            args.pop("max_new_tokens", None)
+        return args