release v0.2.2, fix #1478 #1466

Former-commit-id: c9534c411716e1dceb54c5eb35fe845c93ee2973
fix #424
2023-11-13 23:09:05 +08:00 · 2023-11-13 22:42:23 +08:00 · 2023-11-13 22:20:35 +08:00 · 2023-11-10 18:34:54 +08:00 · 2023-11-10 14:38:18 +08:00 · 2023-11-10 14:16:10 +08:00
66 changed files with 2199 additions and 1241 deletions
--- a/README.md
+++ b/README.md
@@ -12,14 +12,18 @@

 \[ English | [中文](README_zh.md) \]

-## Example: Fine-tuning large language model within 10 minutes
+## LLaMA Board: A One-stop Web UI for Getting Started with LLaMA Factory

-Launch an **all-in-one Web UI** via `python src/train_web.py`.
+Launch **LLaMA Board** via `CUDA_VISIBLE_DEVICES=0 python src/train_web.py`. (multiple GPUs are not supported yet)
+
+Here is an example of altering the self-cognition of an instruction-tuned language model within 10 minutes on a single GPU.

 https://github.com/hiyouga/LLaMA-Factory/assets/16256802/6ba60acc-e2e2-4bec-b846-2d88920d5ba1

 ## Changelog

+[23/10/21] We supported **[NEFTune](https://arxiv.org/abs/2310.05914)** trick for fine-tuning. Try `--neft_alpha` argument to activate NEFTune, e.g., `--neft_alpha 5`.
+
 [23/09/27] We supported **$S^2$-Attn** proposed by [LongLoRA](https://github.com/dvlab-research/LongLoRA) for the LLaMA models. Try `--shift_attn` argument to enable shift short attention.

 [23/09/23] We integrated MMLU, C-Eval and CMMLU benchmarks in this repo. See [this example](#evaluation) to evaluate your models.
@@ -48,24 +52,27 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/6ba60acc-e2e2-4bec-b846

 | Model                                                    | Model size                  | Default module    | Template  |
 | -------------------------------------------------------- | --------------------------- | ----------------- | --------- |
-| [LLaMA](https://github.com/facebookresearch/llama)       | 7B/13B/33B/65B              | q_proj,v_proj     | -         |
-| [LLaMA-2](https://huggingface.co/meta-llama)             | 7B/13B/70B                  | q_proj,v_proj     | llama2    |
-| [BLOOM](https://huggingface.co/bigscience/bloom)         | 560M/1.1B/1.7B/3B/7.1B/176B | query_key_value   | -         |
-| [BLOOMZ](https://huggingface.co/bigscience/bloomz)       | 560M/1.1B/1.7B/3B/7.1B/176B | query_key_value   | -         |
-| [Falcon](https://huggingface.co/tiiuae/falcon-7b)        | 7B/40B                      | query_key_value   | -         |
 | [Baichuan](https://github.com/baichuan-inc/Baichuan-13B) | 7B/13B                      | W_pack            | baichuan  |
 | [Baichuan2](https://github.com/baichuan-inc/Baichuan2)   | 7B/13B                      | W_pack            | baichuan2 |
+| [BLOOM](https://huggingface.co/bigscience/bloom)         | 560M/1.1B/1.7B/3B/7.1B/176B | query_key_value   | -         |
+| [BLOOMZ](https://huggingface.co/bigscience/bloomz)       | 560M/1.1B/1.7B/3B/7.1B/176B | query_key_value   | -         |
+| [ChatGLM3](https://github.com/THUDM/ChatGLM3)            | 6B                          | query_key_value   | chatglm3  |
+| [Falcon](https://huggingface.co/tiiuae/falcon-7b)        | 7B/40B/180B                 | query_key_value   | falcon    |
 | [InternLM](https://github.com/InternLM/InternLM)         | 7B/20B                      | q_proj,v_proj     | intern    |
-| [Qwen](https://github.com/QwenLM/Qwen-7B)                | 7B/14B                      | c_attn            | chatml    |
-| [XVERSE](https://github.com/xverse-ai/XVERSE-13B)        | 13B                         | q_proj,v_proj     | xverse    |
-| [ChatGLM2](https://github.com/THUDM/ChatGLM2-6B)         | 6B                          | query_key_value   | chatglm2  |
+| [LLaMA](https://github.com/facebookresearch/llama)       | 7B/13B/33B/65B              | q_proj,v_proj     | -         |
+| [LLaMA-2](https://huggingface.co/meta-llama)             | 7B/13B/70B                  | q_proj,v_proj     | llama2    |
+| [Mistral](https://huggingface.co/mistralai)              | 7B                          | q_proj,v_proj     | mistral   |
 | [Phi-1.5](https://huggingface.co/microsoft/phi-1_5)      | 1.3B                        | Wqkv              | -         |
+| [Qwen](https://github.com/QwenLM/Qwen)                   | 7B/14B                      | c_attn            | qwen      |
+| [XVERSE](https://github.com/xverse-ai)                   | 7B/13B/65B                  | q_proj,v_proj     | xverse    |

 > [!NOTE]
 > **Default module** is used for the `--lora_target` argument, you can use `--lora_target all` to specify all the available modules.
 >
 > 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 "chat" models.

+Please refer to [template.py](src/llmtuner/extras/template.py) for a full list of models we supported.
+
 ## Supported Training Approaches

 | Approach               |   Full-parameter   | Partial-parameter  |       LoRA         |       QLoRA        |
@@ -81,39 +88,61 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/6ba60acc-e2e2-4bec-b846

 ## Provided Datasets

- For pre-training:
-  - [Wiki Demo (en)](data/wiki_demo.txt)
-  - [RefinedWeb (en)](https://huggingface.co/datasets/tiiuae/falcon-refinedweb)
-  - [StarCoder (en)](https://huggingface.co/datasets/bigcode/starcoderdata)
-  - [Wikipedia (en)](https://huggingface.co/datasets/olm/olm-wikipedia-20221220)
-  - [Wikipedia (zh)](https://huggingface.co/datasets/pleisto/wikipedia-cn-20230720-filtered)
- For supervised fine-tuning:
-  - [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)
-  - [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca)
-  - [GPT-4 Generated Data (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
-  - [Open Assistant (multilingual)](https://huggingface.co/datasets/OpenAssistant/oasst1)
-  - [Self-cognition (zh)](data/self_cognition.json)
-  - [ShareGPT (zh)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT/tree/main/Chinese-instruction-collection)
-  - [Guanaco Dataset (multilingual)](https://huggingface.co/datasets/JosephusCheung/GuanacoDataset)
-  - [BELLE 2M (zh)](https://huggingface.co/datasets/BelleGroup/train_2M_CN)
-  - [BELLE 1M (zh)](https://huggingface.co/datasets/BelleGroup/train_1M_CN)
-  - [BELLE 0.5M (zh)](https://huggingface.co/datasets/BelleGroup/train_0.5M_CN)
-  - [BELLE Dialogue 0.4M (zh)](https://huggingface.co/datasets/BelleGroup/generated_chat_0.4M)
-  - [BELLE School Math 0.25M (zh)](https://huggingface.co/datasets/BelleGroup/school_math_0.25M)
-  - [BELLE Multiturn Chat 0.8M (zh)](https://huggingface.co/datasets/BelleGroup/multiturn_chat_0.8M)
-  - [LIMA (en)](https://huggingface.co/datasets/GAIR/lima)
-  - [CodeAlpaca 20k (en)](https://huggingface.co/datasets/sahil2801/CodeAlpaca-20k)
-  - [Alpaca CoT (multilingual)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT)
-  - [MathInstruct (en)](https://huggingface.co/datasets/TIGER-Lab/MathInstruct)
-  - [Firefly 1.1M (zh)](https://huggingface.co/datasets/YeungNLP/firefly-train-1.1M)
-  - [Web QA (zh)](https://huggingface.co/datasets/suolyer/webqa)
-  - [UltraChat (en)](https://github.com/thunlp/UltraChat)
-  - [WebNovel (zh)](https://huggingface.co/datasets/zxbsmk/webnovel_cn)
-  - [Ad Gen (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)
- For reward modeling or DPO training:
-  - [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
-  - [Open Assistant (multilingual)](https://huggingface.co/datasets/OpenAssistant/oasst1)
-  - [GPT-4 Generated Data (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
+<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)
+- [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>
+
+- [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)
+- [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca)
+- [GPT-4 Generated Data (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
+- [Self-cognition (zh)](data/self_cognition.json)
+- [Open Assistant (multilingual)](https://huggingface.co/datasets/OpenAssistant/oasst1)
+- [ShareGPT (zh)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT/tree/main/Chinese-instruction-collection)
+- [Guanaco Dataset (multilingual)](https://huggingface.co/datasets/JosephusCheung/GuanacoDataset)
+- [BELLE 2M (zh)](https://huggingface.co/datasets/BelleGroup/train_2M_CN)
+- [BELLE 1M (zh)](https://huggingface.co/datasets/BelleGroup/train_1M_CN)
+- [BELLE 0.5M (zh)](https://huggingface.co/datasets/BelleGroup/train_0.5M_CN)
+- [BELLE Dialogue 0.4M (zh)](https://huggingface.co/datasets/BelleGroup/generated_chat_0.4M)
+- [BELLE School Math 0.25M (zh)](https://huggingface.co/datasets/BelleGroup/school_math_0.25M)
+- [BELLE Multiturn Chat 0.8M (zh)](https://huggingface.co/datasets/BelleGroup/multiturn_chat_0.8M)
+- [UltraChat (en)](https://github.com/thunlp/UltraChat)
+- [LIMA (en)](https://huggingface.co/datasets/GAIR/lima)
+- [OpenPlatypus (en)](https://huggingface.co/datasets/garage-bAInd/Open-Platypus)
+- [CodeAlpaca 20k (en)](https://huggingface.co/datasets/sahil2801/CodeAlpaca-20k)
+- [Alpaca CoT (multilingual)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT)
+- [MathInstruct (en)](https://huggingface.co/datasets/TIGER-Lab/MathInstruct)
+- [Firefly 1.1M (zh)](https://huggingface.co/datasets/YeungNLP/firefly-train-1.1M)
+- [Web QA (zh)](https://huggingface.co/datasets/suolyer/webqa)
+- [WebNovel (zh)](https://huggingface.co/datasets/zxbsmk/webnovel_cn)
+- [Ad Gen (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)
+- [ShareGPT Hyperfiltered (en)](https://huggingface.co/datasets/totally-not-an-llm/sharegpt-hyperfiltered-3k)
+- [ShareGPT4 (en&zh)](https://huggingface.co/datasets/shibing624/sharegpt_gpt4)
+- [UltraChat 200k (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k)
+- [AgentInstruct (en)](https://huggingface.co/datasets/THUDM/AgentInstruct)
+- [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)
+
+</details>
+
+<details><summary>Preference datasets</summary>
+
+- [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
+- [Open Assistant (multilingual)](https://huggingface.co/datasets/OpenAssistant/oasst1)
+- [GPT-4 Generated Data (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
+
+</details>

 Please refer to [data/README.md](data/README.md) for details.

@@ -129,9 +158,9 @@ huggingface-cli login
 - Python 3.8+ and PyTorch 1.13.1+
 - 🤗Transformers, Datasets, Accelerate, PEFT and TRL
 - sentencepiece, protobuf and tiktoken
- fire, jieba, rouge-chinese and nltk (used at evaluation and predict)
- gradio and matplotlib (used in web_demo.py)
- uvicorn, fastapi and sse-starlette (used in api_demo.py)
+- jieba, rouge-chinese and nltk (used at evaluation and predict)
+- gradio and matplotlib (used in web UI)
+- uvicorn, fastapi and sse-starlette (used in API)

 And **powerful GPUs**!

@@ -139,7 +168,7 @@ And **powerful GPUs**!

 ### Data Preparation (optional)

-Please refer to `data/example_dataset` for checking the details about the format of dataset files. You can either use a single `.json` file or a [dataset loading script](https://huggingface.co/docs/datasets/dataset_script) with multiple files to create a custom dataset.
+Please refer to [data/README.md](data/README.md) for checking the details about the format of dataset files. You can either use a single `.json` file or a [dataset loading script](https://huggingface.co/docs/datasets/dataset_script) with multiple files to create a custom dataset.

 > [!NOTE]
 > Please update `data/dataset_info.json` to use your custom dataset. About the format of this file, please refer to `data/README.md`.
@@ -160,17 +189,6 @@ If you want to enable the quantized LoRA (QLoRA) on the Windows platform, you wi
 pip install https://github.com/jllllll/bitsandbytes-windows-webui/releases/download/wheels/bitsandbytes-0.39.1-py3-none-win_amd64.whl
 ```

-### All-in-one Web UI
-
-```bash
-CUDA_VISIBLE_DEVICES=0 python src/train_web.py
-```
-
-We **strongly recommend** using the all-in-one Web UI for newcomers since it can also generate training scripts automatically, even without a GPU environment.
-
-> [!WARNING]
-> Currently the web UI only supports training on **a single GPU**.
-
 ### Train on a single GPU

 > [!IMPORTANT]
@@ -377,8 +395,7 @@ python src/export_model.py \
    --template default \
    --finetuning_type lora \
    --checkpoint_dir path_to_checkpoint \
-    --export_dir path_to_export \
-    --fp16
+    --export_dir path_to_export
 ```

 ### API Demo
@@ -449,11 +466,18 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
 > [!NOTE]
 > We recommend using `--per_device_eval_batch_size=1` and `--max_target_length 128` at 4/8-bit predict.

+## Projects using LLaMA Factory
+
+- **[StarWhisper](https://github.com/Yu-Yang-Li/StarWhisper)**: A large language model for Astronomy, based on ChatGLM2-6B and Qwen-14B.
+- **[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.
+- **[Sunsimiao](https://github.com/thomas-yanxin/Sunsimiao)**: A large language model specialized in Chinese medical domain, based on Baichuan-7B and ChatGLM-6B.
+- **[CareGPT](https://github.com/WangRongsheng/CareGPT)**: A series of large language models for Chinese medical domain, based on LLaMA2-7B and Baichuan-13B.
+
 ## License

 This repository is licensed under the [Apache-2.0 License](LICENSE).

-Please follow the model licenses to use the corresponding model weights: [LLaMA](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md) / [LLaMA-2](https://ai.meta.com/llama/license/) / [BLOOM](https://huggingface.co/spaces/bigscience/license) / [Falcon](LICENSE) / [Baichuan](https://huggingface.co/baichuan-inc/baichuan-7B/resolve/main/baichuan-7B%20%E6%A8%A1%E5%9E%8B%E8%AE%B8%E5%8F%AF%E5%8D%8F%E8%AE%AE.pdf) / [Baichuan2](https://huggingface.co/baichuan-inc/Baichuan2-7B-Base/resolve/main/Baichuan%202%E6%A8%A1%E5%9E%8B%E7%A4%BE%E5%8C%BA%E8%AE%B8%E5%8F%AF%E5%8D%8F%E8%AE%AE.pdf) / [InternLM](https://github.com/InternLM/InternLM#open-source-license) / [Qwen](https://huggingface.co/Qwen/Qwen-7B-Chat/blob/main/LICENSE) / [XVERSE](https://github.com/xverse-ai/XVERSE-13B/blob/main/MODEL_LICENSE.pdf) / [ChatGLM2](https://github.com/THUDM/ChatGLM2-6B/blob/main/MODEL_LICENSE) / [Phi-1.5](https://huggingface.co/microsoft/phi-1_5/resolve/main/Research%20License.docx)
+Please follow the model licenses to use the corresponding model weights: [Baichuan](https://huggingface.co/baichuan-inc/Baichuan-13B-Base/resolve/main/Community%20License%20for%20Baichuan-13B%20Model.pdf) / [Baichuan2](https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat/resolve/main/Community%20License%20for%20Baichuan2%20Model.pdf) / [BLOOM](https://huggingface.co/spaces/bigscience/license) / [ChatGLM3](https://github.com/THUDM/ChatGLM3/blob/main/MODEL_LICENSE) / [Falcon](https://huggingface.co/tiiuae/falcon-180B/blob/main/LICENSE.txt) / [InternLM](https://github.com/InternLM/InternLM#license) / [LLaMA](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md) / [LLaMA-2](https://ai.meta.com/llama/license/) / [Mistral](LICENSE) / [Phi-1.5](https://huggingface.co/microsoft/phi-1_5/resolve/main/Research%20License.docx) / [Qwen](https://github.com/QwenLM/Qwen/blob/main/LICENSE) / [XVERSE](https://github.com/xverse-ai/XVERSE-13B/blob/main/MODEL_LICENSE.pdf)

 ## Citation

--- a/README_zh.md
+++ b/README_zh.md
@@ -12,14 +12,18 @@

 \[ [English](README.md) | 中文 \]

-## 示例：在十分钟内微调一个大模型
+## LLaMA Board: 通过一站式网页界面快速上手 LLaMA Factory

-通过 `python src/train_web.py` 开启**训练推理一体化界面**。
+使用 `CUDA_VISIBLE_DEVICES=0 python src/train_web.py` 启动 **LLaMA Board**。（该界面目前仅支持单卡训练）
+
+下面是使用单张 GPU 在 10 分钟内更改对话式大型语言模型自我认知的示例。

 https://github.com/hiyouga/LLaMA-Factory/assets/16256802/6ba60acc-e2e2-4bec-b846-2d88920d5ba1

 ## 更新日志

+[23/10/21] 我们支持了 **[NEFTune](https://arxiv.org/abs/2310.05914)** 训练技巧。请使用 `--neft_alpha` 参数启用 NEFTune，例如 `--neft_alpha 5`。
+
 [23/09/27] 我们针对 LLaMA 模型支持了 [LongLoRA](https://github.com/dvlab-research/LongLoRA) 提出的 **$S^2$-Attn**。请使用 `--shift_attn` 参数以启用该功能。

 [23/09/23] 我们在项目中集成了 MMLU、C-Eval 和 CMMLU 评估集。使用方法请参阅[此示例](#模型评估)。
@@ -34,7 +38,7 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/6ba60acc-e2e2-4bec-b846

 [23/07/29] 我们在 Hugging Face 发布了两个 13B 指令微调模型。详细内容请查阅我们的 Hugging Face 项目（[LLaMA-2](https://huggingface.co/hiyouga/Llama-2-Chinese-13b-chat) / [Baichuan](https://huggingface.co/hiyouga/Baichuan-13B-sft)）。

-[23/07/18] 我们开发了支持训练和测试的**浏览器一体化界面**。请尝试使用 `train_web.py` 在您的浏览器中微调模型。感谢 [@KanadeSiina](https://github.com/KanadeSiina) 和 [@codemayq](https://github.com/codemayq) 在该功能开发中付出的努力。
+[23/07/18] 我们开发了支持训练和测试的**浏览器一体化界面**。请使用 `train_web.py` 在您的浏览器中微调模型。感谢 [@KanadeSiina](https://github.com/KanadeSiina) 和 [@codemayq](https://github.com/codemayq) 在该功能开发中付出的努力。

 [23/07/09] 我们开源了 **[FastEdit](https://github.com/hiyouga/FastEdit)** ⚡🩹，一个简单易用的、能迅速编辑大模型事实记忆的工具包。如果您感兴趣请关注我们的 [FastEdit](https://github.com/hiyouga/FastEdit) 项目。

@@ -42,30 +46,33 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/6ba60acc-e2e2-4bec-b846

 [23/06/22] 我们对齐了[示例 API](src/api_demo.py) 与 [OpenAI API](https://platform.openai.com/docs/api-reference/chat) 的格式，您可以将微调模型接入**任意基于 ChatGPT 的应用**中。

-[23/06/03] 我们实现了 4 比特的 LoRA 训练（也称 **[QLoRA](https://github.com/artidoro/qlora)**）。请尝试使用 `--quantization_bit 4` 参数进行 4 比特量化微调。
+[23/06/03] 我们实现了 4 比特的 LoRA 训练（也称 **[QLoRA](https://github.com/artidoro/qlora)**）。请使用 `--quantization_bit 4` 参数进行 4 比特量化微调。

 ## 模型

 | 模型名                                                   | 模型大小                     | 默认模块           | Template  |
 | -------------------------------------------------------- | --------------------------- | ----------------- | --------- |
-| [LLaMA](https://github.com/facebookresearch/llama)       | 7B/13B/33B/65B              | q_proj,v_proj     | -         |
-| [LLaMA-2](https://huggingface.co/meta-llama)             | 7B/13B/70B                  | q_proj,v_proj     | llama2    |
-| [BLOOM](https://huggingface.co/bigscience/bloom)         | 560M/1.1B/1.7B/3B/7.1B/176B | query_key_value   | -         |
-| [BLOOMZ](https://huggingface.co/bigscience/bloomz)       | 560M/1.1B/1.7B/3B/7.1B/176B | query_key_value   | -         |
-| [Falcon](https://huggingface.co/tiiuae/falcon-7b)        | 7B/40B                      | query_key_value   | -         |
 | [Baichuan](https://github.com/baichuan-inc/Baichuan-13B) | 7B/13B                      | W_pack            | baichuan  |
 | [Baichuan2](https://github.com/baichuan-inc/Baichuan2)   | 7B/13B                      | W_pack            | baichuan2 |
+| [BLOOM](https://huggingface.co/bigscience/bloom)         | 560M/1.1B/1.7B/3B/7.1B/176B | query_key_value   | -         |
+| [BLOOMZ](https://huggingface.co/bigscience/bloomz)       | 560M/1.1B/1.7B/3B/7.1B/176B | query_key_value   | -         |
+| [ChatGLM3](https://github.com/THUDM/ChatGLM3)            | 6B                          | query_key_value   | chatglm3  |
+| [Falcon](https://huggingface.co/tiiuae/falcon-7b)        | 7B/40B/180B                 | query_key_value   | falcon    |
 | [InternLM](https://github.com/InternLM/InternLM)         | 7B/20B                      | q_proj,v_proj     | intern    |
-| [Qwen](https://github.com/QwenLM/Qwen-7B)                | 7B/14B                      | c_attn            | chatml    |
-| [XVERSE](https://github.com/xverse-ai/XVERSE-13B)        | 13B                         | q_proj,v_proj     | xverse    |
-| [ChatGLM2](https://github.com/THUDM/ChatGLM2-6B)         | 6B                          | query_key_value   | chatglm2  |
+| [LLaMA](https://github.com/facebookresearch/llama)       | 7B/13B/33B/65B              | q_proj,v_proj     | -         |
+| [LLaMA-2](https://huggingface.co/meta-llama)             | 7B/13B/70B                  | q_proj,v_proj     | llama2    |
+| [Mistral](https://huggingface.co/mistralai)              | 7B                          | q_proj,v_proj     | mistral   |
 | [Phi-1.5](https://huggingface.co/microsoft/phi-1_5)      | 1.3B                        | Wqkv              | -         |
+| [Qwen](https://github.com/QwenLM/Qwen)                   | 7B/14B                      | c_attn            | qwen      |
+| [XVERSE](https://github.com/xverse-ai)                   | 7B/13B/65B                  | q_proj,v_proj     | xverse    |

 > [!NOTE]
 > **默认模块**应作为 `--lora_target` 参数的默认值，可使用 `--lora_target all` 参数指定全部模块。
 >
 > 对于所有“基座”（Base）模型，`--template` 参数可以是 `default`, `alpaca`, `vicuna` 等任意值。但“对话”（Chat）模型请务必使用**对应的模板**。

+项目所支持模型的完整列表请参阅 [template.py](src/llmtuner/extras/template.py)。
+
 ## 训练方法

 | 方法                   |     全参数训练      |    部分参数训练     |       LoRA         |       QLoRA        |
@@ -81,41 +88,63 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/6ba60acc-e2e2-4bec-b846

 ## 数据集

- 用于预训练：
-  - [Wiki Demo (en)](data/wiki_demo.txt)
-  - [RefinedWeb (en)](https://huggingface.co/datasets/tiiuae/falcon-refinedweb)
-  - [StarCoder (en)](https://huggingface.co/datasets/bigcode/starcoderdata)
-  - [Wikipedia (en)](https://huggingface.co/datasets/olm/olm-wikipedia-20221220)
-  - [Wikipedia (zh)](https://huggingface.co/datasets/pleisto/wikipedia-cn-20230720-filtered)
- 用于指令监督微调：
-  - [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)
-  - [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca)
-  - [GPT-4 Generated Data (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
-  - [Open Assistant (multilingual)](https://huggingface.co/datasets/OpenAssistant/oasst1)
-  - [Self-cognition (zh)](data/self_cognition.json)
-  - [ShareGPT (zh)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT/tree/main/Chinese-instruction-collection)
-  - [Guanaco Dataset (multilingual)](https://huggingface.co/datasets/JosephusCheung/GuanacoDataset)
-  - [BELLE 2M (zh)](https://huggingface.co/datasets/BelleGroup/train_2M_CN)
-  - [BELLE 1M (zh)](https://huggingface.co/datasets/BelleGroup/train_1M_CN)
-  - [BELLE 0.5M (zh)](https://huggingface.co/datasets/BelleGroup/train_0.5M_CN)
-  - [BELLE Dialogue 0.4M (zh)](https://huggingface.co/datasets/BelleGroup/generated_chat_0.4M)
-  - [BELLE School Math 0.25M (zh)](https://huggingface.co/datasets/BelleGroup/school_math_0.25M)
-  - [BELLE Multiturn Chat 0.8M (zh)](https://huggingface.co/datasets/BelleGroup/multiturn_chat_0.8M)
-  - [LIMA (en)](https://huggingface.co/datasets/GAIR/lima)
-  - [CodeAlpaca 20k (en)](https://huggingface.co/datasets/sahil2801/CodeAlpaca-20k)
-  - [Alpaca CoT (multilingual)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT)
-  - [MathInstruct (en)](https://huggingface.co/datasets/TIGER-Lab/MathInstruct)
-  - [Firefly 1.1M (zh)](https://huggingface.co/datasets/YeungNLP/firefly-train-1.1M)
-  - [Web QA (zh)](https://huggingface.co/datasets/suolyer/webqa)
-  - [UltraChat (en)](https://github.com/thunlp/UltraChat)
-  - [WebNovel (zh)](https://huggingface.co/datasets/zxbsmk/webnovel_cn)
-  - [Ad Gen (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)
- 用于训练奖励模型或 DPO 训练：
-  - [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
-  - [Open Assistant (multilingual)](https://huggingface.co/datasets/OpenAssistant/oasst1)
-  - [GPT-4 Generated Data (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
+<details><summary>预训练数据集</summary>

-使用方法请参考 [data/README.md](data/README_zh.md) 文件。
+- [Wiki Demo (en)](data/wiki_demo.txt)
+- [RefinedWeb (en)](https://huggingface.co/datasets/tiiuae/falcon-refinedweb)
+- [RedPajama V2 (en)](https://huggingface.co/datasets/togethercomputer/RedPajama-Data-V2)
+- [Wikipedia (en)](https://huggingface.co/datasets/olm/olm-wikipedia-20221220)
+- [Wikipedia (zh)](https://huggingface.co/datasets/pleisto/wikipedia-cn-20230720-filtered)
+- [Pile (en)](https://huggingface.co/datasets/EleutherAI/pile)
+- [SkyPile (zh)](https://huggingface.co/datasets/Skywork/SkyPile-150B)
+- [The Stack (en)](https://huggingface.co/datasets/bigcode/the-stack)
+- [StarCoder (en)](https://huggingface.co/datasets/bigcode/starcoderdata)
+
+</details>
+
+<details><summary>指令微调数据集</summary>
+
+- [Stanford Alpaca (en)](https://github.com/tatsu-lab/stanford_alpaca)
+- [Stanford Alpaca (zh)](https://github.com/ymcui/Chinese-LLaMA-Alpaca)
+- [GPT-4 Generated Data (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
+- [Self-cognition (zh)](data/self_cognition.json)
+- [Open Assistant (multilingual)](https://huggingface.co/datasets/OpenAssistant/oasst1)
+- [ShareGPT (zh)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT/tree/main/Chinese-instruction-collection)
+- [Guanaco Dataset (multilingual)](https://huggingface.co/datasets/JosephusCheung/GuanacoDataset)
+- [BELLE 2M (zh)](https://huggingface.co/datasets/BelleGroup/train_2M_CN)
+- [BELLE 1M (zh)](https://huggingface.co/datasets/BelleGroup/train_1M_CN)
+- [BELLE 0.5M (zh)](https://huggingface.co/datasets/BelleGroup/train_0.5M_CN)
+- [BELLE Dialogue 0.4M (zh)](https://huggingface.co/datasets/BelleGroup/generated_chat_0.4M)
+- [BELLE School Math 0.25M (zh)](https://huggingface.co/datasets/BelleGroup/school_math_0.25M)
+- [BELLE Multiturn Chat 0.8M (zh)](https://huggingface.co/datasets/BelleGroup/multiturn_chat_0.8M)
+- [UltraChat (en)](https://github.com/thunlp/UltraChat)
+- [LIMA (en)](https://huggingface.co/datasets/GAIR/lima)
+- [OpenPlatypus (en)](https://huggingface.co/datasets/garage-bAInd/Open-Platypus)
+- [CodeAlpaca 20k (en)](https://huggingface.co/datasets/sahil2801/CodeAlpaca-20k)
+- [Alpaca CoT (multilingual)](https://huggingface.co/datasets/QingyiSi/Alpaca-CoT)
+- [MathInstruct (en)](https://huggingface.co/datasets/TIGER-Lab/MathInstruct)
+- [Firefly 1.1M (zh)](https://huggingface.co/datasets/YeungNLP/firefly-train-1.1M)
+- [Web QA (zh)](https://huggingface.co/datasets/suolyer/webqa)
+- [WebNovel (zh)](https://huggingface.co/datasets/zxbsmk/webnovel_cn)
+- [Ad Gen (zh)](https://huggingface.co/datasets/HasturOfficial/adgen)
+- [ShareGPT Hyperfiltered (en)](https://huggingface.co/datasets/totally-not-an-llm/sharegpt-hyperfiltered-3k)
+- [ShareGPT4 (en&zh)](https://huggingface.co/datasets/shibing624/sharegpt_gpt4)
+- [UltraChat 200k (en)](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k)
+- [AgentInstruct (en)](https://huggingface.co/datasets/THUDM/AgentInstruct)
+- [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)
+
+</details>
+
+<details><summary>偏好数据集</summary>
+
+- [HH-RLHF (en)](https://huggingface.co/datasets/Anthropic/hh-rlhf)
+- [Open Assistant (multilingual)](https://huggingface.co/datasets/OpenAssistant/oasst1)
+- [GPT-4 Generated Data (en&zh)](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM)
+
+</details>
+
+使用方法请参考 [data/README_zh.md](data/README_zh.md) 文件。

 部分数据集的使用需要确认，我们推荐使用下述命令登录您的 Hugging Face 账户。

@@ -129,7 +158,7 @@ huggingface-cli login
 - Python 3.8+ 和 PyTorch 1.13.1+
 - 🤗Transformers, Datasets, Accelerate, PEFT 和 TRL
 - sentencepiece, protobuf 和 tiktoken
- fire, jieba, rouge-chinese 和 nltk (用于评估及预测)
+- jieba, rouge-chinese 和 nltk (用于评估及预测)
 - gradio 和 matplotlib (用于网页端交互)
 - uvicorn, fastapi 和 sse-starlette (用于 API)

@@ -139,10 +168,10 @@ huggingface-cli login

 ### 数据准备（可跳过）

-关于数据集文件的格式，请参考 `data/example_dataset` 文件夹的内容。构建自定义数据集时，既可以使用单个 `.json` 文件，也可以使用一个[数据加载脚本](https://huggingface.co/docs/datasets/dataset_script)和多个文件。
+关于数据集文件的格式，请参考 [data/README_zh.md](data/README_zh.md) 的内容。构建自定义数据集时，既可以使用单个 `.json` 文件，也可以使用一个[数据加载脚本](https://huggingface.co/docs/datasets/dataset_script)和多个文件。

 > [!NOTE]
-> 使用自定义数据集时，请更新 `data/dataset_info.json` 文件，该文件的格式请参考 `data/README.md`。
+> 使用自定义数据集时，请更新 `data/dataset_info.json` 文件，该文件的格式请参考 `data/README_zh.md`。

 ### 环境搭建（可跳过）

@@ -160,17 +189,6 @@ pip install -r requirements.txt
 pip install https://github.com/jllllll/bitsandbytes-windows-webui/releases/download/wheels/bitsandbytes-0.39.1-py3-none-win_amd64.whl
 ```

-### 浏览器一体化界面
-
-```bash
-CUDA_VISIBLE_DEVICES=0 python src/train_web.py
-```
-
-我们**极力推荐**新手使用浏览器一体化界面，因为它还可以不依赖 GPU 环境自动生成在 GPU 上运行的命令行脚本。
-
-> [!WARNING]
-> 目前网页 UI 仅支持**单卡训练**。
-
 ### 单 GPU 训练

 > [!IMPORTANT]
@@ -376,8 +394,7 @@ python src/export_model.py \
    --template default \
    --finetuning_type lora \
    --checkpoint_dir path_to_checkpoint \
-    --output_dir path_to_export \
-    --fp16
+    --export_dir path_to_export
 ```

 ### API 服务
@@ -448,11 +465,18 @@ CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
 > [!NOTE]
 > 我们建议在量化模型的预测中使用 `--per_device_eval_batch_size=1` 和 `--max_target_length 128`。

+## 使用了 LLaMA Factory 的项目
+
+- **[StarWhisper](https://github.com/Yu-Yang-Li/StarWhisper)**: 天文大模型 StarWhisper，基于 ChatGLM2-6B 和 Qwen-14B 在天文数据上微调而得。
+- **[DISC-LawLLM](https://github.com/FudanDISC/DISC-LawLLM)**: 中文法律领域大模型 DISC-LawLLM，基于 Baichuan-13B 微调而得，具有法律推理和知识检索能力。
+- **[Sunsimiao](https://github.com/thomas-yanxin/Sunsimiao)**: 孙思邈中文医疗大模型 Sumsimiao，基于 Baichuan-7B 和 ChatGLM-6B 在中文医疗数据上微调而得。
+- **[CareGPT](https://github.com/WangRongsheng/CareGPT)**: 医疗大模型项目 CareGPT，基于 LLaMA2-7B 和 Baichuan-13B 在中文医疗数据上微调而得。
+
 ## 协议

 本仓库的代码依照 [Apache-2.0](LICENSE) 协议开源。

-使用模型权重时，请遵循对应的模型协议：[LLaMA](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md) / [LLaMA-2](https://ai.meta.com/llama/license/) / [BLOOM](https://huggingface.co/spaces/bigscience/license) / [Falcon](LICENSE) / [Baichuan](https://huggingface.co/baichuan-inc/baichuan-7B/resolve/main/baichuan-7B%20%E6%A8%A1%E5%9E%8B%E8%AE%B8%E5%8F%AF%E5%8D%8F%E8%AE%AE.pdf) / [Baichuan2](https://huggingface.co/baichuan-inc/Baichuan2-7B-Base/resolve/main/Baichuan%202%E6%A8%A1%E5%9E%8B%E7%A4%BE%E5%8C%BA%E8%AE%B8%E5%8F%AF%E5%8D%8F%E8%AE%AE.pdf) / [InternLM](https://github.com/InternLM/InternLM#open-source-license) / [Qwen](https://huggingface.co/Qwen/Qwen-7B-Chat/blob/main/LICENSE) / [XVERSE](https://github.com/xverse-ai/XVERSE-13B/blob/main/MODEL_LICENSE.pdf) / [ChatGLM2](https://github.com/THUDM/ChatGLM2-6B/blob/main/MODEL_LICENSE) / [Phi-1.5](https://huggingface.co/microsoft/phi-1_5/resolve/main/Research%20License.docx)
+使用模型权重时，请遵循对应的模型协议：[Baichuan](https://huggingface.co/baichuan-inc/Baichuan-13B-Base/resolve/main/Community%20License%20for%20Baichuan-13B%20Model.pdf) / [Baichuan2](https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat/resolve/main/Community%20License%20for%20Baichuan2%20Model.pdf) / [BLOOM](https://huggingface.co/spaces/bigscience/license) / [ChatGLM3](https://github.com/THUDM/ChatGLM3/blob/main/MODEL_LICENSE) / [Falcon](https://huggingface.co/tiiuae/falcon-180B/blob/main/LICENSE.txt) / [InternLM](https://github.com/InternLM/InternLM#license) / [LLaMA](https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md) / [LLaMA-2](https://ai.meta.com/llama/license/) / [Mistral](LICENSE) / [Phi-1.5](https://huggingface.co/microsoft/phi-1_5/resolve/main/Research%20License.docx) / [Qwen](https://github.com/QwenLM/Qwen/blob/main/LICENSE) / [XVERSE](https://github.com/xverse-ai/XVERSE-13B/blob/main/MODEL_LICENSE.pdf)

 ## 引用

--- a/data/README.md
+++ b/data/README.md
@@ -2,31 +2,106 @@ If you are using a custom dataset, please provide your dataset definition in the

 ```json
 "dataset_name": {
-  "hf_hub_url": "the name of the dataset repository on the HuggingFace hub. (if specified, ignore below 3 arguments)",
+  "hf_hub_url": "the name of the dataset repository on the Hugging Face hub. (if specified, ignore below 3 arguments)",
  "script_url": "the name of the directory containing a dataset loading script. (if specified, ignore below 2 arguments)",
  "file_name": "the name of the dataset file in the this directory. (required if above are not specified)",
-  "file_sha1": "the SHA-1 hash value of the dataset file. (optional)",
-  "ranking": "whether the examples contains ranked responses or not. (default: false)",
+  "file_sha1": "the SHA-1 hash value of the dataset file. (optional, does not affect training)",
+  "subset": "the name of the subset. (optional, default: None)",
+  "ranking": "whether the dataset is a preference dataset or not. (default: false)",
+  "formatting": "the format of the dataset. (optional, default: alpaca, can be chosen from {alpaca, sharegpt})",
  "columns": {
-    "prompt": "the name of the column in the datasets containing the prompts. (default: instruction)",
-    "query": "the name of the column in the datasets containing the queries. (default: input)",
-    "response": "the name of the column in the datasets containing the responses. (default: output)",
-    "history": "the name of the column in the datasets containing the history of chat. (default: None)"
+    "prompt": "the column name in the dataset containing the prompts. (default: instruction, for alpaca)",
+    "query": "the column name in the dataset containing the queries. (default: input, for alpaca)",
+    "response": "the column name in the dataset containing the responses. (default: output, for alpaca)",
+    "history": "the column name in the dataset containing the histories. (default: None, for alpaca)",
+    "messages": "the column name in the dataset containing the messages. (default: conversations, for sharegpt)",
+    "role": "the key in the message represents the identity. (default: from, for sharegpt)",
+    "content": "the key in the message represents the content. (default: value, for sharegpt)"
  }
 }
 ```

-where the `prompt` and `response` columns should contain non-empty values. The `query` column will be concatenated with the `prompt` column and used as input for the model. The `history` column should contain a list where each element is a string tuple representing a query-response pair.
+Given above, you can use the custom dataset via specifying `--dataset dataset_name`.

-For datasets used in reward modeling or DPO training, the `response` column should be a string list, with the preferred answers appearing first, for example:
+Currently we support dataset in **alpaca** or **sharegpt** format, the dataset in alpaca format should follow the below format:
+
+```json
+[
+  {
+    "instruction": "user instruction (required)",
+    "input": "user input (optional)",
+    "output": "model response (required)",
+    "history": [
+      ["user instruction in the first round (optional)", "model response in the first round (optional)"],
+      ["user instruction in the second round (optional)", "model response in the second round (optional)"]
+    ]
+  }
+]
+```
+
+Regarding the above dataset, the `columns` in `dataset_info.json` should be:
+
+```json
+"dataset_name": {
+  "columns": {
+    "prompt": "instruction",
+    "query": "input",
+    "response": "output",
+    "history": "history"
+  }
+}
+```
+
+where the `prompt` and `response` columns should contain non-empty values, represent instruction and response respectively. The `query` column will be concatenated with the `prompt` column and used as input for the model.
+
+The `history` column is a list consisting string tuples representing query-response pairs in history. Note that the responses **in each round will be used for training**.
+
+For the pre-training datasets, only the `prompt` column will be used for training.
+
+For the preference datasets, the `response` column should be a string list whose length is 2, with the preferred answers appearing first, for example:

 ```json
 {
-  "instruction": "Question",
-  "input": "",
+  "instruction": "user instruction",
+  "input": "user input",
  "output": [
-    "Chosen answer",
-    "Rejected answer"
+    "chosen answer",
+    "rejected answer"
  ]
 }
 ```
+
+The dataset in sharegpt format should follow the below format:
+
+```json
+[
+  {
+    "conversations": [
+      {
+        "from": "human",
+        "value": "user instruction"
+      },
+      {
+        "from": "gpt",
+        "value": "model response"
+      }
+    ]
+  }
+]
+```
+
+Regarding the above dataset, the `columns` in `dataset_info.json` should be:
+
+```json
+"dataset_name": {
+  "columns": {
+    "messages": "conversations",
+    "role": "from",
+    "content": "value"
+  }
+}
+```
+
+where the `messages` column should be a list whose length is even, and follow the `u/a/u/a/u/a` order.
+
+Pre-training datasets and preference datasets are incompatible with the sharegpt format yet.
--- a/data/README_zh.md
+++ b/data/README_zh.md
@@ -1,32 +1,107 @@
-如果您使用自定义数据集，请务必在 `dataset_info.json` 文件中以如下格式提供您的数据集定义。
+如果您使用自定义数据集，请务必在 `dataset_info.json` 文件中按照以下格式提供数据集定义。

 ```json
 "数据集名称": {
-  "hf_hub_url": "HuggingFace上的项目地址（若指定，则忽略下列三个参数）",
+  "hf_hub_url": "Hugging Face 上的项目地址（若指定，则忽略下列三个参数）",
  "script_url": "包含数据加载脚本的本地文件夹名称（若指定，则忽略下列两个参数）",
  "file_name": "该目录下数据集文件的名称（若上述参数未指定，则此项必需）",
-  "file_sha1": "数据集文件的SHA-1哈希值（可选）",
-  "ranking": "数据集是否包含排序后的回答（默认：false）",
+  "file_sha1": "数据集文件的SHA-1哈希值（可选，留空不影响训练）",
+  "subset": "数据集子集的名称（可选，默认：None）",
+  "ranking": "是否为偏好数据集（可选，默认：False）",
+  "formatting": "数据集格式（可选，默认：alpaca，可以为 alpaca 或 sharegpt）",
  "columns": {
-    "prompt": "数据集代表提示词的表头名称（默认：instruction）",
-    "query": "数据集代表请求的表头名称（默认：input）",
-    "response": "数据集代表回答的表头名称（默认：output）",
-    "history": "数据集代表历史对话的表头名称（默认：None）"
+    "prompt": "数据集代表提示词的表头名称（默认：instruction，用于 alpaca 格式）",
+    "query": "数据集代表请求的表头名称（默认：input，用于 alpaca 格式）",
+    "response": "数据集代表回答的表头名称（默认：output，用于 alpaca 格式）",
+    "history": "数据集代表历史对话的表头名称（默认：None，用于 alpaca 格式）",
+    "messages": "数据集代表消息列表的表头名称（默认：conversations，用于 sharegpt 格式）",
+    "role": "消息中代表发送者身份的键名（默认：from，用于 sharegpt 格式）",
+    "content": "消息中代表文本内容的键名（默认：value，用于 sharegpt 格式）"
  }
 }
 ```

-其中 `prompt` 和 `response` 列应当是非空的字符串。`query` 列的内容将会和 `prompt` 列拼接作为模型输入。`history` 列应当是一个列表，其中每个元素是一个字符串二元组，分别代表用户请求和模型答复。
+添加后可通过指定 `--dataset 数据集名称` 参数使用自定义数据集。

-对于训练奖励模型或 DPO 训练的数据集，`response` 列应当是一个字符串列表，排在前面的代表更优的答案，例如：
+该项目目前支持两种格式的数据集：**alpaca** 和 **sharegpt**，其中 alpaca 格式的数据集按照以下方式组织：
+
+```json
+[
+  {
+    "instruction": "用户指令（必填）",
+    "input": "用户输入（选填）",
+    "output": "模型回答（必填）",
+    "history": [
+      ["第一轮指令（选填）", "第一轮回答（选填）"],
+      ["第二轮指令（选填）", "第二轮回答（选填）"]
+    ]
+  }
+]
+```
+
+对于上述格式的数据，`dataset_info.json` 中的 `columns` 应为：
+
+```json
+"数据集名称": {
+  "columns": {
+    "prompt": "instruction",
+    "query": "input",
+    "response": "output",
+    "history": "history"
+  }
+}
+```
+
+其中 `prompt` 和 `response` 列应当是非空的字符串，分别代表用户指令和模型回答。`query` 列的内容将会和 `prompt` 列拼接作为模型输入。
+
+`history` 列是由多个字符串二元组构成的列表，分别代表历史消息中每轮的指令和回答。注意每轮的模型回答**均会被用于训练**。
+
+对于预训练数据集，仅 `prompt` 列中的内容会用于模型训练。
+
+对于偏好数据集，`response` 列应当是一个长度为 2 的字符串列表，排在前面的代表更优的回答，例如：

 ```json
 {
-  "instruction": "Question",
-  "input": "",
+  "instruction": "用户指令",
+  "input": "用户输入",
  "output": [
-    "Chosen answer",
-    "Rejected answer"
+    "优质回答",
+    "劣质回答"
  ]
 }
 ```
+
+而 sharegpt 格式的数据集按照以下方式组织：
+
+```json
+[
+  {
+    "conversations": [
+      {
+        "from": "human",
+        "value": "用户指令"
+      },
+      {
+        "from": "gpt",
+        "value": "模型回答"
+      }
+    ]
+  }
+]
+```
+
+对于上述格式的数据，`dataset_info.json` 中的 `columns` 应为：
+
+```json
+"数据集名称": {
+  "columns": {
+    "messages": "conversations",
+    "role": "from",
+    "content": "value"
+  }
+}
+```
+
+其中 `messages` 列必须为偶数长度的列表，且符合 `用户/模型/用户/模型/用户/模型` 的顺序。
+
+预训练数据集和偏好数据集尚不支持 sharegpt 格式。
--- a/data/belle_multiturn/belle_multiturn.py
+++ b/data/belle_multiturn/belle_multiturn.py
@@ -1,6 +1,5 @@
 import json
 import datasets
-from typing import Any, Dict, List


 _DESCRIPTION = "BELLE multiturn chat dataset."
@@ -23,7 +22,7 @@ class BelleMultiturn(datasets.GeneratorBasedBuilder):

    VERSION = datasets.Version("0.0.0")

-    def _info(self) -> datasets.DatasetInfo:
+    def _info(self):
        features = datasets.Features({
            "instruction": datasets.Value("string"),
            "output": datasets.Value("string"),
@@ -37,7 +36,7 @@ class BelleMultiturn(datasets.GeneratorBasedBuilder):
            citation=_CITATION
        )

-    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
+    def _split_generators(self, dl_manager: datasets.DownloadManager):
        file_path = dl_manager.download(_URL)
        return [
            datasets.SplitGenerator(
@@ -48,7 +47,7 @@ class BelleMultiturn(datasets.GeneratorBasedBuilder):
            )
        ]

-    def _generate_examples(self, filepath: str) -> Dict[int, Dict[str, Any]]: # generate multi-turn chat with history
+    def _generate_examples(self, filepath: str):
        with open(filepath, "r", encoding="utf-8") as f:
            for key, row in enumerate(f):
                data = json.loads(row)
--- a/data/example_dataset/example_dataset.py
+++ b/data/example_dataset/example_dataset.py
@@ -3,7 +3,7 @@ import datasets
 from typing import Any, Dict, List


-_DESCRIPTION = "An example of dataset for LLaMA."
+_DESCRIPTION = "An example of dataset."
 _CITATION = ""
 _HOMEPAGE = ""
 _LICENSE = ""
--- a/data/hh_rlhf_en/hh_rlhf_en.py
+++ b/data/hh_rlhf_en/hh_rlhf_en.py
@@ -1,9 +1,9 @@
 import json
 import datasets
-from typing import Any, Dict, List
+from typing import List


-_DESCRIPTION = "Human preference data about helpfulness and harmlessness for ChatGLM."
+_DESCRIPTION = "Human preference data about helpfulness and harmlessness."
 _CITATION = ""
 _HOMEPAGE = "https://huggingface.co/datasets/Anthropic/hh-rlhf"
 _LICENSE = "mit"
@@ -42,7 +42,7 @@ class HhRlhfEn(datasets.GeneratorBasedBuilder):
            citation=_CITATION
        )

-    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
+    def _split_generators(self, dl_manager: datasets.DownloadManager):
        file_path = dl_manager.download_and_extract(_URLS)
        return [
            datasets.SplitGenerator(
@@ -59,7 +59,7 @@ class HhRlhfEn(datasets.GeneratorBasedBuilder):
            )
        ]

-    def _generate_examples(self, filepaths: List[str]) -> Dict[int, Dict[str, Any]]: # generate multi-turn chat for ChatGLM
+    def _generate_examples(self, filepaths: List[str]):
        key = 0
        for filepath in filepaths:
            with open(filepath, "r", encoding="utf-8") as f:
--- a/data/ultra_chat/ultra_chat.py
+++ b/data/ultra_chat/ultra_chat.py
@@ -1,6 +1,6 @@
 import json
 import datasets
-from typing import Any, Dict, List
+from typing import List


 _DESCRIPTION = "UltraChat: Large-scale, Informative, and Diverse Multi-round Dialogue Data."
@@ -21,15 +21,13 @@ _LICENSE = "cc-by-nc-4.0"
 _BASE_DATA_URL = "https://huggingface.co/datasets/stingning/ultrachat/resolve/main/train_{idx}.jsonl"


-class BelleMultiturn(datasets.GeneratorBasedBuilder):
+class UltraChat(datasets.GeneratorBasedBuilder):

    VERSION = datasets.Version("0.0.0")

-    def _info(self) -> datasets.DatasetInfo:
+    def _info(self):
        features = datasets.Features({
-            "instruction": datasets.Value("string"),
-            "output": datasets.Value("string"),
-            "history": datasets.Sequence(datasets.Sequence(datasets.Value("string")))
+            "conversations": [{"from": datasets.Value("string"), "value": datasets.Value("string")}]
        })
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
@@ -39,8 +37,8 @@ class BelleMultiturn(datasets.GeneratorBasedBuilder):
            citation=_CITATION
        )

-    def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
-        file_paths = [dl_manager.download(_BASE_DATA_URL.format(idx=idx)) for idx in range(9)] # multiple shards
+    def _split_generators(self, dl_manager: datasets.DownloadManager):
+        file_paths = [dl_manager.download(_BASE_DATA_URL.format(idx=idx)) for idx in range(10)] # multiple shards
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
@@ -50,7 +48,7 @@ class BelleMultiturn(datasets.GeneratorBasedBuilder):
            )
        ]

-    def _generate_examples(self, filepaths: List[str]) -> Dict[int, Dict[str, Any]]: # generate multi-turn chat for ChatGLM
+    def _generate_examples(self, filepaths: List[str]):
        for filepath in filepaths:
            with open(filepath, "r", encoding="utf-8") as f:
                for row in f:
@@ -58,19 +56,16 @@ class BelleMultiturn(datasets.GeneratorBasedBuilder):
                        data = json.loads(row)
                    except:
                        continue
-                    key = data["id"]
-                    content = data["data"]
+                    key: int = data["id"]
+                    content: List[str] = data["data"]
                    if len(content) % 2 == 1:
                        content.pop(-1)
                    if len(content) < 2:
                        continue
-
-                    query = content[-2]
-                    response = content[-1]
-                    history = [[content[2*i], content[2*i+1]] for i in range(len(content) // 2 - 1)]
-
+                    conversations = [{
+                        "from": "human" if i % 2 == 0 else "gpt",
+                        "value": content[i]
+                    } for i in range(len(content))]
                    yield key, {
-                        "instruction": query,
-                        "output": response,
-                        "history": history
+                        "conversations": conversations
                    }
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,20 +1,19 @@
 torch>=1.13.1
-transformers>=4.31.0
-datasets>=2.12.0
+transformers>=4.31.0,<4.35.0
+datasets>=2.14.0
 accelerate>=0.21.0
-peft>=0.4.0
-trl>=0.7.1
+peft>=0.6.0
+trl>=0.7.4
+gradio>=3.38.0,<4.0.0
 scipy
 sentencepiece
 protobuf
 tiktoken
-fire
 jieba
 rouge-chinese
 nltk
-gradio==3.38.0
 uvicorn
-pydantic==1.10.11
-fastapi==0.95.1
+pydantic
+fastapi
 sse-starlette
 matplotlib
--- a/src/api_demo.py
+++ b/src/api_demo.py
@@ -6,8 +6,8 @@ from llmtuner import ChatModel, create_app
 def main():
    chat_model = ChatModel()
    app = create_app(chat_model)
-    uvicorn.run(app, host="0.0.0.0", port=8000, workers=1)
    print("Visit http://localhost:8000/docs for API document.")
+    uvicorn.run(app, host="0.0.0.0", port=8000, workers=1)


 if __name__ == "__main__":
--- a/src/cli_demo.py
+++ b/src/cli_demo.py
@@ -1,3 +1,4 @@
+import readline
 from llmtuner import ChatModel


--- a/src/evaluate.py
+++ b/src/evaluate.py
@@ -1,185 +1,10 @@
-# coding=utf-8
-# Evaluates the performance of pre-trained models.
-# Usage: python evaluate.py --model_name_or_path path_to_model --checkpoint_dir path_to_ckpt --template vanilla
-#                           --task ceval --split validation --lang zh --n_shot 5 --batch_size 4 --save_name result
-# Inspired by: https://github.com/hendrycks/test/blob/master/evaluate_flan.py
-
-import os
-import fire
-import json
-import torch
-import numpy as np
-from collections import Counter
-from datasets import load_dataset
-from dataclasses import dataclass
-from tqdm import tqdm, trange
-from typing import TYPE_CHECKING, Dict, List, Literal, Optional, Tuple
-
-from llmtuner import ChatModel
-
-if TYPE_CHECKING:
-    from datasets import Dataset
+from llmtuner import Evaluator


-choices = ["A", "B", "C", "D"]
-
-
-@dataclass
-class EvalTemplate:
-
-    system: str
-    choice: str
-    answer: str
-    prefix: str
-
-    def parse_example(
-        self,
-        example: Dict[str, str]
-    ) -> Tuple[str, str]:
-        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: "Dataset",
-        subject_name: str,
-        use_history: bool
-    ) -> Tuple[str, str, List[Tuple[str, str]]]:
-        query, resp = self.parse_example(target_data)
-        history = [self.parse_example(support_set[k]) for k in range(len(support_set))]
-
-        if len(history):
-            temp = history.pop(0)
-            history.insert(0, (self.system.format(subject=subject_name) + temp[0], temp[1]))
-        else:
-            query = self.system.format(subject=subject_name) + query
-
-        if not use_history:
-            query = "\n\n".join(["".join(item) for item in history] + [query])
-            history = []
-        return query.strip(), resp, history
-
-
-eval_templates = {
-    "en": EvalTemplate(
-        system="The following are multiple choice questions (with answers) about {subject}.\n\n",
-        choice="\n{choice}. {content}",
-        answer="\nAnswer: ",
-        prefix=" "
-    ),
-    "zh": EvalTemplate(
-        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：",
-        prefix="\n"
-    )
-}
-
-
-@torch.inference_mode()
-def batch_inference(
-    chat_model: ChatModel,
-    batch_input: Dict[str, torch.Tensor],
-    prefix_char: str
-) -> List[str]:
-    logits = chat_model.model(**batch_input).logits
-    probs = torch.nn.functional.softmax(
-        torch.stack(
-            [
-                logits[:, -1, chat_model.tokenizer.encode(prefix_char + choice, add_special_tokens=False)[-1]]
-                for choice in choices
-            ],
-            dim=-1
-        ),
-        dim=-1
-    ).detach()
-    return [chr(ord("A") + offset.item()) for offset in torch.argmax(probs, dim=-1)]
-
-
-def evaluate(
-    model_name_or_path: str,
-    finetuning_type: Optional[str] = "lora",
-    checkpoint_dir: Optional[str] = None,
-    template: Optional[str] = "vanilla",
-    task: Optional[str] = "ceval",
-    dataset_dir: Optional[str] = "evaluation",
-    split: Optional[Literal["validation", "test"]] = "validation",
-    lang: Optional[Literal["zh", "en"]] = "zh",
-    n_shot: Optional[int] = 5,
-    n_avg: Optional[int] = 1,
-    batch_size: Optional[int] = 4,
-    save_name: Optional[str] = None
-):
-    with open(os.path.join(dataset_dir, task, "mapping.json"), "r", encoding="utf-8") as f:
-        categorys: Dict[str, Dict[str, str]] = json.load(f)
-
-    chat_model = ChatModel(dict(
-        model_name_or_path=model_name_or_path,
-        finetuning_type=finetuning_type,
-        checkpoint_dir=checkpoint_dir,
-        template=template
-    ))
-    eval_template = eval_templates[lang]
-    assert chat_model.tokenizer.padding_side == "left", "only left-padded tensor can be accepted."
-
-    category_corrects: Dict[str, np.ndarray] = {
-        subj: np.array([], dtype="bool") for subj in ["Average", "STEM", "Social Sciences", "Humanities", "Other"]
-    }
-    pbar = tqdm(categorys.keys(), desc="Processing subjects", position=0)
-    results = {}
-    for subject in pbar:
-        dataset = load_dataset(os.path.join(dataset_dir, task), subject)
-        labels, answers, all_outputs = [], [], []
-        for epoch in range(n_avg):
-            pbar.set_postfix_str("{} Trial: {}".format(categorys[subject]["name"], epoch))
-            inputs, outputs = [], []
-            for i in trange(len(dataset[split]), desc="Formatting batches", position=1, leave=False):
-                support_set = dataset["train"].shuffle().select(range(min(n_shot, len(dataset["train"]))))
-                query, resp, history = eval_template.format_example(
-                    target_data=dataset[split][i],
-                    support_set=support_set,
-                    subject_name=categorys[subject]["name"],
-                    use_history=chat_model.template.use_history
-                )
-                input_ids, _ = chat_model.template.encode_oneturn(
-                    tokenizer=chat_model.tokenizer, query=query, resp=resp, history=history
-                )
-                inputs.append({"input_ids": input_ids, "attention_mask": [1] * len(input_ids)})
-                if epoch == 0:
-                    labels.append(resp)
-
-            for i in trange(0, len(inputs), batch_size, desc="Predicting batches", position=1, leave=False):
-                batch_input = chat_model.tokenizer.pad(
-                    inputs[i : i + batch_size], return_attention_mask=True, return_tensors="pt"
-                ).to(chat_model.model.device)
-                preds = batch_inference(chat_model, batch_input, eval_template.prefix)
-                outputs += preds
-            all_outputs.append(outputs)
-
-        for i in range(len(all_outputs[0])):
-            count = Counter([all_outputs[epoch][i] for epoch in range(n_avg)])
-            answers.append(count.most_common(1)[0][0])
-
-        corrects = (np.array(answers) == 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): answers[i] for i in range(len(answers))}
-
-    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 save_name is not None:
-        with open(save_name + ".json", "w", encoding="utf-8", newline="\n") as f:
-            json.dump(results, f, indent=2)
-
-        with open(save_name + ".log", "w", encoding="utf-8", newline="\n") as f:
-            f.write(score_info)
+def main():
+    evaluator = Evaluator()
+    evaluator.eval()


 if __name__ == "__main__":
-    fire.Fire(evaluate)
+    main()
--- a/src/llmtuner/init.py
+++ b/src/llmtuner/init.py
@@ -1,9 +1,10 @@
-# Level: api, webui > chat > tuner > dsets > extras, hparams
+# Level: api, webui > chat, eval > tuner > dsets > extras, hparams

 from llmtuner.api import create_app
 from llmtuner.chat import ChatModel
+from llmtuner.eval import Evaluator
 from llmtuner.tuner import export_model, run_exp
 from llmtuner.webui import create_ui, create_web_demo


-__version__ = "0.2.0"
+__version__ = "0.2.2"
--- a/src/llmtuner/api/app.py
+++ b/src/llmtuner/api/app.py
@@ -1,9 +1,11 @@
+import json
 import uvicorn
-from fastapi import FastAPI, HTTPException
+from fastapi import FastAPI, HTTPException, status
 from fastapi.middleware.cors import CORSMiddleware
 from contextlib import asynccontextmanager
 from sse_starlette import EventSourceResponse
 from typing import List, Tuple
+from pydantic import BaseModel

 from llmtuner.extras.misc import torch_gc
 from llmtuner.chat import ChatModel
@@ -29,6 +31,13 @@ async def lifespan(app: FastAPI): # collects GPU memory
    torch_gc()


+def to_json(data: BaseModel) -> str:
+    try: # pydantic v2
+        return json.dumps(data.model_dump(exclude_unset=True), ensure_ascii=False)
+    except: # pydantic v1
+        return data.json(exclude_unset=True, ensure_ascii=False)
+
+
 def create_app(chat_model: ChatModel) -> FastAPI:
    app = FastAPI(lifespan=lifespan)

@@ -45,10 +54,10 @@ def create_app(chat_model: ChatModel) -> FastAPI:
        model_card = ModelCard(id="gpt-3.5-turbo")
        return ModelList(data=[model_card])

-    @app.post("/v1/chat/completions", response_model=ChatCompletionResponse)
+    @app.post("/v1/chat/completions", response_model=ChatCompletionResponse, status_code=status.HTTP_200_OK)
    async def create_chat_completion(request: ChatCompletionRequest):
        if len(request.messages) < 1 or request.messages[-1].role != Role.USER:
-            raise HTTPException(status_code=400, detail="Invalid request")
+            raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Invalid request")

        query = request.messages[-1].content
        prev_messages = request.messages[:-1]
@@ -62,6 +71,8 @@ def create_app(chat_model: ChatModel) -> FastAPI:
            for i in range(0, len(prev_messages), 2):
                if prev_messages[i].role == Role.USER and prev_messages[i+1].role == Role.ASSISTANT:
                    history.append([prev_messages[i].content, prev_messages[i+1].content])
+                else:
+                    raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="Only supports u/a/u/a/u...")

        if request.stream:
            generate = predict(query, history, system, request)
@@ -72,7 +83,8 @@ def create_app(chat_model: ChatModel) -> FastAPI:
            do_sample=request.do_sample,
            temperature=request.temperature,
            top_p=request.top_p,
-            max_new_tokens=request.max_tokens
+            max_new_tokens=request.max_tokens,
+            num_return_sequences=request.n
        )

        usage = ChatCompletionResponseUsage(
@@ -81,13 +93,13 @@ def create_app(chat_model: ChatModel) -> FastAPI:
            total_tokens=prompt_length+response_length
        )

-        choice_data = ChatCompletionResponseChoice(
-            index=0,
-            message=ChatMessage(role=Role.ASSISTANT, content=response),
+        choices = [ChatCompletionResponseChoice(
+            index=i,
+            message=ChatMessage(role=Role.ASSISTANT, content=choice),
            finish_reason=Finish.STOP
-        )
+        ) for i, choice in enumerate(response)]

-        return ChatCompletionResponse(model=request.model, choices=[choice_data], usage=usage)
+        return ChatCompletionResponse(model=request.model, choices=choices, usage=usage)

    async def predict(query: str, history: List[Tuple[str, str]], system: str, request: ChatCompletionRequest):
        choice_data = ChatCompletionResponseStreamChoice(
@@ -96,7 +108,7 @@ def create_app(chat_model: ChatModel) -> FastAPI:
            finish_reason=None
        )
        chunk = ChatCompletionStreamResponse(model=request.model, choices=[choice_data])
-        yield chunk.json(exclude_unset=True, ensure_ascii=False)
+        yield to_json(chunk)

        for new_text in chat_model.stream_chat(
            query, history, system,
@@ -114,7 +126,7 @@ def create_app(chat_model: ChatModel) -> FastAPI:
                finish_reason=None
            )
            chunk = ChatCompletionStreamResponse(model=request.model, choices=[choice_data])
-            yield chunk.json(exclude_unset=True, ensure_ascii=False)
+            yield to_json(chunk)

        choice_data = ChatCompletionResponseStreamChoice(
            index=0,
@@ -122,7 +134,7 @@ def create_app(chat_model: ChatModel) -> FastAPI:
            finish_reason=Finish.STOP
        )
        chunk = ChatCompletionStreamResponse(model=request.model, choices=[choice_data])
-        yield chunk.json(exclude_unset=True, ensure_ascii=False)
+        yield to_json(chunk)
        yield "[DONE]"

    return app
--- a/src/llmtuner/api/protocol.py
+++ b/src/llmtuner/api/protocol.py
@@ -20,9 +20,6 @@ class ModelCard(BaseModel):
    object: Optional[str] = "model"
    created: Optional[int] = Field(default_factory=lambda: int(time.time()))
    owned_by: Optional[str] = "owner"
-    root: Optional[str] = None
-    parent: Optional[str] = None
-    permission: Optional[list] = []


 class ModelList(BaseModel):
--- a/src/llmtuner/chat/stream_chat.py
+++ b/src/llmtuner/chat/stream_chat.py
@@ -26,17 +26,17 @@ class ChatModel:
        **input_kwargs
    ) -> Tuple[Dict[str, Any], int]:
        system = system or self.system_prompt
-
        prompt, _ = self.template.encode_oneturn(
            tokenizer=self.tokenizer, query=query, resp="", history=history, system=system
        )
+        prompt_length = len(prompt)
        input_ids = torch.tensor([prompt], device=self.model.device)
-        prompt_length = len(input_ids[0])

        do_sample = input_kwargs.pop("do_sample", None)
        temperature = input_kwargs.pop("temperature", None)
        top_p = input_kwargs.pop("top_p", None)
        top_k = input_kwargs.pop("top_k", None)
+        num_return_sequences = input_kwargs.pop("num_return_sequences", None)
        repetition_penalty = input_kwargs.pop("repetition_penalty", None)
        max_length = input_kwargs.pop("max_length", None)
        max_new_tokens = input_kwargs.pop("max_new_tokens", None)
@@ -47,11 +47,15 @@ class ChatModel:
            temperature=temperature or generating_args["temperature"],
            top_p=top_p or generating_args["top_p"],
            top_k=top_k or generating_args["top_k"],
+            num_return_sequences=num_return_sequences or 1,
            repetition_penalty=repetition_penalty or generating_args["repetition_penalty"],
            eos_token_id=[self.tokenizer.eos_token_id] + self.tokenizer.additional_special_tokens_ids,
            pad_token_id=self.tokenizer.pad_token_id
        ))

+        if isinstance(num_return_sequences, int) and num_return_sequences > 1:
+            generating_args["do_sample"] = True
+
        if max_length:
            generating_args.pop("max_new_tokens", None)
            generating_args["max_length"] = max_length
@@ -75,12 +79,16 @@ class ChatModel:
        history: Optional[List[Tuple[str, str]]] = None,
        system: Optional[str] = None,
        **input_kwargs
-    ) -> Tuple[str, Tuple[int, int]]:
+    ) -> Tuple[List[str], Tuple[int, int]]:
        gen_kwargs, prompt_length = self.process_args(query, history, system, **input_kwargs)
-        generation_output = self.model.generate(**gen_kwargs)
-        outputs = generation_output.tolist()[0][prompt_length:]
-        response = self.tokenizer.decode(outputs, skip_special_tokens=True)
-        response_length = len(outputs)
+        generate_output = self.model.generate(**gen_kwargs)
+        response_ids = generate_output[:, prompt_length:]
+        response = self.tokenizer.batch_decode(response_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True)
+        response_length = 0
+        for i in range(len(response_ids)):
+            eos_index = (response_ids[i] == self.tokenizer.eos_token_id).nonzero()
+            response_length += eos_index[0].item() if len(eos_index) else len(response_ids[i])
+
        return response, (prompt_length, response_length)

    @torch.inference_mode()
--- a/src/llmtuner/dsets/loader.py
+++ b/src/llmtuner/dsets/loader.py
@@ -1,5 +1,5 @@
 import os
-from typing import TYPE_CHECKING, List, Union
+from typing import TYPE_CHECKING, Any, Dict, List, Union

 from datasets import concatenate_datasets, interleave_datasets, load_dataset

@@ -26,22 +26,23 @@ def get_dataset(

        if dataset_attr.load_from == "hf_hub":
            data_path = dataset_attr.dataset_name
+            data_name = dataset_attr.subset
            data_files = None
        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_files = None
        elif dataset_attr.load_from == "file":
-            data_path = None
+            data_path, data_name = None, None
            data_files: List[str] = []
-
-            if os.path.isdir(os.path.join(data_args.dataset_dir, dataset_attr.dataset_name)): # directory
+            if os.path.isdir(os.path.join(data_args.dataset_dir, dataset_attr.dataset_name)): # is directory
                for file_name in os.listdir(os.path.join(data_args.dataset_dir, dataset_attr.dataset_name)):
                    data_files.append(os.path.join(data_args.dataset_dir, dataset_attr.dataset_name, file_name))
                    if data_path is None:
                        data_path = EXT2TYPE.get(file_name.split(".")[-1], None)
                    else:
-                        assert data_path == EXT2TYPE.get(file_name.split(".")[-1], None), "file type does not match."
-            elif os.path.isfile(os.path.join(data_args.dataset_dir, dataset_attr.dataset_name)): # single file
+                        assert data_path == EXT2TYPE.get(file_name.split(".")[-1], None), "file types are not identical."
+            elif os.path.isfile(os.path.join(data_args.dataset_dir, dataset_attr.dataset_name)): # is file
                data_files.append(os.path.join(data_args.dataset_dir, dataset_attr.dataset_name))
                data_path = EXT2TYPE.get(dataset_attr.dataset_name.split(".")[-1], None)
            else:
@@ -53,29 +54,75 @@ def get_dataset(
            raise NotImplementedError

        dataset = load_dataset(
-            data_path,
+            path=data_path,
+            name=data_name,
            data_files=data_files,
            split=data_args.split,
            cache_dir=model_args.cache_dir,
-            streaming=data_args.streaming,
-            use_auth_token=True if model_args.use_auth_token else None
+            token=model_args.hf_hub_token,
+            streaming=data_args.streaming
        )

-        if max_samples is not None:
-            max_samples_temp = min(len(dataset), max_samples)
-            dataset = dataset.select(range(max_samples_temp))
+        if max_samples is not None: # truncate dataset
+            dataset = dataset.select(range(min(len(dataset), max_samples)))

-        # TODO: adapt to the sharegpt format
+        def convert_format(examples: Dict[str, List[Any]]) -> Dict[str, List[Any]]:
+            # convert dataset from sharegpt format to alpaca format
+            outputs = {"prompt": [], "query": [], "response": [], "history": []}
+            for msg_list in examples[dataset_attr.messages]:
+                msg_list = msg_list[:len(msg_list) // 2 * 2] # should be multiples of 2
+                if len(msg_list) == 0:
+                    continue

-        for column_name in ["prompt", "query", "response", "history"]: # align datasets
-            if getattr(dataset_attr, column_name) and getattr(dataset_attr, column_name) != column_name:
-                dataset = dataset.rename_column(getattr(dataset_attr, column_name), column_name)
+                msg_pairs = []
+                user_role, assistant_role = None, None
+                for idx in range(0, len(msg_list), 2):
+                    if user_role is None and assistant_role is None:
+                        user_role = msg_list[idx][dataset_attr.role]
+                        assistant_role = msg_list[idx + 1][dataset_attr.role]
+                    else:
+                        if (
+                            msg_list[idx][dataset_attr.role] != user_role
+                            or msg_list[idx+1][dataset_attr.role] != assistant_role
+                        ):
+                            raise ValueError("Only accepts conversation in u/a/u/a/u/a order.")
+                    msg_pairs.append((msg_list[idx][dataset_attr.content], msg_list[idx + 1][dataset_attr.content]))
+
+                if len(msg_pairs) != 0:
+                    outputs["prompt"].append(msg_pairs[-1][0])
+                    outputs["query"].append("")
+                    outputs["response"].append(msg_pairs[-1][1])
+                    outputs["history"].append(msg_pairs[:-1])
+
+            return outputs
+
+        if dataset_attr.formatting == "sharegpt": # convert format
+            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),
+                    desc="Converting format of dataset"
+                )
+
+            dataset = dataset.map(
+                convert_format,
+                batched=True,
+                remove_columns=column_names,
+                **kwargs
+            )
+        else:
+            for column_name in ["prompt", "query", "response", "history"]: # align dataset
+                if getattr(dataset_attr, column_name) and getattr(dataset_attr, column_name) != column_name:
+                    dataset = dataset.rename_column(getattr(dataset_attr, column_name), column_name)

        if dataset_attr.system_prompt: # add system prompt
+            system_prompt = dataset_attr.system_prompt
            if data_args.streaming:
-                dataset = dataset.map(lambda _: {"system": dataset_attr.system_prompt})
+                dataset = dataset.map(lambda _: {"system": system_prompt})
            else:
-                dataset = dataset.add_column("system", [dataset_attr.system_prompt] * len(dataset))
+                dataset = dataset.add_column("system", [system_prompt] * len(dataset))

        all_datasets.append(dataset)

@@ -88,7 +135,11 @@ def get_dataset(
    elif data_args.mix_strategy.startswith("interleave"):
        if not data_args.streaming:
            logger.warning("We recommend using `mix_strategy=concat` in non-streaming mode.")
-        stopping_strategy = "first_exhausted" if data_args.mix_strategy.endswith("under") else "all_exhausted"
-        return interleave_datasets(all_datasets, data_args.interleave_probs, stopping_strategy=stopping_strategy)
+        return interleave_datasets(
+            datasets=all_datasets,
+            probabilities=data_args.interleave_probs,
+            seed=data_args.seed,
+            stopping_strategy="first_exhausted" if data_args.mix_strategy.endswith("under") else "all_exhausted"
+        )
    else:
        raise ValueError("Unknown mixing strategy.")
--- a/src/llmtuner/dsets/preprocess.py
+++ b/src/llmtuner/dsets/preprocess.py
@@ -1,8 +1,12 @@
+import os
 import tiktoken
-from typing import TYPE_CHECKING, Any, Dict, Generator, List, Literal, Union
 from itertools import chain
+from typing import TYPE_CHECKING, Any, Dict, Generator, List, Literal, Union
+
+from datasets import load_from_disk

 from llmtuner.extras.constants import IGNORE_INDEX
+from llmtuner.extras.logging import get_logger
 from llmtuner.extras.template import get_template_and_fix_tokenizer

 if TYPE_CHECKING:
@@ -12,6 +16,9 @@ if TYPE_CHECKING:
    from llmtuner.hparams import DataArguments


+logger = get_logger(__name__)
+
+
 def preprocess_dataset(
    dataset: Union["Dataset", "IterableDataset"],
    tokenizer: "PreTrainedTokenizer",
@@ -19,7 +26,6 @@ def preprocess_dataset(
    training_args: "Seq2SeqTrainingArguments",
    stage: Literal["pt", "sft", "rm", "ppo"]
 ) -> Union["Dataset", "IterableDataset"]:
-    column_names = list(next(iter(dataset)).keys())
    template = get_template_and_fix_tokenizer(data_args.template, tokenizer)

    if data_args.train_on_prompt and template.efficient_eos:
@@ -33,7 +39,7 @@ def preprocess_dataset(
            system = examples["system"][i] if "system" in examples else None
            yield query, response, history, system

-    def preprocess_pretrain_dataset(examples: Dict[str, List[Any]]) -> Dict[str, Any]:
+    def preprocess_pretrain_dataset(examples: Dict[str, List[Any]]) -> Dict[str, List[List[int]]]:
        # build grouped texts with format `X1 X2 X3 ...`
        if isinstance(getattr(tokenizer, "tokenizer", None), tiktoken.Encoding): # for tiktoken tokenizer (Qwen)
            kwargs = dict(allowed_special="all")
@@ -41,6 +47,7 @@ def preprocess_dataset(
            kwargs = dict(add_special_tokens=True)

        if hasattr(tokenizer, "add_eos_token"): # for LLaMA tokenizer
+            add_eos_token_flag = getattr(tokenizer, "add_eos_token")
            setattr(tokenizer, "add_eos_token", True)

        tokenized_examples = tokenizer(examples["prompt"], **kwargs)
@@ -54,16 +61,21 @@ def preprocess_dataset(
            k: [t[i: i + block_size] for i in range(0, total_length, block_size)]
            for k, t in concatenated_examples.items()
        }
+        # make sure the saved tokenizer is the same as the original one
+        if hasattr(tokenizer, "add_eos_token"):
+            setattr(tokenizer, "add_eos_token", add_eos_token_flag)
        return result

-    def preprocess_supervised_dataset(examples: Dict[str, List[Any]]) -> Dict[str, Any]:
+    def preprocess_supervised_dataset(examples: Dict[str, List[Any]]) -> 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": []}

        for query, response, history, system in construct_example(examples):
-            input_ids, labels = [], []
+            if not (isinstance(query, str) and isinstance(response, str) and query != "" and response != ""):
+                continue

+            input_ids, labels = [], []
            for turn_idx, (source_ids, target_ids) in enumerate(template.encode_multiturn(
                tokenizer, query, response, history, system
            )):
@@ -100,12 +112,15 @@ def preprocess_dataset(

        return model_inputs

-    def preprocess_packed_supervised_dataset(examples: Dict[str, List[Any]]) -> Dict[str, Any]:
+    def preprocess_packed_supervised_dataset(examples: Dict[str, List[Any]]) -> 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>`
        model_inputs = {"input_ids": [], "attention_mask": [], "labels": []}
        input_ids, labels = [], []
        for query, response, history, system in construct_example(examples):
+            if not (isinstance(query, str) and isinstance(response, str) and query != "" and response != ""):
+                continue
+
            for turn_idx, (source_ids, target_ids) in enumerate(template.encode_multiturn(
                tokenizer, query, response, history, system
            )):
@@ -134,11 +149,14 @@ def preprocess_dataset(

        return model_inputs

-    def preprocess_unsupervised_dataset(examples: Dict[str, List[Any]]) -> Dict[str, Any]:
+    def preprocess_unsupervised_dataset(examples: Dict[str, List[Any]]) -> Dict[str, List[List[int]]]:
        # build inputs with format `<bos> X` and labels with format `Y <eos>`
        model_inputs = {"input_ids": [], "attention_mask": [], "labels": []}

        for query, response, history, system in construct_example(examples):
+            if not (isinstance(query, str) and query != ""):
+                continue
+
            input_ids, labels = template.encode_oneturn(tokenizer, query, response, history, system)

            if template.efficient_eos:
@@ -155,10 +173,13 @@ def preprocess_dataset(

        return model_inputs

-    def preprocess_pairwise_dataset(examples):
+    def preprocess_pairwise_dataset(examples: Dict[str, List[Any]]) -> Dict[str, List[List[int]]]:
        # build input pairs with format `<bos> X`, `Y1 <eos>` and `Y2 <eos>`
        model_inputs = {"prompt_ids": [], "chosen_ids": [], "rejected_ids": []}
        for query, response, history, system in construct_example(examples):
+            if not (isinstance(query, str) and isinstance(response, list) and query != "" and len(response) > 1):
+                continue
+
            prompt_ids, chosen_ids = template.encode_oneturn(tokenizer, query, response[0], history, system)
            _, rejected_ids = template.encode_oneturn(tokenizer, query, response[1], history, system)

@@ -180,9 +201,10 @@ def preprocess_dataset(
            model_inputs["prompt_ids"].append(prompt_ids)
            model_inputs["chosen_ids"].append(chosen_ids)
            model_inputs["rejected_ids"].append(rejected_ids)
+
        return model_inputs

-    def print_supervised_dataset_example(example):
+    def print_supervised_dataset_example(example: Dict[str, List[int]]) -> None:
        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"]))
@@ -190,7 +212,7 @@ def preprocess_dataset(
            tokenizer.decode(list(filter(lambda x: x != IGNORE_INDEX, example["labels"])), skip_special_tokens=False)
        ))

-    def print_pairwise_dataset_example(example):
+    def print_pairwise_dataset_example(example: Dict[str, List[int]]) -> None:
        print("prompt_ids:\n{}".format(example["prompt_ids"]))
        print("prompt:\n{}".format(tokenizer.decode(example["prompt_ids"], skip_special_tokens=False)))
        print("chosen_ids:\n{}".format(example["chosen_ids"]))
@@ -198,33 +220,34 @@ def preprocess_dataset(
        print("rejected_ids:\n{}".format(example["rejected_ids"]))
        print("rejected:\n{}".format(tokenizer.decode(example["rejected_ids"], skip_special_tokens=False)))

-    def print_unsupervised_dataset_example(example):
+    def print_unsupervised_dataset_example(example: Dict[str, List[int]]) -> None:
        print("input_ids:\n{}".format(example["input_ids"]))
        print("inputs:\n{}".format(tokenizer.decode(example["input_ids"], skip_special_tokens=False)))

    if stage == "pt":
-        dataset = dataset.filter(lambda example: example["prompt"])
        preprocess_func = preprocess_pretrain_dataset
        print_function = print_unsupervised_dataset_example
    elif stage == "sft" and not training_args.predict_with_generate:
-        dataset = dataset.filter(lambda example: example["prompt"] and example["response"])
        preprocess_func = preprocess_packed_supervised_dataset if data_args.sft_packing else preprocess_supervised_dataset
        print_function = print_supervised_dataset_example
    elif stage == "rm":
-        dataset = dataset.filter(lambda example: example["prompt"] and len(example["response"]) > 1)
        preprocess_func = preprocess_pairwise_dataset
        print_function = print_pairwise_dataset_example
    else:
-        dataset = dataset.filter(lambda example: example["prompt"])
        preprocess_func = preprocess_unsupervised_dataset
        print_function = print_unsupervised_dataset_example

+    if data_args.cache_path is not None and os.path.exists(data_args.cache_path):
+        logger.warning("Loading dataset from disk will ignore other data arguments.")
+        return load_from_disk(data_args.cache_path)
+
    with training_args.main_process_first(desc="dataset map pre-processing"):
+        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,
+                load_from_cache_file=(not data_args.overwrite_cache),
                desc="Running tokenizer on dataset"
            )

@@ -235,9 +258,15 @@ def preprocess_dataset(
            **kwargs
        )

-        try:
-            print_function(next(iter(dataset)))
-        except StopIteration:
-            raise ValueError("Empty dataset!")
+        if data_args.cache_path is not None and not os.path.exists(data_args.cache_path):
+            if training_args.should_save:
+                dataset.save_to_disk(data_args.cache_path)
+            raise SystemExit("Dataset saved, rerun this script with the same `--cache_path`.")
+
+        if training_args.should_log:
+            try:
+                print_function(next(iter(dataset)))
+            except StopIteration:
+                raise RuntimeError("Empty dataset!")

        return dataset
--- a/src/llmtuner/dsets/utils.py
+++ b/src/llmtuner/dsets/utils.py
@@ -13,9 +13,11 @@ logger = get_logger(__name__)


 EXT2TYPE = {
+    "arrow": "arrow",
    "csv": "csv",
    "json": "json",
    "jsonl": "json",
+    "parquet": "parquet",
    "txt": "text"
 }

--- a/src/llmtuner/eval/init.py
+++ b/src/llmtuner/eval/init.py
@@ -0,0 +1 @@
+from llmtuner.eval.engine import Evaluator
--- a/src/llmtuner/eval/constants.py
+++ b/src/llmtuner/eval/constants.py
@@ -0,0 +1,3 @@
+CHOICES = ["A", "B", "C", "D"]
+
+SUBJECTS = ["Average", "STEM", "Social Sciences", "Humanities", "Other"]
--- a/src/llmtuner/eval/engine.py
+++ b/src/llmtuner/eval/engine.py
@@ -0,0 +1,110 @@
+# Inspired by: https://github.com/hendrycks/test/blob/master/evaluate_flan.py
+
+import os
+import json
+import torch
+import tiktoken
+import numpy as np
+from tqdm import tqdm, trange
+from datasets import load_dataset
+from typing import Any, Dict, List, Optional
+
+from llmtuner.eval.constants import CHOICES, SUBJECTS
+from llmtuner.eval.parser import get_eval_args
+from llmtuner.eval.template import get_eval_template
+from llmtuner.extras.misc import dispatch_model
+from llmtuner.extras.template import get_template_and_fix_tokenizer
+from llmtuner.tuner.core import load_model_and_tokenizer
+
+
+class Evaluator:
+
+    def __init__(self, args: Optional[Dict[str, Any]] = None) -> None:
+        model_args, self.data_args, self.eval_args, finetuning_args = get_eval_args(args)
+        self.model, self.tokenizer = load_model_and_tokenizer(model_args, finetuning_args)
+        self.tokenizer.padding_side = "right" # avoid overflow issue in batched inference for llama2
+        self.model = dispatch_model(self.model)
+        self.template = get_template_and_fix_tokenizer(self.data_args.template, self.tokenizer)
+        self.eval_template = get_eval_template(self.eval_args.lang)
+        self.choice_inputs = self._encode_choices()
+
+    def _encode_choices(self) -> List[int]:
+        if isinstance(getattr(self.tokenizer, "tokenizer", None), tiktoken.Encoding): # for tiktoken tokenizer (Qwen)
+            kwargs = dict(allowed_special="all")
+        else:
+            kwargs = dict(add_special_tokens=False)
+
+        return [self.tokenizer.encode(self.eval_template.prefix + ch, **kwargs)[-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:
+        mapping = os.path.join(self.eval_args.task_dir, self.eval_args.task, "mapping.json")
+        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, self.eval_args.task),
+                name=subject,
+                download_mode="force_redownload"
+            )
+            pbar.set_postfix_str(categorys[subject]["name"])
+            inputs, outputs, labels = [], [], []
+            for i in trange(len(dataset[self.data_args.split]), desc="Formatting batches", position=1, leave=False):
+                support_set = dataset["train"].shuffle().select(range(min(self.eval_args.n_shot, len(dataset["train"]))))
+                query, resp, history = self.eval_template.format_example(
+                    target_data=dataset[self.data_args.split][i],
+                    support_set=support_set,
+                    subject_name=categorys[subject]["name"],
+                    use_history=self.template.use_history
+                )
+                input_ids, _ = self.template.encode_oneturn(
+                    tokenizer=self.tokenizer, query=query, resp=resp, history=history
+                )
+                inputs.append({"input_ids": input_ids, "attention_mask": [1] * len(input_ids)})
+                labels.append(resp)
+
+            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)
+
+
+if __name__ == "__main__":
+    evaluator = Evaluator()
+    evaluator.eval()
--- a/src/llmtuner/eval/parser.py
+++ b/src/llmtuner/eval/parser.py
@@ -0,0 +1,49 @@
+import transformers
+from typing import Any, Dict, Optional, Tuple
+from transformers import HfArgumentParser
+
+from llmtuner.extras.misc import parse_args
+from llmtuner.hparams import (
+    ModelArguments,
+    DataArguments,
+    EvaluationArguments,
+    FinetuningArguments
+)
+
+
+def parse_eval_args(
+    args: Optional[Dict[str, Any]] = None
+) -> Tuple[
+    ModelArguments,
+    DataArguments,
+    EvaluationArguments,
+    FinetuningArguments
+]:
+    parser = HfArgumentParser((
+        ModelArguments,
+        DataArguments,
+        EvaluationArguments,
+        FinetuningArguments
+    ))
+    return parse_args(parser, args)
+
+
+def get_eval_args(
+    args: Optional[Dict[str, Any]] = None
+) -> Tuple[
+    ModelArguments,
+    DataArguments,
+    EvaluationArguments,
+    FinetuningArguments
+]:
+    model_args, data_args, eval_args, finetuning_args = parse_eval_args(args)
+
+    if data_args.template is None:
+        raise ValueError("Please specify which `template` to use.")
+
+    if model_args.quantization_bit is not None and finetuning_args.finetuning_type != "lora":
+        raise ValueError("Quantization is only compatible with the LoRA method.")
+
+    transformers.set_seed(eval_args.seed)
+
+    return model_args, data_args, eval_args, finetuning_args
--- a/src/llmtuner/eval/template.py
+++ b/src/llmtuner/eval/template.py
@@ -0,0 +1,86 @@
+from dataclasses import dataclass
+from typing import TYPE_CHECKING, Dict, List, Tuple
+
+from llmtuner.eval.constants import CHOICES
+
+if TYPE_CHECKING:
+    from datasets import Dataset
+
+
+@dataclass
+class EvalTemplate:
+
+    system: str
+    choice: str
+    answer: str
+    prefix: str
+
+    def parse_example(
+        self,
+        example: Dict[str, str]
+    ) -> Tuple[str, str]:
+        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: "Dataset",
+        subject_name: str,
+        use_history: bool
+    ) -> Tuple[str, str, List[Tuple[str, str]]]:
+        query, resp = self.parse_example(target_data)
+        history = [self.parse_example(support_set[k]) for k in range(len(support_set))]
+
+        if len(history):
+            temp = history.pop(0)
+            history.insert(0, (self.system.format(subject=subject_name) + temp[0], temp[1]))
+        else:
+            query = self.system.format(subject=subject_name) + query
+
+        if not use_history:
+            query = "\n\n".join(["".join(item) for item in history] + [query])
+            history = []
+        return query.strip(), resp, history
+
+
+eval_templates: Dict[str, EvalTemplate] = {}
+
+
+def register_eval_template(
+    name: str,
+    system: str,
+    choice: str,
+    answer: str,
+    prefix: str
+) -> None:
+    eval_templates[name] = EvalTemplate(
+        system=system,
+        choice=choice,
+        answer=answer,
+        prefix=prefix
+    )
+
+
+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: ",
+    prefix=" "
+)
+
+
+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：",
+    prefix="\n"
+)
--- a/src/llmtuner/extras/constants.py
+++ b/src/llmtuner/extras/constants.py
@@ -1,9 +1,11 @@
+from collections import defaultdict, OrderedDict
+from typing import Dict, Optional
+
+
 IGNORE_INDEX = -100

 LOG_FILE_NAME = "trainer_log.jsonl"

-LAYERNORM_NAMES = ["norm", "ln_f", "ln_attn", "ln_mlp", "ln_1", "ln_2"]
-
 METHODS = ["full", "freeze", "lora"]

 TRAINING_STAGES = {
@@ -14,75 +16,222 @@ TRAINING_STAGES = {
    "Pre-Training": "pt"
 }

-SUPPORTED_MODELS = {
-    "LLaMA-7B": "huggyllama/llama-7b",
-    "LLaMA-13B": "huggyllama/llama-13b",
-    "LLaMA-30B": "huggyllama/llama-30b",
-    "LLaMA-65B": "huggyllama/llama-65b",
-    "LLaMA2-7B": "meta-llama/Llama-2-7b-hf",
-    "LLaMA2-13B": "meta-llama/Llama-2-13b-hf",
-    "LLaMA2-70B": "meta-llama/Llama-2-70b-hf",
-    "LLaMA2-7B-Chat": "meta-llama/Llama-2-7b-chat-hf",
-    "LLaMA2-13B-Chat": "meta-llama/Llama-2-13b-chat-hf",
-    "LLaMA2-70B-Chat": "meta-llama/Llama-2-70b-chat-hf",
-    "ChineseLLaMA2-7B": "ziqingyang/chinese-llama-2-7b",
-    "ChineseLLaMA2-13B": "ziqingyang/chinese-llama-2-13b",
-    "ChineseLLaMA2-7B-Chat": "ziqingyang/chinese-alpaca-2-7b",
-    "ChineseLLaMA2-13B-Chat": "ziqingyang/chinese-alpaca-2-13b",
-    "BLOOM-560M": "bigscience/bloom-560m",
-    "BLOOM-3B": "bigscience/bloom-3b",
-    "BLOOM-7B1": "bigscience/bloom-7b1",
-    "BLOOMZ-560M": "bigscience/bloomz-560m",
-    "BLOOMZ-3B": "bigscience/bloomz-3b",
-    "BLOOMZ-7B1-mt": "bigscience/bloomz-7b1-mt",
-    "Falcon-7B": "tiiuae/falcon-7b",
-    "Falcon-40B": "tiiuae/falcon-40b",
-    "Falcon-7B-Chat": "tiiuae/falcon-7b-instruct",
-    "Falcon-40B-Chat": "tiiuae/falcon-40b-instruct",
-    "Baichuan-7B": "baichuan-inc/Baichuan-7B",
-    "Baichuan-13B": "baichuan-inc/Baichuan-13B-Base",
-    "Baichuan-13B-Chat": "baichuan-inc/Baichuan-13B-Chat",
-    "Baichuan2-7B": "baichuan-inc/Baichuan2-7B-Base",
-    "Baichuan2-13B": "baichuan-inc/Baichuan2-13B-Base",
-    "Baichuan2-7B-Chat": "baichuan-inc/Baichuan2-7B-Chat",
-    "Baichuan2-13B-Chat": "baichuan-inc/Baichuan2-13B-Chat",
-    "InternLM-7B": "internlm/internlm-7b",
-    "InternLM-20B": "internlm/internlm-20b",
-    "InternLM-7B-Chat": "internlm/internlm-chat-7b",
-    "InternLM-20B-Chat": "internlm/internlm-chat-20b",
-    "Qwen-7B": "Qwen/Qwen-7B",
-    "Qwen-14B": "Qwen/Qwen-14B",
-    "Qwen-7B-Chat": "Qwen/Qwen-7B-Chat",
-    "Qwen-14B-Chat": "Qwen/Qwen-14B-Chat",
-    "XVERSE-13B": "xverse/XVERSE-13B",
-    "XVERSE-13B-Chat": "xverse/XVERSE-13B-Chat",
-    "ChatGLM2-6B-Chat": "THUDM/chatglm2-6b",
-    "Phi1.5-1.3B": "microsoft/phi-1_5"
-}
+LAYERNORM_NAMES = {"norm", "ln"}

-DEFAULT_MODULE = {
-    "LLaMA": "q_proj,v_proj",
-    "LLaMA2": "q_proj,v_proj",
-    "ChineseLLaMA2": "q_proj,v_proj",
-    "BLOOM": "query_key_value",
-    "BLOOMZ": "query_key_value",
-    "Falcon": "query_key_value",
-    "Baichuan": "W_pack",
-    "Baichuan2": "W_pack",
-    "InternLM": "q_proj,v_proj",
-    "Qwen": "c_attn",
-    "XVERSE": "q_proj,v_proj",
-    "ChatGLM2": "query_key_value",
-    "Phi1.5": "Wqkv"
-}
+SUPPORTED_MODELS = OrderedDict()

-DEFAULT_TEMPLATE = {
-    "LLaMA2": "llama2",
-    "ChineseLLaMA2": "llama2_zh",
-    "Baichuan": "baichuan",
-    "Baichuan2": "baichuan2",
-    "InternLM": "intern",
-    "Qwen": "chatml",
-    "XVERSE": "xverse",
-    "ChatGLM2": "chatglm2"
-}
+DEFAULT_MODULE = defaultdict(str)
+
+DEFAULT_TEMPLATE = defaultdict(str)
+
+
+def register_model_group(
+    models: Dict[str, str],
+    module: Optional[str] = None,
+    template: Optional[str] = None
+) -> 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 module is not None:
+        DEFAULT_MODULE[prefix] = module
+    if template is not None:
+        DEFAULT_TEMPLATE[prefix] = template
+
+
+register_model_group(
+    models={
+        "Baichuan-7B-Base": "baichuan-inc/Baichuan-7B",
+        "Baichuan-13B-Base": "baichuan-inc/Baichuan-13B-Base",
+        "Baichuan-13B-Chat": "baichuan-inc/Baichuan-13B-Chat"
+    },
+    module="W_pack",
+    template="baichuan"
+)
+
+
+register_model_group(
+    models={
+        "Baichuan2-7B-Base": "baichuan-inc/Baichuan2-7B-Base",
+        "Baichuan2-13B-Base": "baichuan-inc/Baichuan2-13B-Base",
+        "Baichuan2-7B-Chat": "baichuan-inc/Baichuan2-7B-Chat",
+        "Baichuan2-13B-Chat": "baichuan-inc/Baichuan2-13B-Chat"
+    },
+    module="W_pack",
+    template="baichuan2"
+)
+
+
+register_model_group(
+    models={
+        "BLOOM-560M": "bigscience/bloom-560m",
+        "BLOOM-3B": "bigscience/bloom-3b",
+        "BLOOM-7B1": "bigscience/bloom-7b1"
+    },
+    module="query_key_value"
+)
+
+
+register_model_group(
+    models={
+        "BLOOMZ-560M": "bigscience/bloomz-560m",
+        "BLOOMZ-3B": "bigscience/bloomz-3b",
+        "BLOOMZ-7B1-mt": "bigscience/bloomz-7b1-mt"
+    },
+    module="query_key_value"
+)
+
+
+register_model_group(
+    models={
+        "BlueLM-7B-Base": "vivo-ai/BlueLM-7B-Base",
+        "BlueLM-7B-Chat": "vivo-ai/BlueLM-7B-Chat"
+    },
+    template="bluelm"
+)
+
+
+register_model_group(
+    models={
+        "ChatGLM2-6B-Chat": "THUDM/chatglm2-6b"
+    },
+    module="query_key_value",
+    template="chatglm2"
+)
+
+
+register_model_group(
+    models={
+        "ChatGLM3-6B-Base": "THUDM/chatglm3-6b-base",
+        "ChatGLM3-6B-Chat": "THUDM/chatglm3-6b"
+    },
+    module="query_key_value",
+    template="chatglm3"
+)
+
+
+register_model_group(
+    models={
+        "ChineseLLaMA2-7B": "ziqingyang/chinese-llama-2-7b",
+        "ChineseLLaMA2-13B": "ziqingyang/chinese-llama-2-13b",
+        "ChineseLLaMA2-7B-Chat": "ziqingyang/chinese-alpaca-2-7b",
+        "ChineseLLaMA2-13B-Chat": "ziqingyang/chinese-alpaca-2-13b"
+    },
+    template="llama2_zh"
+)
+
+
+register_model_group(
+    models={
+        "Falcon-7B": "tiiuae/falcon-7b",
+        "Falcon-40B": "tiiuae/falcon-40b",
+        "Falcon-180B": "tiiuae/falcon-180B",
+        "Falcon-7B-Chat": "tiiuae/falcon-7b-instruct",
+        "Falcon-40B-Chat": "tiiuae/falcon-40b-instruct",
+        "Falcon-180B-Chat": "tiiuae/falcon-180B-chat"
+    },
+    module="query_key_value",
+    template="falcon"
+)
+
+
+register_model_group(
+    models={
+        "InternLM-7B": "internlm/internlm-7b",
+        "InternLM-20B": "internlm/internlm-20b",
+        "InternLM-7B-Chat": "internlm/internlm-chat-7b",
+        "InternLM-20B-Chat": "internlm/internlm-chat-20b"
+    },
+    template="intern"
+)
+
+
+register_model_group(
+    models={
+        "LingoWhale-8B": "deeplang-ai/LingoWhale-8B"
+    },
+    module="qkv_proj"
+)
+
+
+register_model_group(
+    models={
+        "LLaMA-7B": "huggyllama/llama-7b",
+        "LLaMA-13B": "huggyllama/llama-13b",
+        "LLaMA-30B": "huggyllama/llama-30b",
+        "LLaMA-65B": "huggyllama/llama-65b"
+    }
+)
+
+
+register_model_group(
+    models={
+        "LLaMA2-7B": "meta-llama/Llama-2-7b-hf",
+        "LLaMA2-13B": "meta-llama/Llama-2-13b-hf",
+        "LLaMA2-70B": "meta-llama/Llama-2-70b-hf",
+        "LLaMA2-7B-Chat": "meta-llama/Llama-2-7b-chat-hf",
+        "LLaMA2-13B-Chat": "meta-llama/Llama-2-13b-chat-hf",
+        "LLaMA2-70B-Chat": "meta-llama/Llama-2-70b-chat-hf"
+    },
+    template="llama2"
+)
+
+
+register_model_group(
+    models={
+        "Mistral-7B": "mistralai/Mistral-7B-v0.1",
+        "Mistral-7B-Chat": "mistralai/Mistral-7B-Instruct-v0.1"
+    },
+    template="mistral"
+)
+
+
+register_model_group(
+    models={
+        "Phi1.5-1.3B": "microsoft/phi-1_5"
+    },
+    module="Wqkv"
+)
+
+
+register_model_group(
+    models={
+        "Qwen-7B": "Qwen/Qwen-7B",
+        "Qwen-14B": "Qwen/Qwen-14B",
+        "Qwen-7B-Chat": "Qwen/Qwen-7B-Chat",
+        "Qwen-14B-Chat": "Qwen/Qwen-14B-Chat"
+    },
+    module="c_attn",
+    template="qwen"
+)
+
+
+register_model_group(
+    models={
+        "Skywork-13B-Base": "Skywork/Skywork-13B-base"
+    }
+)
+
+
+register_model_group(
+    models={
+        "XVERSE-7B": "xverse/XVERSE-7B",
+        "XVERSE-13B": "xverse/XVERSE-13B",
+        "XVERSE-65B": "xverse/XVERSE-65B",
+        "XVERSE-7B-Chat": "xverse/XVERSE-7B-Chat",
+        "XVERSE-13B-Chat": "xverse/XVERSE-13B-Chat"
+    },
+    template="xverse"
+)
+
+
+register_model_group(
+    models={
+        "Yi-6B": "01-ai/Yi-6B",
+        "Yi-34B": "01-ai/Yi-34B"
+    }
+)
--- a/src/llmtuner/extras/misc.py
+++ b/src/llmtuner/extras/misc.py
@@ -1,6 +1,8 @@
 import gc
+import os
+import sys
 import torch
-from typing import TYPE_CHECKING, Tuple
+from typing import TYPE_CHECKING, Any, Dict, Optional, Tuple
 from transformers import InfNanRemoveLogitsProcessor, LogitsProcessorList

 try:
@@ -17,6 +19,7 @@ except ImportError:
    _is_bf16_available = torch.cuda.is_bf16_supported()

 if TYPE_CHECKING:
+    from transformers import HfArgumentParser
    from transformers.modeling_utils import PreTrainedModel


@@ -74,7 +77,7 @@ def infer_optim_dtype(model_dtype: torch.dtype) -> torch.dtype:
        return torch.float32


-def get_logits_processor() -> LogitsProcessorList:
+def get_logits_processor() -> "LogitsProcessorList":
    r"""
    Gets logits processor that removes NaN and Inf logits.
    """
@@ -93,6 +96,17 @@ def torch_gc() -> None:
        torch.cuda.ipc_collect()


+def parse_args(parser: "HfArgumentParser", args: Optional[Dict[str, Any]] = None) -> Tuple[Any]:
+    if args is not None:
+        return parser.parse_dict(args)
+    elif len(sys.argv) == 2 and sys.argv[1].endswith(".yaml"):
+        return parser.parse_yaml_file(os.path.abspath(sys.argv[1]))
+    elif len(sys.argv) == 2 and sys.argv[1].endswith(".json"):
+        return parser.parse_json_file(os.path.abspath(sys.argv[1]))
+    else:
+        return parser.parse_args_into_dataclasses()
+
+
 def dispatch_model(model: "PreTrainedModel") -> "PreTrainedModel":
    r"""
    Dispatches a pre-trained model to GPUs with balanced memory.
--- a/src/llmtuner/extras/patches/llama_patch.py
+++ b/src/llmtuner/extras/patches/llama_patch.py
@@ -5,11 +5,14 @@ from typing import Optional, Tuple
 from transformers.utils import logging
 from transformers.models.llama.modeling_llama import LlamaAttention, apply_rotary_pos_emb, repeat_kv

+is_flash_attn_2_available = False
+
 try:
    from flash_attn import flash_attn_func, flash_attn_varlen_func # type: ignore
    from flash_attn.bert_padding import pad_input, unpad_input # type: ignore
+    is_flash_attn_2_available = True
 except ImportError:
-    print("FlashAttention-2 is not installed, ignore this if you are not using FlashAttention.")
+    is_flash_attn_2_available = False


 logger = logging.get_logger(__name__)
--- a/src/llmtuner/extras/save_and_load.py
+++ b/src/llmtuner/extras/save_and_load.py
@@ -1,21 +0,0 @@
-import os
-import torch
-from transformers.trainer import WEIGHTS_NAME
-
-from llmtuner.extras.logging import get_logger
-
-
-logger = get_logger(__name__)
-
-
-def load_valuehead_params(model: torch.nn.Module, checkpoint_dir: os.PathLike) -> bool:
-    vhead_file = os.path.join(checkpoint_dir, WEIGHTS_NAME)
-    if not os.path.exists(vhead_file):
-        logger.warning("Provided path ({}) does not contain valuehead weights.".format(checkpoint_dir))
-        return False
-    vhead_params = torch.load(vhead_file, map_location="cpu")
-    model.register_buffer("reward_head_weight", vhead_params["v_head.summary.weight"], persistent=False)
-    model.register_buffer("reward_head_bias", vhead_params["v_head.summary.bias"], persistent=False)
-    model.register_buffer("default_head_weight", torch.zeros_like(vhead_params["v_head.summary.weight"]), persistent=False)
-    model.register_buffer("default_head_bias", torch.zeros_like(vhead_params["v_head.summary.bias"]), persistent=False)
-    return True
--- a/src/llmtuner/extras/template.py
+++ b/src/llmtuner/extras/template.py
@@ -225,89 +225,8 @@ def get_template_and_fix_tokenizer(
    return template


-r"""
-Supports language model inference without histories.
-"""
-register_template(
-    name="vanilla",
-    prefix=[],
-    prompt=[
-        "{{query}}"
-    ],
-    system="",
-    sep=[],
-    use_history=False
-)
-
-
-r"""
-Default template.
-"""
-register_template(
-    name="default",
-    prefix=[
-        "{{system}}"
-    ],
-    prompt=[
-        "Human: {{query}}\nAssistant: "
-    ],
-    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."
-    ),
-    sep=[
-        "\n"
-    ]
-)
-
-
-r"""
-Supports: https://huggingface.co/meta-llama/Llama-2-7b-chat-hf
-          https://huggingface.co/meta-llama/Llama-2-13b-chat-hf
-          https://huggingface.co/meta-llama/Llama-2-70b-chat-hf
-"""
-register_template(
-    name="llama2",
-    prefix=[
-        "<<SYS>>\n{{system}}\n<</SYS>>\n\n"
-    ],
-    prompt=[
-        "[INST] {{query}} [/INST] "
-    ],
-    system=(
-        "You are a helpful, respectful and honest assistant. "
-        "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."
-    ),
-    sep=[]
-)
-
-
-r"""
-Supports: https://github.com/ymcui/Chinese-LLaMA-Alpaca-2
-          https://huggingface.co/ziqingyang/chinese-alpaca-2-7b
-"""
-register_template(
-    name="llama2_zh",
-    prefix=[
-        "<<SYS>>\n{{system}}\n<</SYS>>\n\n"
-    ],
-    prompt=[
-        "[INST] {{query}} [/INST] "
-    ],
-    system="You are a helpful assistant. 你是一个乐于助人的助手。",
-    sep=[]
-)
-
-
 r"""
 Supports: https://huggingface.co/tatsu-lab/alpaca-7b-wdiff
-          https://github.com/ymcui/Chinese-LLaMA-Alpaca
 """
 register_template(
    name="alpaca",
@@ -327,103 +246,10 @@ register_template(
 )


-r"""
-Supports: https://huggingface.co/lmsys/vicuna-7b-delta-v1.1
-          https://huggingface.co/lmsys/vicuna-13b-delta-v1.1
-"""
-register_template(
-    name="vicuna",
-    prefix=[
-        "{{system}}"
-    ],
-    prompt=[
-        "USER: {{query}} ASSISTANT: "
-    ],
-    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."
-    ),
-    sep=[]
-)
-
-
-r"""
-Supports: https://huggingface.co/BelleGroup/BELLE-LLaMA-EXT-13B
-"""
-register_template(
-    name="belle",
-    prefix=[
-        "{{system}}"
-    ],
-    prompt=[
-        "Human: {{query}}\n\nBelle: "
-    ],
-    system="",
-    sep=[
-        "\n\n"
-    ]
-)
-
-
-r"""
-Supports: https://github.com/CVI-SZU/Linly
-"""
-register_template(
-    name="linly",
-    prefix=[
-        "{{system}}"
-    ],
-    prompt=[
-        "User: {{query}}\nBot: "
-    ],
-    system="",
-    sep=[
-        "\n"
-    ]
-)
-
-
-r"""
-Supports: https://github.com/Neutralzz/BiLLa
-"""
-register_template(
-    name="billa",
-    prefix=[
-        "{{system}}"
-    ],
-    prompt=[
-        "Human: {{query}}\nAssistant: "
-    ],
-    system="",
-    sep=[
-        "\n"
-    ]
-)
-
-
-r"""
-Supports: https://huggingface.co/IDEA-CCNL/Ziya-LLaMA-13B-v1
-"""
-register_template(
-    name="ziya",
-    prefix=[
-        "{{system}}"
-    ],
-    prompt=[
-        {"token": "<human>"},
-        ":{{query}}\n",
-        {"token": "<bot>"},
-        ":"
-    ],
-    system="",
-    sep=[
-        "\n"
-    ]
-)
-
-
 r"""
 Supports: https://huggingface.co/BAAI/AquilaChat-7B
+          https://huggingface.co/BAAI/AquilaChat2-7B
+          https://huggingface.co/BAAI/AquilaChat2-34B
 """
 register_template(
    name="aquila",
@@ -431,7 +257,7 @@ register_template(
        "{{system}}"
    ],
    prompt=[
-        "Human: {{query}}###Assistant: "
+        "Human: {{query}}###Assistant:"
    ],
    system=(
        "A chat between a curious human and an artificial intelligence assistant. "
@@ -447,31 +273,6 @@ register_template(
 )


-r"""
-Supports: https://huggingface.co/internlm/internlm-chat-7b
-"""
-register_template(
-    name="intern",
-    prefix=[
-        "{{system}}"
-    ],
-    prompt=[
-        "<|User|>:{{query}}",
-        {"token": "<eoh>"},
-        "\n<|Bot|>:"
-    ],
-    system="",
-    sep=[
-        {"token": "<eoa>"},
-        "\n"
-    ],
-    stop_words=[
-        "<eoa>"
-    ],
-    efficient_eos=True
-)
-
-
 r"""
 Supports: https://huggingface.co/baichuan-inc/Baichuan-13B-Chat
 """
@@ -511,6 +312,300 @@ register_template(
 )


+r"""
+Supports: https://huggingface.co/BelleGroup/BELLE-LLaMA-EXT-13B
+"""
+register_template(
+    name="belle",
+    prefix=[
+        "{{system}}"
+    ],
+    prompt=[
+        "Human: {{query}}\n\nBelle: "
+    ],
+    system="",
+    sep=[
+        "\n\n"
+    ]
+)
+
+
+r"""
+Supports: https://huggingface.co/vivo-ai/BlueLM-7B-Chat
+"""
+register_template(
+    name="bluelm",
+    prefix=[
+        "{{system}}"
+    ],
+    prompt=[
+        {"token": "[|Human|]:"},
+        "{{query}}",
+        {"token": "[|AI|]:"}
+    ],
+    system="",
+    sep=[]
+)
+
+
+r"""
+Supports: https://huggingface.co/THUDM/chatglm2-6b
+"""
+register_template(
+    name="chatglm2",
+    prefix=[
+        {"token": "[gMASK]"},
+        {"token": "sop"},
+        "{{system}}"
+    ],
+    prompt=[
+        "[Round {{idx}}]\n\n问：{{query}}\n\n答："
+    ],
+    system="",
+    sep=[
+        "\n\n"
+    ],
+    efficient_eos=True
+)
+
+
+r"""
+Supports: https://huggingface.co/THUDM/chatglm3-6b
+"""
+register_template(
+    name="chatglm3",
+    prefix=[
+        {"token": "[gMASK]"},
+        {"token": "sop"},
+        "{{system}}"
+    ],
+    prompt=[
+        {"token": "<|user|>"},
+        "\n",
+        "{{query}}",
+        {"token": "<|assistant|>"}
+    ],
+    system="",
+    sep=[],
+    stop_words=[
+        "<|user|>",
+        "<|observation|>"
+    ],
+    efficient_eos=True
+)
+
+
+r"""
+Supports: https://huggingface.co/deepseek-ai/deepseek-coder-1.3b-instruct
+          https://huggingface.co/deepseek-ai/deepseek-coder-6.7b-instruct
+          https://huggingface.co/deepseek-ai/deepseek-coder-33b-instruct
+"""
+register_template(
+    name="deepseek",
+    prefix=[
+        "{{system}}"
+    ],
+    prompt=[
+        "### Instruction:\n{{query}}\n\n### Response:\n"
+    ],
+    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."
+    ),
+    sep=[
+        "\n",
+        {"token": "<|EOT|>"},
+        "\n\n"
+    ],
+    stop_words=[
+        "<|EOT|>"
+    ],
+    efficient_eos=True
+)
+
+
+r"""
+Default template.
+"""
+register_template(
+    name="default",
+    prefix=[
+        "{{system}}"
+    ],
+    prompt=[
+        "Human: {{query}}\nAssistant:"
+    ],
+    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."
+    ),
+    sep=[
+        "\n"
+    ]
+)
+
+
+r"""
+Supports: https://huggingface.co/tiiuae/falcon-180B-chat
+"""
+register_template(
+    name="falcon",
+    prefix=[
+        "{{system}}"
+    ],
+    prompt=[
+        "User: {{query}}\nFalcon:"
+    ],
+    system="",
+    sep=[
+        "\n"
+    ],
+    efficient_eos=True
+)
+
+
+r"""
+Supports: https://huggingface.co/internlm/internlm-chat-7b
+          https://huggingface.co/internlm/internlm-chat-20b
+"""
+register_template(
+    name="intern",
+    prefix=[
+        "{{system}}"
+    ],
+    prompt=[
+        "<|User|>:{{query}}",
+        {"token": "<eoh>"},
+        "\n<|Bot|>:"
+    ],
+    system="",
+    sep=[
+        {"token": "<eoa>"},
+        "\n"
+    ],
+    stop_words=[
+        "<eoa>"
+    ],
+    efficient_eos=True
+)
+
+
+r"""
+Supports: https://huggingface.co/meta-llama/Llama-2-7b-chat-hf
+          https://huggingface.co/meta-llama/Llama-2-13b-chat-hf
+          https://huggingface.co/meta-llama/Llama-2-70b-chat-hf
+"""
+register_template(
+    name="llama2",
+    prefix=[
+        "<<SYS>>\n{{system}}\n<</SYS>>\n\n"
+    ],
+    prompt=[
+        "[INST] {{query}} [/INST]"
+    ],
+    system=(
+        "You are a helpful, respectful and honest assistant. "
+        "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."
+    ),
+    sep=[]
+)
+
+
+r"""
+Supports: https://huggingface.co/ziqingyang/chinese-alpaca-2-7b
+          https://huggingface.co/ziqingyang/chinese-alpaca-2-13b
+"""
+register_template(
+    name="llama2_zh",
+    prefix=[
+        "<<SYS>>\n{{system}}\n<</SYS>>\n\n"
+    ],
+    prompt=[
+        "[INST] {{query}} [/INST]"
+    ],
+    system="You are a helpful assistant. 你是一个乐于助人的助手。",
+    sep=[]
+)
+
+
+r"""
+Supports: https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1
+"""
+register_template(
+    name="mistral",
+    prefix=[
+        "{{system}}"
+    ],
+    prompt=[
+        "[INST] {{query}} [/INST]"
+    ],
+    system="",
+    sep=[]
+)
+
+
+r"""
+Supports: https://huggingface.co/openchat/openchat_3.5
+"""
+register_template(
+    name="openchat",
+    prefix=[
+        "{{system}}"
+    ],
+    prompt=[
+        "GPT4 Correct User: {{query}}",
+        {"token": "<|end_of_turn|>"},
+        "GPT4 Correct Assistant:"
+    ],
+    system="",
+    sep=[
+        {"token": "<|end_of_turn|>"}
+    ],
+    stop_words=[
+        "<|end_of_turn|>"
+    ],
+    efficient_eos=True
+)
+
+
+r"""
+Supports: https://huggingface.co/Qwen/Qwen-7B-Chat
+          https://huggingface.co/Qwen/Qwen-14B-Chat
+"""
+register_template(
+    name="qwen",
+    prefix=[
+        {"token": "<|im_start|>"},
+        "system\n{{system}}"
+    ],
+    prompt=[
+        {"token": "<|im_start|>"},
+        "user\n{{query}}",
+        {"token": "<|im_end|>"},
+        "\n",
+        {"token": "<|im_start|>"},
+        "assistant\n"
+    ],
+    system="You are a helpful assistant.",
+    sep=[
+        {"token": "<|im_end|>"},
+        "\n"
+    ],
+    stop_words=[
+        "<|im_end|>"
+    ],
+    efficient_eos=True
+)
+
+
 r"""
 Supports: https://huggingface.co/HuggingFaceH4/starchat-alpha
          https://huggingface.co/HuggingFaceH4/starchat-beta
@@ -541,57 +636,43 @@ register_template(


 r"""
-Supports: https://huggingface.co/Qwen/Qwen-7B-Chat
+Supports language model inference without histories.
 """
 register_template(
-    name="chatml",
-    prefix=[
-        {"token": "<|im_start|>"},
-        "system\n{{system}}"
-    ],
+    name="vanilla",
+    prefix=[],
    prompt=[
-        {"token": "<|im_start|>"},
-        "user\n{{query}}",
-        {"token": "<|im_end|>"},
-        "\n",
-        {"token": "<|im_start|>"},
-        "assistant\n"
+        "{{query}}"
    ],
-    system="You are a helpful assistant.",
-    sep=[
-        {"token": "<|im_end|>"},
-        "\n"
-    ],
-    stop_words=[
-        "<|im_end|>"
-    ],
-    efficient_eos=True
+    system="",
+    sep=[],
+    use_history=False
 )


 r"""
-Supports: https://huggingface.co/THUDM/chatglm2-6b
+Supports: https://huggingface.co/lmsys/vicuna-7b-v1.5
+          https://huggingface.co/lmsys/vicuna-13b-v1.5
 """
 register_template(
-    name="chatglm2",
+    name="vicuna",
    prefix=[
-        {"token": "[gMASK]"},
-        {"token": "sop"},
        "{{system}}"
    ],
    prompt=[
-        "[Round {{idx}}]\n\n问：{{query}}\n\n答："
+        "USER: {{query}} ASSISTANT:"
    ],
-    system="",
-    sep=[
-        "\n\n"
-    ],
-    efficient_eos=True
+    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."
+    ),
+    sep=[]
 )


 r"""
-Supports: https://huggingface.co/xverse/XVERSE-13B-Chat
+Supports: https://huggingface.co/xverse/XVERSE-7B-Chat
+          https://huggingface.co/xverse/XVERSE-13B-Chat
 """
 register_template(
    name="xverse",
@@ -604,3 +685,85 @@ register_template(
    system="",
    sep=[]
 )
+
+
+r"""
+Supports: https://huggingface.co/wenge-research/yayi-7b
+          https://huggingface.co/wenge-research/yayi-7b-llama2
+          https://huggingface.co/wenge-research/yayi-13b-llama2
+"""
+register_template(
+    name="yayi",
+    prefix=[
+        {"token": "<|System|>"},
+        ":\n{{system}}"
+    ],
+    prompt=[
+        {"token": "<|Human|>"},
+        ":\n{{query}}\n\n",
+        {"token": "<|YaYi|>"},
+        ":"
+    ],
+    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."
+    ),
+    sep=[
+        "\n\n"
+    ],
+    stop_words=[
+        "<|End|>"
+    ]
+)
+
+
+r"""
+Supports: https://huggingface.co/HuggingFaceH4/zephyr-7b-alpha
+          https://huggingface.co/HuggingFaceH4/zephyr-7b-beta
+"""
+register_template(
+    name="zephyr",
+    prefix=[
+        {"token": "<|system|>"},
+        "\n{{system}}",
+        {"token": "</s>"}
+    ],
+    prompt=[
+        {"token": "<|user|>"},
+        "\n{{query}}",
+        {"token": "</s>"},
+        {"token": "<|assistant|>"}
+    ],
+    system="You are a friendly chatbot who always responds in the style of a pirate",
+    sep=[]
+)
+
+
+r"""
+Supports: https://huggingface.co/IDEA-CCNL/Ziya-LLaMA-13B-v1
+          https://huggingface.co/IDEA-CCNL/Ziya-LLaMA-13B-v1.1
+          https://huggingface.co/IDEA-CCNL/Ziya2-13B-Chat
+"""
+register_template(
+    name="ziya",
+    prefix=[
+        "{{system}}"
+    ],
+    prompt=[
+        {"token": "<human>"},
+        ":{{query}}\n",
+        {"token": "<bot>"},
+        ":"
+    ],
+    system="",
+    sep=[
+        "\n"
+    ]
+)
--- a/src/llmtuner/hparams/init.py
+++ b/src/llmtuner/hparams/init.py
@@ -1,4 +1,5 @@
 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
--- a/src/llmtuner/hparams/data_args.py
+++ b/src/llmtuner/hparams/data_args.py
@@ -11,11 +11,17 @@ class DatasetAttr:
    dataset_name: Optional[str] = None
    dataset_sha1: Optional[str] = None
    system_prompt: Optional[str] = None
+    subset: Optional[str] = None
    ranking: Optional[bool] = False
+    formatting: Optional[Literal["alpaca", "sharegpt"]] = "alpaca"
+
    prompt: Optional[str] = "instruction"
    query: Optional[str] = "input"
    response: Optional[str] = "output"
    history: Optional[str] = None
+    messages: Optional[str] = "conversations"
+    role: Optional[str] = "from"
+    content: Optional[str] = "value"

    def __repr__(self) -> str:
        return self.dataset_name
@@ -36,7 +42,7 @@ class DataArguments:
    )
    dataset_dir: Optional[str] = field(
        default="data",
-        metadata={"help": "The name of the folder containing datasets."}
+        metadata={"help": "Path to the folder containing the datasets."}
    )
    split: Optional[str] = field(
        default="train",
@@ -60,7 +66,7 @@ class DataArguments:
    )
    mix_strategy: Optional[Literal["concat", "interleave_under", "interleave_over"]] = field(
        default="concat",
-        metadata={"help": "Strategy to use in dataset mixing."}
+        metadata={"help": "Strategy to use in dataset mixing (concat/interleave) (undersampling/oversampling)."}
    )
    interleave_probs: Optional[str] = field(
        default=None,
@@ -98,6 +104,10 @@ class DataArguments:
        default=False,
        metadata={"help": "Packing the questions and answers in the supervised fine-tuning stage."}
    )
+    cache_path: Optional[str] = field(
+        default=None,
+        metadata={"help": "Path to save or load the preprocessed datasets."}
+    )

    def __post_init__(self):
        if self.streaming and self.val_size > 1e-6 and self.val_size < 1:
@@ -106,12 +116,18 @@ class DataArguments:
        if self.streaming and self.max_samples is not None:
            raise ValueError("`max_samples` is incompatible with `streaming`.")

-    def init_for_training(self): # support mixing multiple datasets
+        if self.streaming and self.cache_path:
+            raise ValueError("`cache_path` is incompatible with `streaming`.")
+
+    def init_for_training(self, seed: int): # support mixing multiple datasets
+        self.seed = seed
        dataset_names = [ds.strip() for ds in self.dataset.split(",")] if self.dataset is not None else []
        try:
            with open(os.path.join(self.dataset_dir, "dataset_info.json"), "r") as f:
                dataset_info = json.load(f)
        except Exception:
+            if self.dataset is not None:
+                raise ValueError("Cannot find dataset_info.json in `dataset_dir`.")
            dataset_info = None

        prompt_list = self.system_prompt.split("|") if self.system_prompt else [None]
@@ -142,7 +158,12 @@ class DataArguments:
                dataset_attr.query = dataset_info[name]["columns"].get("query", None)
                dataset_attr.response = dataset_info[name]["columns"].get("response", None)
                dataset_attr.history = dataset_info[name]["columns"].get("history", None)
+                dataset_attr.messages = dataset_info[name]["columns"].get("messages", None)
+                dataset_attr.role = dataset_info[name]["columns"].get("role", None)
+                dataset_attr.content = dataset_info[name]["columns"].get("content", None)

+            dataset_attr.subset = dataset_info[name].get("subset", None)
            dataset_attr.ranking = dataset_info[name].get("ranking", False)
+            dataset_attr.formatting = dataset_info[name].get("formatting", "alpaca")
            dataset_attr.system_prompt = prompt_list[i]
            self.dataset_list.append(dataset_attr)
--- a/src/llmtuner/hparams/evaluation_args.py
+++ b/src/llmtuner/hparams/evaluation_args.py
@@ -0,0 +1,55 @@
+import os
+from typing import Literal, Optional
+from dataclasses import dataclass, field
+
+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: Optional[str] = field(
+        default="evaluation",
+        metadata={"help": "Path to the folder containing the evaluation datasets."}
+    )
+    batch_size: Optional[int] = field(
+        default=4,
+        metadata={"help": "The batch size per GPU for evaluation."}
+    )
+    seed: Optional[int] = field(
+        default=42,
+        metadata={"help": "Random seed to be used with data loaders."}
+    )
+    lang: Optional[Literal["en", "zh"]] = field(
+        default="en",
+        metadata={"help": "Language used at evaluation."}
+    )
+    n_shot: Optional[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: Optional[DownloadMode] = field(
+        default=DownloadMode.REUSE_DATASET_IF_EXISTS,
+        metadata={"help": "Download mode used for the evaluation datasets."}
+    )
+
+    def __post_init__(self):
+        task_available = []
+        for folder in os.listdir(self.task_dir):
+            if os.path.isdir(os.path.join(self.task_dir, folder)):
+                task_available.append(folder)
+
+        if self.task not in task_available:
+            raise ValueError("Task {} not found in {}.".format(self.task, self.task_dir))
+
+        if self.save_dir is not None and os.path.exists(self.save_dir):
+            raise ValueError("`save_dir` already exists, use another one.")
--- a/src/llmtuner/hparams/finetuning_args.py
+++ b/src/llmtuner/hparams/finetuning_args.py
@@ -12,7 +12,7 @@ class FinetuningArguments:
        default="sft",
        metadata={"help": "Which stage will be performed in training."}
    )
-    finetuning_type: Optional[Literal["lora", "freeze", "full", "none"]] = field(
+    finetuning_type: Optional[Literal["lora", "freeze", "full"]] = field(
        default="lora",
        metadata={"help": "Which fine-tuning method to use."}
    )
@@ -24,10 +24,10 @@ class FinetuningArguments:
        default="mlp",
        metadata={"help": "Name of trainable modules for partial-parameter (freeze) fine-tuning. \
                  LLaMA choices: [\"mlp\", \"self_attn\"], \
-                  BLOOM & Falcon & ChatGLM2 choices: [\"mlp\", \"self_attention\"], \
+                  BLOOM & Falcon & ChatGLM choices: [\"mlp\", \"self_attention\"], \
                  Qwen choices: [\"mlp\", \"attn\"], \
                  Phi-1.5 choices: [\"mlp\", \"mixer\"], \
-                  LLaMA-2, Baichuan, InternLM, XVERSE choices: the same as LLaMA."}
+                  LLaMA-2, BlueLM, Baichuan, InternLM, Mistral, Skywork, XVERSE, Yi choices: the same as LLaMA."}
    )
    lora_rank: Optional[int] = field(
        default=8,
@@ -45,11 +45,11 @@ class FinetuningArguments:
        default=None,
        metadata={"help": "Name(s) of target modules to apply LoRA. Use commas to separate multiple modules. \
                  LLaMA choices: [\"q_proj\", \"k_proj\", \"v_proj\", \"o_proj\", \"gate_proj\", \"up_proj\", \"down_proj\"], \
-                  BLOOM & Falcon & ChatGLM2 choices: [\"query_key_value\", \"self_attention.dense\", \"mlp.dense\"], \
+                  BLOOM & Falcon & ChatGLM choices: [\"query_key_value\", \"dense\", \"dense_h_to_4h\", \"dense_4h_to_h\"], \
                  Baichuan choices: [\"W_pack\", \"o_proj\", \"gate_proj\", \"up_proj\", \"down_proj\"], \
                  Qwen choices: [\"c_attn\", \"attn.c_proj\", \"w1\", \"w2\", \"mlp.c_proj\"], \
                  Phi-1.5 choices: [\"Wqkv\", \"out_proj\", \"fc1\", \"fc2\"], \
-                  LLaMA-2, InternLM, XVERSE choices: the same as LLaMA."}
+                  LLaMA-2, BlueLM, InternLM, Mistral, Skywork, XVERSE, Yi choices: the same as LLaMA."}
    )
    additional_target: Optional[str] = field(
        default=None,
@@ -75,6 +75,22 @@ class FinetuningArguments:
        default=0.1,
        metadata={"help": "The beta parameter for the DPO loss."}
    )
+    dpo_ref_model: Optional[str] = field(
+        default=None,
+        metadata={"help": "Path to the reference model used for the DPO training."}
+    )
+    dpo_ref_model_checkpoint: Optional[str] = field(
+        default=None,
+        metadata={"help": "Path to the directory(s) containing the model checkpoints of the reference model."}
+    )
+    upcast_layernorm: Optional[bool] = field(
+        default=False,
+        metadata={"help": "Whether to upcast the layernorm weights in fp32."}
+    )
+    neft_alpha: Optional[float] = field(
+        default=0,
+        metadata={"help": "The alpha parameter to control the noise magnitude in NEFTune."}
+    )

    def __post_init__(self):
        if isinstance(self.lora_target, str): # support custom target modules/layers of LoRA
@@ -83,7 +99,7 @@ class FinetuningArguments:
        if isinstance(self.additional_target, str):
            self.additional_target = [target.strip() for target in self.additional_target.split(",")]

-        assert self.finetuning_type in ["lora", "freeze", "full", "none"], "Invalid fine-tuning method."
+        assert self.finetuning_type in ["lora", "freeze", "full"], "Invalid fine-tuning method."

    def save_to_json(self, json_path: str):
        r"""Saves the content of this instance in JSON format inside `json_path`."""
--- a/src/llmtuner/hparams/general_args.py
+++ b/src/llmtuner/hparams/general_args.py
@@ -1,13 +0,0 @@
-from typing import Literal, Optional
-from dataclasses import dataclass, field
-
-
-@dataclass
-class GeneralArguments:
-    r"""
-    Arguments pertaining to which stage we are going to perform.
-    """
-    stage: Optional[Literal["pt", "sft", "rm", "ppo", "dpo"]] = field(
-        default="sft",
-        metadata={"help": "Which stage will be performed in training."}
-    )
--- a/src/llmtuner/hparams/generating_args.py
+++ b/src/llmtuner/hparams/generating_args.py
@@ -28,7 +28,7 @@ class GeneratingArguments:
        metadata={"help": "Number of beams for beam search. 1 means no beam search."}
    )
    max_length: Optional[int] = field(
-        default=None,
+        default=512,
        metadata={"help": "The maximum length the generated tokens can have. It can be overridden by max_new_tokens."}
    )
    max_new_tokens: Optional[int] = field(
@@ -46,6 +46,8 @@ class GeneratingArguments:

    def to_dict(self) -> Dict[str, Any]:
        args = asdict(self)
-        if args.get("max_new_tokens", None):
+        if args.get("max_new_tokens", -1) > 0:
            args.pop("max_length", None)
+        else:
+            args.pop("max_new_tokens", None)
        return args
--- a/src/llmtuner/hparams/model_args.py
+++ b/src/llmtuner/hparams/model_args.py
@@ -1,5 +1,5 @@
-from typing import Literal, Optional
-from dataclasses import dataclass, field
+from typing import Any, Dict, Literal, Optional
+from dataclasses import asdict, dataclass, field


@dataclass
@@ -22,10 +22,6 @@ class ModelArguments:
        default=False,
        metadata={"help": "Whether or not the special tokens should be split during the tokenization process."}
    )
-    use_auth_token: Optional[bool] = field(
-        default=False,
-        metadata={"help": "Will use the token generated when running `huggingface-cli login`."}
-    )
    model_revision: Optional[str] = field(
        default="main",
        metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."}
@@ -48,7 +44,7 @@ class ModelArguments:
    )
    checkpoint_dir: Optional[str] = field(
        default=None,
-        metadata={"help": "Path to the directory(s) containing the delta model checkpoints as well as the configurations."}
+        metadata={"help": "Path to the directory(s) containing the model checkpoints as well as the configurations."}
    )
    flash_attn: Optional[bool] = field(
        default=False,
@@ -58,19 +54,15 @@ class ModelArguments:
        default=False,
        metadata={"help": "Enable shift short attention (S^2-Attn) proposed by LongLoRA."}
    )
-    reward_model: Optional[str] = field(
+    reward_model: Optional[str] = field( # TODO: move it to FinetuningArguments
        default=None,
        metadata={"help": "Path to the directory containing the checkpoints of the reward model."}
    )
-    upcast_layernorm: Optional[bool] = field(
-        default=False,
-        metadata={"help": "Whether to upcast the layernorm weights in fp32."}
-    )
-    plot_loss: Optional[bool] = field(
+    plot_loss: Optional[bool] = field( # TODO: move it to FinetuningArguments
        default=False,
        metadata={"help": "Whether to plot the training loss after fine-tuning or not."}
    )
-    hf_auth_token: Optional[str] = field(
+    hf_hub_token: Optional[str] = field(
        default=None,
        metadata={"help": "Auth token to log in with Hugging Face Hub."}
    )
@@ -92,6 +84,5 @@ class ModelArguments:
        if self.quantization_bit is not None:
            assert self.quantization_bit in [4, 8], "We only accept 4-bit or 8-bit quantization."

-        if self.use_auth_token == True and self.hf_auth_token is not None:
-            from huggingface_hub.hf_api import HfFolder # lazy load
-            HfFolder.save_token(self.hf_auth_token)
+    def to_dict(self) -> Dict[str, Any]:
+        return asdict(self)
--- a/src/llmtuner/tuner/core/init.py
+++ b/src/llmtuner/tuner/core/init.py
@@ -1,2 +1,3 @@
 from llmtuner.tuner.core.parser import get_train_args, get_infer_args
 from llmtuner.tuner.core.loader import load_model_and_tokenizer
+from llmtuner.tuner.core.utils import generate_model_card
--- a/src/llmtuner/tuner/core/adapter.py
+++ b/src/llmtuner/tuner/core/adapter.py
@@ -1,6 +1,9 @@
+import os
 import torch
 from typing import TYPE_CHECKING

+from transformers.utils import cached_file
+from transformers.trainer import WEIGHTS_NAME, SAFE_WEIGHTS_NAME
 from peft import (
    PeftModel,
    TaskType,
@@ -23,8 +26,7 @@ def init_adapter(
    model: "PreTrainedModel",
    model_args: "ModelArguments",
    finetuning_args: "FinetuningArguments",
-    is_trainable: bool,
-    is_mergeable: bool
+    is_trainable: bool
 ) -> "PreTrainedModel":
    r"""
    Initializes the adapters.
@@ -34,14 +36,15 @@ def init_adapter(
    Note that the trainable parameters must be cast to float32.
    """

-    if finetuning_args.finetuning_type == "none" and is_trainable:
-        raise ValueError("You cannot use finetuning_type=none while training.")
+    if (not is_trainable) and model_args.checkpoint_dir is None:
+        logger.info("Checkpoint is not found at evaluation, load the original model.")
+        return model

    if finetuning_args.finetuning_type == "full" and is_trainable:
        logger.info("Fine-tuning method: Full")
        model = model.float()

-    if finetuning_args.finetuning_type == "freeze":
+    if finetuning_args.finetuning_type == "freeze" and is_trainable:
        logger.info("Fine-tuning method: Freeze")
        num_layers = getattr(model.config, "num_layers")
        if finetuning_args.num_layer_trainable > 0: # fine-tuning the last n layers if num_layer_trainable > 0
@@ -58,11 +61,11 @@ def init_adapter(

    if finetuning_args.finetuning_type == "lora":
        logger.info("Fine-tuning method: LoRA")
-        latest_checkpoint = None
+        checkpoint_to_resume = None

        if model_args.checkpoint_dir is not None:
-            if (is_trainable and finetuning_args.resume_lora_training) or (not is_mergeable): # continually fine-tuning
-                checkpoints_to_merge, latest_checkpoint = model_args.checkpoint_dir[:-1], model_args.checkpoint_dir[-1]
+            if is_trainable and finetuning_args.resume_lora_training:
+                checkpoints_to_merge, checkpoint_to_resume = model_args.checkpoint_dir[:-1], model_args.checkpoint_dir[-1]
            else:
                checkpoints_to_merge = model_args.checkpoint_dir

@@ -73,10 +76,10 @@ def init_adapter(
            if len(checkpoints_to_merge) > 0:
                logger.info("Merged {} model checkpoint(s).".format(len(checkpoints_to_merge)))

-            if latest_checkpoint is not None: # resume lora training or quantized inference
-                model = PeftModel.from_pretrained(model, latest_checkpoint, is_trainable=is_trainable)
+            if checkpoint_to_resume is not None: # resume lora training
+                model = PeftModel.from_pretrained(model, checkpoint_to_resume, is_trainable=is_trainable)

-        if is_trainable and latest_checkpoint is None: # create new lora weights while training
+        if is_trainable and checkpoint_to_resume is None: # create new lora weights while training
            if len(finetuning_args.lora_target) == 1 and finetuning_args.lora_target[0] == "all":
                target_modules = find_all_linear_modules(model, model_args.quantization_bit)
            else:
@@ -92,10 +95,35 @@ def init_adapter(
                modules_to_save=finetuning_args.additional_target
            )
            model = get_peft_model(model, lora_config)
-            if id(model.peft_config) != id(model.base_model.peft_config): # https://github.com/huggingface/peft/issues/923
-                model.base_model.peft_config = model.peft_config

    if model_args.checkpoint_dir is not None:
        logger.info("Loaded fine-tuned model from checkpoint(s): {}".format(",".join(model_args.checkpoint_dir)))

    return model
+
+
+def load_valuehead_params(
+    model: "PreTrainedModel",
+    model_args: "ModelArguments"
+) -> bool:
+    kwargs = {
+        "path_or_repo_id": model_args.reward_model,
+        "cache_dir": model_args.cache_dir,
+        "token": model_args.hf_hub_token,
+        "revision": model_args.model_revision
+    }
+    try:
+        vhead_file = cached_file(filename=WEIGHTS_NAME, **kwargs)
+    except:
+        try:
+            vhead_file = cached_file(filename=SAFE_WEIGHTS_NAME, **kwargs)
+        except:
+            logger.warning("Provided path ({}) does not contain valuehead weights.".format(model_args.reward_model))
+            return False
+
+    vhead_params = torch.load(vhead_file, map_location="cpu")
+    model.register_buffer("reward_head_weight", vhead_params["v_head.summary.weight"], persistent=False)
+    model.register_buffer("reward_head_bias", vhead_params["v_head.summary.bias"], persistent=False)
+    model.register_buffer("default_head_weight", torch.zeros_like(vhead_params["v_head.summary.weight"]), persistent=False)
+    model.register_buffer("default_head_bias", torch.zeros_like(vhead_params["v_head.summary.bias"]), persistent=False)
+    return True
--- a/src/llmtuner/tuner/core/loader.py
+++ b/src/llmtuner/tuner/core/loader.py
@@ -14,8 +14,8 @@ from transformers import (
    PreTrainedTokenizerBase
 )
 from transformers.models.llama import modeling_llama as LlamaModule
-from transformers.utils import check_min_version
 from transformers.utils.versions import require_version
+from peft import PeftModel
 from trl import AutoModelForCausalLMWithValueHead

 try:
@@ -26,9 +26,8 @@ except ImportError: # https://github.com/huggingface/transformers/releases/tag/v
 from llmtuner.extras.logging import reset_logging, get_logger
 from llmtuner.extras.misc import count_parameters, infer_optim_dtype
 from llmtuner.extras.patches import llama_patch as LlamaPatches
-from llmtuner.extras.save_and_load import load_valuehead_params
 from llmtuner.hparams import FinetuningArguments
-from llmtuner.tuner.core.adapter import init_adapter
+from llmtuner.tuner.core.adapter import init_adapter, load_valuehead_params
 from llmtuner.tuner.core.utils import prepare_model_for_training

 if TYPE_CHECKING:
@@ -39,11 +38,11 @@ if TYPE_CHECKING:
 logger = get_logger(__name__)


-check_min_version("4.31.0")
-require_version("datasets>=2.12.0", "To fix: pip install datasets>=2.12.0")
+require_version("transformers>=4.31.0,<4.35.0", "To fix: pip install \"transformers>=4.31.0,<4.35.0\"")
+require_version("datasets>=2.14.0", "To fix: pip install datasets>=2.14.0")
 require_version("accelerate>=0.21.0", "To fix: pip install accelerate>=0.21.0")
-require_version("peft>=0.4.0", "To fix: pip install peft>=0.4.0")
-require_version("trl>=0.7.1", "To fix: pip install trl>=0.7.1")
+require_version("peft>=0.6.0", "To fix: pip install peft>=0.6.0")
+require_version("trl>=0.7.4", "To fix: pip install trl>=0.7.4")


 def load_model_and_tokenizer(
@@ -57,15 +56,12 @@ def load_model_and_tokenizer(

    Support both training and inference.
    """
-    if (not is_trainable) and model_args.checkpoint_dir is None:
-        logger.warning("Checkpoint is not found at evaluation, load the original model.")
-        finetuning_args = FinetuningArguments(finetuning_type="none")

    config_kwargs = {
        "trust_remote_code": True,
        "cache_dir": model_args.cache_dir,
        "revision": model_args.model_revision,
-        "use_auth_token": True if model_args.use_auth_token else None,
+        "token": model_args.hf_hub_token
    }

    tokenizer = AutoTokenizer.from_pretrained(
@@ -83,7 +79,7 @@ def load_model_and_tokenizer(

    config = AutoConfig.from_pretrained(model_to_load, **config_kwargs)

-    # Fix tokenizer (for ChatGLM2)
+    # Fix tokenizer (for ChatGLM2 and ChatGLM3)
    if getattr(config, "model_type", None) == "chatglm":
        tokenizer._pad = MethodType(PreTrainedTokenizerBase._pad, tokenizer)

@@ -100,15 +96,9 @@ def load_model_and_tokenizer(

    # Set RoPE scaling
    if model_args.rope_scaling is not None:
-        if hasattr(config, "use_dynamic_ntk"): # for Qwen models
-            if is_trainable:
-                logger.warning("Qwen model does not support RoPE scaling in training.")
-            else:
-                setattr(config, "use_dynamic_ntk", True)
-                setattr(config, "use_logn_attn", True)
-                logger.info("Using dynamic NTK scaling.")
-
-        elif hasattr(config, "rope_scaling"): # for LLaMA and Falcon models
+        if not hasattr(config, "rope_scaling"):
+            logger.warning("Current model does not support RoPE scaling.")
+        else:
            if is_trainable:
                if model_args.rope_scaling == "dynamic":
                    logger.warning(
@@ -130,17 +120,17 @@ def load_model_and_tokenizer(
                model_args.rope_scaling, scaling_factor
            ))

-        else:
-            logger.warning("Current model does not support RoPE scaling.")
-
    # Set FlashAttention-2
    if model_args.flash_attn:
        if getattr(config, "model_type", None) == "llama":
-            LlamaModule.LlamaAttention = LlamaPatches.LlamaFlashAttention2
-            LlamaModule.LlamaModel._prepare_decoder_attention_mask = LlamaPatches._prepare_decoder_attention_mask
-            logger.info("Using FlashAttention-2 for faster training and inference.")
-        elif getattr(config, "model_type", None) == "qwen":
-            logger.info("Qwen models automatically enable FlashAttention if installed.")
+            if LlamaPatches.is_flash_attn_2_available:
+                LlamaModule.LlamaAttention = LlamaPatches.LlamaFlashAttention2
+                LlamaModule.LlamaModel._prepare_decoder_attention_mask = LlamaPatches._prepare_decoder_attention_mask
+                logger.info("Using FlashAttention-2 for faster training and inference.")
+            else:
+                logger.warning("FlashAttention-2 is not installed.")
+        elif getattr(config, "model_type", None) in ["qwen", "Yi"]:
+            logger.info("Current model automatically enables FlashAttention if installed.")
        else:
            logger.warning("Current model does not support FlashAttention-2.")
    elif is_trainable and model_args.shift_attn and getattr(config, "model_type", None) == "llama":
@@ -156,7 +146,6 @@ def load_model_and_tokenizer(
            logger.warning("Current model does not support shift short attention.")

    # Quantization configurations (using bitsandbytes library).
-    is_mergeable = True
    if model_args.quantization_bit is not None:
        if is_deepspeed_zero3_enabled():
            raise ValueError("DeepSpeed ZeRO-3 is incompatible with quantization.")
@@ -166,7 +155,7 @@ def load_model_and_tokenizer(
            config_kwargs["load_in_8bit"] = True
            config_kwargs["quantization_config"] = BitsAndBytesConfig(load_in_8bit=True)

-        elif model_args.quantization_bit == 4:
+        if model_args.quantization_bit == 4:
            require_version("bitsandbytes>=0.39.0", "To fix: pip install bitsandbytes>=0.39.0")
            config_kwargs["load_in_4bit"] = True
            config_kwargs["quantization_config"] = BitsAndBytesConfig(
@@ -176,7 +165,6 @@ def load_model_and_tokenizer(
                bnb_4bit_quant_type=model_args.quantization_type
            )

-        is_mergeable = False
        config_kwargs["device_map"] = {"": int(os.environ.get("LOCAL_RANK", "0"))} if is_trainable else "auto"
        logger.info("Quantizing model to {} bit.".format(model_args.quantization_bit))

@@ -193,9 +181,10 @@ def load_model_and_tokenizer(
    if isinstance(model, PreTrainedModel) and "GenerationMixin" not in str(model.generate.__func__):
        model.generate = MethodType(PreTrainedModel.generate, model)

-    # Fix LM head (for ChatGLM2)
+    # Fix LM head (for ChatGLM2 and ChatGLM3)
    if getattr(config, "model_type", None) == "chatglm":
        setattr(model, "lm_head", model.transformer.output_layer)
+        setattr(model, "_keys_to_ignore_on_save", ["lm_head.weight"])

    # Register auto class to save the custom code files.
    if isinstance(config, PretrainedConfig) and "AutoConfig" in getattr(config, "auto_map", {}):
@@ -206,19 +195,17 @@ def load_model_and_tokenizer(
        tokenizer.__class__.register_for_auto_class()

    # Initialize adapters
-    if is_trainable:
-        model = prepare_model_for_training(model, model_args.upcast_layernorm, finetuning_args.finetuning_type)
-    model = init_adapter(model, model_args, finetuning_args, is_trainable, is_mergeable)
+    model = prepare_model_for_training(model=model, finetuning_args=finetuning_args) if is_trainable else model
+    model = init_adapter(model, model_args, finetuning_args, is_trainable)
    model = model.train() if is_trainable else model.eval()

    # Prepare model with valuehead for RLHF
    if stage == "rm" or stage == "ppo":
-        model = AutoModelForCausalLMWithValueHead.from_pretrained(model)
-        model._keys_to_ignore_on_save = None
+        model: "AutoModelForCausalLMWithValueHead" = AutoModelForCausalLMWithValueHead.from_pretrained(model)
        reset_logging()
        if stage == "rm" and model_args.checkpoint_dir is not None: # load valuehead weights to evaluate reward model
            logger.warning("Only the last checkpoint containing valuehead will be loaded.")
-            if load_valuehead_params(model, model_args.checkpoint_dir[-1]):
+            if load_valuehead_params(model, model_args):
                model.v_head.load_state_dict({
                    "summary.weight": getattr(model, "reward_head_weight"),
                    "summary.bias": getattr(model, "reward_head_bias")
@@ -226,9 +213,12 @@ def load_model_and_tokenizer(

        if stage == "ppo": # load reward model
            logger.info("Load reward model from {}".format(model_args.reward_model))
-            if getattr(model, "is_peft_model", False):
+            if isinstance(model.pretrained_model, PeftModel):
                model.pretrained_model.load_adapter(model_args.reward_model, "reward")
-            assert load_valuehead_params(model, model_args.reward_model), "Reward model is not correctly loaded."
+            for name, param in model.named_parameters(): # https://github.com/huggingface/peft/issues/1090
+                if "default" in name:
+                    param.data = param.data.to(torch.float32) # trainable params should in fp32
+            assert load_valuehead_params(model, model_args), "Reward model is not correctly loaded."

    # Prepare model for inference
    if not is_trainable:
@@ -240,4 +230,7 @@ def load_model_and_tokenizer(
        trainable_params, all_param, 100 * trainable_params / all_param
    ))

+    if not is_trainable:
+        logger.info("This IS expected that the trainable params is 0 if you are using model for inference only.")
+
    return model, tokenizer
--- a/src/llmtuner/tuner/core/parser.py
+++ b/src/llmtuner/tuner/core/parser.py
@@ -1,5 +1,4 @@
 import os
-import sys
 import torch
 import datasets
 import transformers
@@ -8,6 +7,7 @@ from transformers import HfArgumentParser, Seq2SeqTrainingArguments
 from transformers.trainer_utils import get_last_checkpoint

 from llmtuner.extras.logging import get_logger
+from llmtuner.extras.misc import parse_args
 from llmtuner.hparams import (
    ModelArguments,
    DataArguments,
@@ -19,17 +19,6 @@ from llmtuner.hparams import (
 logger = get_logger(__name__)


-def _parse_args(parser: HfArgumentParser, args: Optional[Dict[str, Any]] = None) -> Tuple[Any]:
-    if args is not None:
-        return parser.parse_dict(args)
-    elif len(sys.argv) == 2 and sys.argv[1].endswith(".yaml"):
-        return parser.parse_yaml_file(os.path.abspath(sys.argv[1]))
-    elif len(sys.argv) == 2 and sys.argv[1].endswith(".json"):
-        return parser.parse_json_file(os.path.abspath(sys.argv[1]))
-    else:
-        return parser.parse_args_into_dataclasses()
-
-
 def parse_train_args(
    args: Optional[Dict[str, Any]] = None
 ) -> Tuple[
@@ -46,7 +35,7 @@ def parse_train_args(
        FinetuningArguments,
        GeneratingArguments
    ))
-    return _parse_args(parser, args)
+    return parse_args(parser, args)


 def parse_infer_args(
@@ -63,7 +52,7 @@ def parse_infer_args(
        FinetuningArguments,
        GeneratingArguments
    ))
-    return _parse_args(parser, args)
+    return parse_args(parser, args)


 def get_train_args(
@@ -88,8 +77,8 @@ def get_train_args(
    transformers.utils.logging.enable_default_handler()
    transformers.utils.logging.enable_explicit_format()

-    # Check arguments (do not check finetuning_args since it may be loaded from checkpoints)
-    data_args.init_for_training()
+    # Check arguments
+    data_args.init_for_training(training_args.seed)

    if finetuning_args.stage != "pt" and data_args.template is None:
        raise ValueError("Please specify which `template` to use.")
@@ -100,14 +89,16 @@ def get_train_args(
    if finetuning_args.stage == "sft" and training_args.do_predict and not training_args.predict_with_generate:
        raise ValueError("Please enable `predict_with_generate` to save model predictions.")

-    if finetuning_args.stage in ["rm", "ppo"] and finetuning_args.finetuning_type != "lora":
-        raise ValueError("RM and PPO stages can only be performed with the LoRA method.")
+    if finetuning_args.stage in ["rm", "ppo"]:
+        if finetuning_args.finetuning_type != "lora":
+            raise ValueError("RM and PPO stages can only be performed with the LoRA method.")
+        if training_args.resume_from_checkpoint is not None:
+            raise ValueError("RM and PPO stages do not support `resume_from_checkpoint`.")
+        if training_args.load_best_model_at_end:
+            raise ValueError("RM and PPO stages do not support `load_best_model_at_end`.")

-    if finetuning_args.stage in ["rm", "ppo"] and training_args.resume_from_checkpoint is not None:
-        raise ValueError("RM and PPO stages do not support `resume_from_checkpoint`.")
-
-    if finetuning_args.stage in ["ppo", "dpo"] and not training_args.do_train:
-        raise ValueError("PPO and DPO stages can only be performed at training.")
+    if finetuning_args.stage == "ppo" and not training_args.do_train:
+        raise ValueError("PPO training does not support evaluation.")

    if finetuning_args.stage in ["rm", "dpo"]:
        for dataset_attr in data_args.dataset_list:
@@ -117,9 +108,6 @@ def get_train_args(
    if finetuning_args.stage == "ppo" and model_args.reward_model is None:
        raise ValueError("Reward model is necessary for PPO training.")

-    if finetuning_args.stage == "ppo" and data_args.streaming:
-        raise ValueError("Streaming mode does not suppport PPO training currently.")
-
    if finetuning_args.stage == "ppo" and model_args.shift_attn:
        raise ValueError("PPO training is incompatible with S^2-Attn.")

@@ -135,18 +123,14 @@ def get_train_args(
    if model_args.quantization_bit is not None and finetuning_args.finetuning_type != "lora":
        raise ValueError("Quantization is only compatible with the LoRA method.")

-    if model_args.checkpoint_dir is not None:
-        if finetuning_args.finetuning_type != "lora" and len(model_args.checkpoint_dir) != 1:
-            raise ValueError("Only LoRA tuning accepts multiple checkpoints.")
+    if (
+        model_args.checkpoint_dir is not None
+        and len(model_args.checkpoint_dir) != 1
+        and finetuning_args.finetuning_type != "lora"
+    ):
+        raise ValueError("Only LoRA tuning accepts multiple checkpoints.")

-        if model_args.quantization_bit is not None:
-            if len(model_args.checkpoint_dir) != 1:
-                raise ValueError("Quantized model only accepts a single checkpoint. Merge them first.")
-            
-            if not finetuning_args.resume_lora_training:
-                raise ValueError("Quantized model cannot create new LoRA weight. Merge them first.")
-
-    if training_args.do_train and model_args.quantization_bit is not None and (not model_args.upcast_layernorm):
+    if training_args.do_train and model_args.quantization_bit is not None and (not finetuning_args.upcast_layernorm):
        logger.warning("We recommend enable `upcast_layernorm` in quantized training.")

    if training_args.do_train and (not training_args.fp16) and (not training_args.bf16):
@@ -219,11 +203,11 @@ def get_infer_args(
    if model_args.quantization_bit is not None and finetuning_args.finetuning_type != "lora":
        raise ValueError("Quantization is only compatible with the LoRA method.")

-    if model_args.checkpoint_dir is not None:
-        if finetuning_args.finetuning_type != "lora" and len(model_args.checkpoint_dir) != 1:
-            raise ValueError("Only LoRA tuning accepts multiple checkpoints.")
-
-        if model_args.quantization_bit is not None and len(model_args.checkpoint_dir) != 1:
-            raise ValueError("Quantized model only accepts a single checkpoint. Merge them first.")
+    if (
+        model_args.checkpoint_dir is not None
+        and len(model_args.checkpoint_dir) != 1
+        and finetuning_args.finetuning_type != "lora"
+    ):
+        raise ValueError("Only LoRA tuning accepts multiple checkpoints.")

    return model_args, data_args, finetuning_args, generating_args
--- a/src/llmtuner/tuner/core/utils.py
+++ b/src/llmtuner/tuner/core/utils.py
@@ -1,16 +1,20 @@
 import torch
-from typing import TYPE_CHECKING, List, Optional
+from typing import TYPE_CHECKING, Any, Dict, List, Optional, Set, Tuple

 from llmtuner.extras.constants import LAYERNORM_NAMES
+from llmtuner.extras.logging import get_logger

 if TYPE_CHECKING:
    from transformers.modeling_utils import PreTrainedModel
+    from llmtuner.hparams import ModelArguments, DataArguments, FinetuningArguments
+
+
+logger = get_logger(__name__)


 def find_all_linear_modules(
    model: "PreTrainedModel",
-    quantization_bit: Optional[int] = None,
-    output_layer_name: Optional[str] = "lm_head"
+    quantization_bit: Optional[int] = None
 ) -> List[str]:
    if quantization_bit is not None:
        import bitsandbytes as bnb
@@ -18,24 +22,41 @@ def find_all_linear_modules(
    else:
        linear_cls = torch.nn.Linear

+    output_layer_names = ["lm_head"]
+    if model.config.model_type == "chatglm":
+        output_layer_names.append("output_layer")
+
    module_names = set()
    for name, module in model.named_modules():
-        if output_layer_name not in name and isinstance(module, linear_cls):
+        if (
+            isinstance(module, linear_cls)
+            and not any([output_layer in name for output_layer in output_layer_names])
+        ):
            module_names.add(name.split(".")[-1])

-    if output_layer_name in module_names:
-        module_names.pop(output_layer_name)
-
+    logger.info("Found linear modules: {}".format(",".join(module_names)))
    return list(module_names)


+def generate_model_card(
+    model_args: "ModelArguments",
+    data_args: "DataArguments",
+    finetuning_args: "FinetuningArguments"
+) -> Dict[str, Any]:
+    return {
+        "tasks": "text-generation",
+        "finetuned_from": model_args.model_name_or_path,
+        "dataset": [dataset.strip() for dataset in data_args.dataset.split(",")],
+        "tags": ["llama-factory"] + (["lora"] if finetuning_args.finetuning_type == "lora" else [])
+    }
+
+
 def prepare_model_for_training(
    model: "PreTrainedModel",
-    upcast_layernorm: bool,
-    finetuning_type: str,
+    finetuning_args: "FinetuningArguments",
    output_layer_name: Optional[str] = "lm_head",
    use_gradient_checkpointing: Optional[bool] = True,
-    layernorm_names: Optional[List[str]] = LAYERNORM_NAMES
+    layernorm_names: Optional[Set[str]] = LAYERNORM_NAMES
 ) -> "PreTrainedModel":
    r"""
    Includes:
@@ -44,31 +65,43 @@ def prepare_model_for_training(
        (3) upcast the lm_head to fp32
    Inspired by: https://github.com/huggingface/peft/blob/v0.2.0/src/peft/utils/other.py#L33
    """
-    if upcast_layernorm:
+    if finetuning_args.upcast_layernorm:
        for name, param in model.named_parameters():
            if param.ndim == 1 and any(ln_name in name for ln_name in layernorm_names):
                param.data = param.data.to(torch.float32)
+        logger.info("Upcasting weights in layernorm in float32.")
+
+    if finetuning_args.neft_alpha > 1e-6:
+        def neftune_forward_hook(module: torch.nn.Module, args: Tuple[torch.Tensor], output: torch.Tensor):
+            if module.training:
+                dims = torch.tensor(output.size(1) * output.size(2))
+                mag_norm = finetuning_args.neft_alpha / torch.sqrt(dims)
+                output = output + torch.zeros_like(output).uniform_(-mag_norm, mag_norm)
+            return output
+
+        model.get_input_embeddings().register_forward_hook(neftune_forward_hook)
+        logger.info("Using noisy embedding with alpha={:.2f}".format(finetuning_args.neft_alpha))

    if use_gradient_checkpointing:
        if hasattr(model, "enable_input_require_grads"):
            model.enable_input_require_grads()
        else:
-            def make_inputs_require_grad(module, input, output):
+            def make_inputs_require_grad(module: torch.nn.Module, args: Tuple[torch.Tensor], output: torch.Tensor):
                output.requires_grad_(True)
            model.get_input_embeddings().register_forward_hook(make_inputs_require_grad)

        model.gradient_checkpointing_enable()
        model.config.use_cache = False # turn off when gradient checkpointing is enabled
+        logger.info("Gradient checkpointing enabled.")

-    if finetuning_type != "full" and hasattr(model, output_layer_name):
-        output_layer: torch.nn.Linear = getattr(model, output_layer_name)
-        input_dtype = output_layer.weight.dtype
-
-        class CastOutputToFloat(torch.nn.Sequential):
-
-            def forward(self, x: torch.Tensor) -> torch.Tensor:
-                return super().forward(x.to(input_dtype)).to(torch.float32)
-
-        setattr(model, output_layer_name, CastOutputToFloat(output_layer))
+    if finetuning_args.finetuning_type != "full" and hasattr(model, output_layer_name):
+        output_layer = getattr(model, output_layer_name)
+        if isinstance(output_layer, torch.nn.Linear):
+            def fp32_forward_pre_hook(module: torch.nn.Module, args: Tuple[torch.Tensor]):
+                return args[0].to(output_layer.weight.dtype)
+            def fp32_forward_post_hook(module: torch.nn.Module, args: Tuple[torch.Tensor], output: torch.Tensor):
+                return output.to(torch.float32)
+            output_layer.register_forward_pre_hook(fp32_forward_pre_hook)
+            output_layer.register_forward_hook(fp32_forward_post_hook)

    return model
--- a/src/llmtuner/tuner/dpo/trainer.py
+++ b/src/llmtuner/tuner/dpo/trainer.py
@@ -1,6 +1,6 @@
 import torch
 from collections import defaultdict
-from typing import TYPE_CHECKING, Dict, Optional, Tuple, Union
+from typing import TYPE_CHECKING, Dict, Literal, Optional, Tuple, Union
 from transformers import BatchEncoding, Trainer
 from trl import DPOTrainer
 from trl.trainer.utils import disable_dropout_in_model
@@ -19,6 +19,7 @@ class CustomDPOTrainer(DPOTrainer):
        model: Union["PreTrainedModel", torch.nn.Module],
        ref_model: Optional[Union["PreTrainedModel", torch.nn.Module]] = None,
        disable_dropout: Optional[bool] = True,
+        loss_type: Optional[Literal["sigmoid", "hinge"]] = "sigmoid",
        **kwargs
    ):
        if disable_dropout:
@@ -29,9 +30,11 @@ class CustomDPOTrainer(DPOTrainer):
        self.is_encoder_decoder = model.config.is_encoder_decoder
        self.ref_model = ref_model
        self.use_dpo_data_collator = True # hack to avoid warning
+        self.generate_during_eval = False # disable at evaluation
        self.label_pad_token_id = IGNORE_INDEX
        self.padding_value = 0
        self.beta = beta
+        self.loss_type = loss_type
        self._stored_metrics = defaultdict(lambda: defaultdict(list))

        Trainer.__init__(self, model=model, **kwargs)
@@ -40,8 +43,7 @@ class CustomDPOTrainer(DPOTrainer):

        if ref_model is not None:
            if self.is_deepspeed_enabled:
-                self.ref_model, = self.accelerator._prepare_deepspeed(self.ref_model)
-                self.ref_model.eval()
+                self.ref_model = self._prepare_deepspeed(self.ref_model)
            else:
                self.ref_model = self.accelerator.prepare_model(self.ref_model, evaluation_mode=True)

--- a/src/llmtuner/tuner/dpo/workflow.py
+++ b/src/llmtuner/tuner/dpo/workflow.py
@@ -1,20 +1,24 @@
 # Inspired by: https://github.com/huggingface/trl/blob/main/examples/research_projects/stack_llama_2/scripts/dpo_llama2.py

-from copy import deepcopy
 from peft import PeftModel
 from typing import TYPE_CHECKING, Optional, List
 from transformers import Seq2SeqTrainingArguments

 from llmtuner.dsets import get_dataset, preprocess_dataset, split_dataset
 from llmtuner.extras.constants import IGNORE_INDEX
+from llmtuner.extras.logging import get_logger
 from llmtuner.extras.ploting import plot_loss
-from llmtuner.tuner.core import load_model_and_tokenizer
+from llmtuner.hparams import ModelArguments
+from llmtuner.tuner.core import generate_model_card, load_model_and_tokenizer
 from llmtuner.tuner.dpo.collator import DPODataCollatorWithPadding
 from llmtuner.tuner.dpo.trainer import CustomDPOTrainer

 if TYPE_CHECKING:
    from transformers import TrainerCallback
-    from llmtuner.hparams import ModelArguments, DataArguments, FinetuningArguments
+    from llmtuner.hparams import DataArguments, FinetuningArguments
+
+
+logger = get_logger(__name__)


 def run_dpo(
@@ -33,6 +37,26 @@ def run_dpo(
        label_pad_token_id=IGNORE_INDEX if data_args.ignore_pad_token_for_loss else tokenizer.pad_token_id
    )

+    # Create reference model
+    if finetuning_args.dpo_ref_model is not None:
+        ref_model_args_dict = model_args.to_dict()
+        ref_model_args_dict.update(dict(
+            model_name_or_path=finetuning_args.dpo_ref_model,
+            checkpoint_dir=finetuning_args.dpo_ref_model_checkpoint
+        ))
+        ref_model_args = ModelArguments(**ref_model_args_dict)
+        ref_model, _ = load_model_and_tokenizer(ref_model_args, finetuning_args, is_trainable=False, stage="sft")
+        logger.info("Created reference model from {}".format(finetuning_args.dpo_ref_model))
+    elif training_args.do_train:
+        if isinstance(model, PeftModel):
+            ref_model = None
+        else:
+            ref_model, _ = load_model_and_tokenizer(model_args, finetuning_args, is_trainable=False, stage="sft")
+            logger.info("Created reference model from the model itself.")
+    else:
+        ref_model = model
+
+    # Update arguments
    training_args_dict = training_args.to_dict()
    training_args_dict.update(dict(remove_unused_columns=False)) # important for pairwise dataset
    training_args = Seq2SeqTrainingArguments(**training_args_dict)
@@ -41,7 +65,7 @@ def run_dpo(
    trainer = CustomDPOTrainer(
        beta=finetuning_args.dpo_beta,
        model=model,
-        ref_model=deepcopy(model) if not isinstance(model, PeftModel) else None,
+        ref_model=ref_model,
        args=training_args,
        tokenizer=tokenizer,
        data_collator=data_collator,
@@ -52,9 +76,27 @@ def run_dpo(
    # Training
    if training_args.do_train:
        train_result = trainer.train(resume_from_checkpoint=training_args.resume_from_checkpoint)
+        trainer.save_model()
        trainer.log_metrics("train", train_result.metrics)
        trainer.save_metrics("train", train_result.metrics)
        trainer.save_state()
-        trainer.save_model()
        if trainer.is_world_process_zero() and model_args.plot_loss:
            plot_loss(training_args.output_dir, keys=["loss", "eval_loss"])
+
+    # Evaluation
+    if training_args.do_eval:
+        metrics = trainer.evaluate(metric_key_prefix="eval")
+        if id(model) == id(ref_model): # unable to compute rewards without a reference model
+            logger.warning("Pass `dpo_ref_model` for computing rewards at evaluation.")
+            remove_keys = [key for key in metrics.keys() if "rewards" in key]
+            for key in remove_keys:
+                metrics.pop(key)
+        trainer.log_metrics("eval", metrics)
+        trainer.save_metrics("eval", metrics)
+
+    # Create model card
+    if training_args.do_train:
+        if training_args.push_to_hub:
+            trainer.push_to_hub(**generate_model_card(model_args, data_args, finetuning_args))
+        else:
+            trainer.create_model_card(**generate_model_card(model_args, data_args, finetuning_args))
--- a/src/llmtuner/tuner/ppo/trainer.py
+++ b/src/llmtuner/tuner/ppo/trainer.py
@@ -1,10 +1,11 @@
 import os
+import sys
 import math
 import torch
 from tqdm import tqdm
-from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple
+from typing import TYPE_CHECKING, Dict, List, Optional, Tuple

-from transformers import GenerationConfig, Trainer, TrainerState, TrainerControl
+from transformers import BatchEncoding, GenerationConfig, Trainer, TrainerState, TrainerControl
 from transformers.trainer_utils import PREFIX_CHECKPOINT_DIR

 from trl import PPOTrainer
@@ -18,7 +19,7 @@ from llmtuner.tuner.ppo.utils import dump_layernorm, restore_layernorm, replace_
 if TYPE_CHECKING:
    from transformers import Seq2SeqTrainingArguments, TrainerCallback
    from trl import AutoModelForCausalLMWithValueHead
-    from llmtuner.hparams import ModelArguments, GeneratingArguments
+    from llmtuner.hparams import ModelArguments, FinetuningArguments, GeneratingArguments


 logger = get_logger(__name__)
@@ -33,16 +34,15 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
        self,
        model_args: "ModelArguments",
        training_args: "Seq2SeqTrainingArguments",
+        finetuning_args: "FinetuningArguments",
        generating_args: "GeneratingArguments",
        callbacks: List["TrainerCallback"],
        **kwargs
    ):
        PPOTrainer.__init__(self, **kwargs)
-        if getattr(self.accelerator.state, "deepspeed_plugin", None) is not None:
-            raise ValueError("PPOTrainer is incompatible with DeepSpeed.")
-
        self.args = training_args
        self.model_args = model_args
+        self.finetuning_args = finetuning_args
        self.generation_config = GenerationConfig(
            pad_token_id=self.tokenizer.pad_token_id,
            eos_token_id=[self.tokenizer.eos_token_id] + self.tokenizer.additional_special_tokens_ids,
@@ -52,6 +52,8 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
        self.control = TrainerControl()
        self.log_callback, self.save_callback = callbacks[0], callbacks[1]
        assert isinstance(self.log_callback, LogCallback) and isinstance(self.save_callback, SavePeftModelCallback)
+        if self.args.max_steps > 0:
+            logger.info("max_steps is given, it will override any value given in num_train_epochs")

    def ppo_train(self) -> None:
        r"""
@@ -60,10 +62,17 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
        total_train_batch_size = (
            self.args.per_device_train_batch_size * self.args.gradient_accumulation_steps * self.args.world_size
        )
-        len_dataloader = len(self.dataloader)
-        num_examples = len(self.dataset)
-        num_train_epochs = self.args.num_train_epochs
-        max_steps = math.ceil(num_train_epochs * len_dataloader)
+        if self.args.max_steps > 0:
+            num_examples = total_train_batch_size * self.args.max_steps
+            num_train_epochs = sys.maxsize
+            max_steps = self.args.max_steps
+            steps_in_epoch = self.args.max_steps * self.args.gradient_accumulation_steps
+        else:
+            len_dataloader = len(self.dataloader)
+            num_examples = len(self.dataset)
+            num_train_epochs = self.args.num_train_epochs
+            max_steps = math.ceil(num_train_epochs * len_dataloader)
+            steps_in_epoch = len_dataloader

        self.state.max_steps = max_steps
        self.state.num_train_epochs = num_train_epochs
@@ -82,14 +91,16 @@ class CustomPPOTrainer(PPOTrainer, Trainer):

        unwrapped_model: "AutoModelForCausalLMWithValueHead" = self.accelerator.unwrap_model(self.model)
        dataiter = iter(self.dataloader)
-        steps_trained = 0
        loss_meter = AverageMeter()
        reward_meter = AverageMeter()
        self.log_callback.on_train_begin(self.args, self.state, self.control)

        for step in tqdm(range(max_steps), disable=not self.is_local_process_zero()):
-            batch = next(dataiter)
-            steps_trained += 1
+            try:
+                batch = next(dataiter)
+            except StopIteration:
+                dataiter = iter(self.dataloader)
+                batch = next(dataiter)

            # Cast to inference mode
            unwrapped_model.gradient_checkpointing_disable()
@@ -97,9 +108,14 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
            self.model.eval()

            # Get inputs
-            queries, responses = self.get_inputs(batch)
            self.tokenizer.padding_side = "right" # change padding side
-            rewards = self.get_rewards(queries, responses, unwrapped_model)
+            queries, responses, rewards = [], [], []
+            for idx in range(0, self.config.batch_size, self.config.mini_batch_size):
+                mini_batch_queries, mini_batch_responses = self.get_inputs(batch[idx:idx+self.config.mini_batch_size])
+                mini_batch_rewards = self.get_rewards(mini_batch_queries, mini_batch_responses, unwrapped_model)
+                queries.extend(mini_batch_queries)
+                responses.extend(mini_batch_responses)
+                rewards.extend(mini_batch_rewards)

            # Cast to training mode
            unwrapped_model.gradient_checkpointing_enable()
@@ -128,7 +144,7 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
                    loss=round(loss_meter.avg, 4),
                    reward=round(reward_meter.avg, 4),
                    learning_rate=stats["ppo/learning_rate"],
-                    epoch=round(step / len_dataloader, 2)
+                    epoch=round(step / steps_in_epoch, 2)
                )
                tqdm.write(str(logs))
                logs["step"] = step
@@ -148,21 +164,17 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
            if self.control.should_epoch_stop or self.control.should_training_stop:
                break

-            if steps_trained == len_dataloader:
-                dataiter = iter(self.dataloader)
-                steps_trained = 0
-
        self.log_callback.on_train_end(self.args, self.state, self.control)
        self.save_callback.on_train_end(
            self.args, self.state, self.control, model=self.accelerator.unwrap_model(self.model)
        )

    @torch.no_grad()
-    def get_inputs(self, batch: Dict[str, torch.Tensor]) -> Tuple[List[torch.Tensor], List[torch.Tensor]]:
+    def get_inputs(self, batch: BatchEncoding) -> Tuple[List[torch.Tensor], List[torch.Tensor]]:
        r"""
        Generates model's responses given queries.
        """
-        if self.model_args.upcast_layernorm:
+        if self.finetuning_args.upcast_layernorm:
            layernorm_params = dump_layernorm(self.model)

        unwrapped_model: "AutoModelForCausalLMWithValueHead" = self.accelerator.unwrap_model(self.model)
@@ -172,21 +184,19 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
            **batch
        )

-        if self.model_args.upcast_layernorm:
+        if self.finetuning_args.upcast_layernorm:
            restore_layernorm(self.model, layernorm_params)

        query, response = batch["input_ids"].detach().cpu(), response[:, batch["input_ids"].size(-1):].detach().cpu()
        queries, responses = [], []
        for i in range(len(query)):
-            query_length = (query[i] != self.tokenizer.pad_token_id).nonzero()[0]
+            query_length = (query[i] != self.tokenizer.pad_token_id).nonzero()[0].item()
            response_index = (response[i] != self.tokenizer.pad_token_id).nonzero()

            if len(response_index) == 0:
                response_length = 1 # allow empty response
-            elif self.tokenizer.pad_token_id == self.tokenizer.eos_token_id:
-                response_length = response_index[-1] + 2 # save the EOS token
            else:
-                response_length = response_index[-1] + 1
+                response_length = response_index[-1].item() + 1

            queries.append(query[i, query_length:]) # remove padding from left
            responses.append(response[i, :response_length]) # remove padding from right
@@ -214,13 +224,14 @@ class CustomPPOTrainer(PPOTrainer, Trainer):

        rewards = []
        for i in range(values.size(0)):
-            end_index = batch["attention_mask"][i].nonzero()[-1] # use the score on the EOS token
+            end_indexes = (batch["input_ids"][i] != self.tokenizer.pad_token_id).nonzero()
+            end_index = end_indexes[-1].item() if len(end_indexes) else 0
            rewards.append(values[i, end_index].float().detach().cpu()) # use fp32 type

        replace_model(unwrapped_model, target="default")
        return rewards

-    @PPODecorators.empty_cuda_cache()
+    @PPODecorators.empty_device_cache()
    def batched_forward_pass(
        self,
        model: "AutoModelForCausalLMWithValueHead",
@@ -264,7 +275,7 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
            for j in range(len(query_batch)):
                start = len(query_batch[j]) - 1
                if attention_mask[j, 0] == 0: # offset left padding
-                    start += attention_mask[j, :].nonzero()[0]
+                    start += attention_mask[j, :].nonzero()[0].item()
                end = start + len(response_batch[j])

                if response_masks is not None:
--- a/src/llmtuner/tuner/ppo/workflow.py
+++ b/src/llmtuner/tuner/ppo/workflow.py
@@ -42,7 +42,7 @@ def run_ppo(
        ppo_epochs=1,
        max_grad_norm=training_args.max_grad_norm,
        seed=training_args.seed,
-        optimize_cuda_cache=True,
+        optimize_device_cache=True,
        target=finetuning_args.ppo_target,
        log_with=finetuning_args.ppo_logger,
        use_score_scaling=finetuning_args.ppo_score_norm,
@@ -51,10 +51,14 @@ def run_ppo(
    )

    optimizer = AdamW(filter(lambda p: p.requires_grad, model.parameters()), lr=training_args.learning_rate)
-    total_train_batch_size = (
-        training_args.per_device_train_batch_size * training_args.gradient_accumulation_steps * training_args.world_size
-    )
-    num_training_steps = training_args.num_train_epochs * math.ceil(len(dataset) / total_train_batch_size)
+    if training_args.max_steps > 0:
+        num_training_steps = training_args.max_steps
+    else:
+        total_train_batch_size = (
+            training_args.per_device_train_batch_size * training_args.gradient_accumulation_steps * training_args.world_size
+        )
+        num_training_steps = training_args.num_train_epochs * math.ceil(len(dataset) / total_train_batch_size)
+
    lr_scheduler = get_scheduler(
        training_args.lr_scheduler_type,
        optimizer=optimizer,
@@ -66,6 +70,7 @@ def run_ppo(
    ppo_trainer = CustomPPOTrainer(
        model_args=model_args,
        training_args=training_args,
+        finetuning_args=finetuning_args,
        generating_args=generating_args,
        callbacks=callbacks + [SavePeftModelCallback()],
        config=ppo_config,
--- a/src/llmtuner/tuner/pt/workflow.py
+++ b/src/llmtuner/tuner/pt/workflow.py
@@ -1,4 +1,4 @@
-# Inspired by: https://github.com/huggingface/transformers/blob/v4.29.2/examples/pytorch/language-modeling/run_clm.py
+# Inspired by: https://github.com/huggingface/transformers/blob/v4.34.1/examples/pytorch/language-modeling/run_clm.py

 import math
 from typing import TYPE_CHECKING, Optional, List
@@ -6,7 +6,7 @@ from transformers import DataCollatorForLanguageModeling, Trainer

 from llmtuner.dsets import get_dataset, preprocess_dataset, split_dataset
 from llmtuner.extras.ploting import plot_loss
-from llmtuner.tuner.core import load_model_and_tokenizer
+from llmtuner.tuner.core import generate_model_card, load_model_and_tokenizer

 if TYPE_CHECKING:
    from transformers import Seq2SeqTrainingArguments, TrainerCallback
@@ -38,10 +38,10 @@ def run_pt(
    # Training
    if training_args.do_train:
        train_result = trainer.train(resume_from_checkpoint=training_args.resume_from_checkpoint)
+        trainer.save_model()
        trainer.log_metrics("train", train_result.metrics)
        trainer.save_metrics("train", train_result.metrics)
        trainer.save_state()
-        trainer.save_model()
        if trainer.is_world_process_zero() and model_args.plot_loss:
            plot_loss(training_args.output_dir, keys=["loss", "eval_loss"])

@@ -56,3 +56,10 @@ def run_pt(
        metrics["perplexity"] = perplexity
        trainer.log_metrics("eval", metrics)
        trainer.save_metrics("eval", metrics)
+
+    # Create model card
+    if training_args.do_train:
+        if training_args.push_to_hub:
+            trainer.push_to_hub(**generate_model_card(model_args, data_args, finetuning_args))
+        else:
+            trainer.create_model_card(**generate_model_card(model_args, data_args, finetuning_args))
--- a/src/llmtuner/tuner/rm/trainer.py
+++ b/src/llmtuner/tuner/rm/trainer.py
@@ -45,9 +45,6 @@ class PairwiseTrainer(Trainer):
        # Split the inputs and rewards into two parts, chosen and rejected
        batch_size = inputs["input_ids"].size(0) // 2
        chosen_input_ids, rejected_input_ids = inputs["input_ids"][:batch_size], inputs["input_ids"][batch_size:]
-        chosen_attn_mask, rejected_attn_mask = (
-            inputs["attention_mask"][:batch_size], inputs["attention_mask"][batch_size:]
-        )
        chosen_rewards, rejected_rewards = values[:batch_size], values[batch_size:]
        chosen_scores, rejected_scores = [], []

@@ -55,8 +52,8 @@ class PairwiseTrainer(Trainer):
        # Inspired by: https://github.com/CarperAI/trlx/blob/main/examples/summarize_rlhf/reward_model/reward_model.py
        loss = 0
        for i in range(batch_size):
-            chosen_length = chosen_attn_mask[i].nonzero()[-1] + 1
-            rejected_length = rejected_attn_mask[i].nonzero()[-1] + 1
+            chosen_length = (chosen_input_ids[i] != self.tokenizer.pad_token_id).nonzero()[-1] + 1
+            rejected_length = (rejected_input_ids[i] != self.tokenizer.pad_token_id).nonzero()[-1] + 1
            check_divergence = (chosen_input_ids[i] != rejected_input_ids[i]).nonzero()

            if len(check_divergence) == 0:
@@ -69,7 +66,7 @@ class PairwiseTrainer(Trainer):
            assert div_index > 0
            chosen_trunc_rewards = chosen_rewards[i, div_index:end_index]
            rejected_trunc_rewards = rejected_rewards[i, div_index:end_index]
-            if return_outputs: # use the score on the EOS token for inference
+            if return_outputs: # use the score on the last token except pad token for inference
                chosen_scores.append(chosen_rewards[i, chosen_length-1])
                rejected_scores.append(rejected_rewards[i, rejected_length-1])
            loss += -torch.nn.functional.logsigmoid(chosen_trunc_rewards - rejected_trunc_rewards).mean()
@@ -95,7 +92,6 @@ class PairwiseTrainer(Trainer):

        output_prediction_file = os.path.join(self.args.output_dir, "generated_predictions.jsonl")
        logger.info(f"Saving prediction results to {output_prediction_file}")
-
        chosen_scores, rejected_scores = predict_results.predictions

        with open(output_prediction_file, "w", encoding="utf-8") as writer:
--- a/src/llmtuner/tuner/rm/workflow.py
+++ b/src/llmtuner/tuner/rm/workflow.py
@@ -1,5 +1,4 @@
-# Inspired by:
-# https://github.com/CarperAI/trlx/blob/main/examples/summarize_rlhf/reward_model/train_reward_model_gptj.py
+# Inspired by: https://github.com/CarperAI/trlx/blob/main/examples/summarize_rlhf/reward_model/train_reward_model_gptj.py

 from typing import TYPE_CHECKING, Optional, List
 from transformers import Seq2SeqTrainingArguments
@@ -7,7 +6,7 @@ from transformers import Seq2SeqTrainingArguments
 from llmtuner.dsets import get_dataset, preprocess_dataset, split_dataset
 from llmtuner.extras.callbacks import SavePeftModelCallback
 from llmtuner.extras.ploting import plot_loss
-from llmtuner.tuner.core import load_model_and_tokenizer
+from llmtuner.tuner.core import generate_model_card, load_model_and_tokenizer
 from llmtuner.tuner.rm.metric import compute_accuracy
 from llmtuner.tuner.rm.collator import PairwiseDataCollatorWithPadding
 from llmtuner.tuner.rm.trainer import PairwiseTrainer
@@ -29,6 +28,7 @@ def run_rm(
    dataset = preprocess_dataset(dataset, tokenizer, data_args, training_args, stage="rm")
    data_collator = PairwiseDataCollatorWithPadding(tokenizer, pad_to_multiple_of=4)

+    # Update arguments
    training_args_dict = training_args.to_dict()
    training_args_dict.update(dict(remove_unused_columns=False)) # important for pairwise dataset
    training_args = Seq2SeqTrainingArguments(**training_args_dict)
@@ -47,10 +47,10 @@ def run_rm(
    # Training
    if training_args.do_train:
        train_result = trainer.train()
+        trainer.save_model()
        trainer.log_metrics("train", train_result.metrics)
        trainer.save_metrics("train", train_result.metrics)
        trainer.save_state()
-        trainer.save_model()
        if trainer.is_world_process_zero() and model_args.plot_loss:
            plot_loss(training_args.output_dir, keys=["loss", "eval_loss"])

@@ -66,3 +66,10 @@ def run_rm(
        trainer.log_metrics("predict", predict_results.metrics)
        trainer.save_metrics("predict", predict_results.metrics)
        trainer.save_predictions(predict_results)
+
+    # Create model card
+    if training_args.do_train:
+        if training_args.push_to_hub:
+            trainer.push_to_hub(**generate_model_card(model_args, data_args, finetuning_args))
+        else:
+            trainer.create_model_card(**generate_model_card(model_args, data_args, finetuning_args))
--- a/src/llmtuner/tuner/sft/trainer.py
+++ b/src/llmtuner/tuner/sft/trainer.py
@@ -33,28 +33,20 @@ class CustomSeq2SeqTrainer(Seq2SeqTrainer):

        Subclass and override to inject custom behavior.
        """
+        labels = inputs["labels"].detach().clone() if "labels" in inputs else None # backup labels
        if self.args.predict_with_generate:
            assert self.tokenizer.padding_side == "left", "This method only accepts left-padded tensor."
-            assert self.tokenizer.pad_token_id is not None, "Pad token is required."
            prompt_len, label_len = inputs["input_ids"].size(-1), inputs["labels"].size(-1)
            if prompt_len > label_len:
                inputs["labels"] = self._pad_tensors_to_target_len(inputs["labels"], inputs["input_ids"])
            if label_len > prompt_len:
-                inputs["input_ids"] = self._pad_tensors_to_target_len(inputs["input_ids"], inputs["labels"])
-                if "attention_mask" in inputs:
-                    inputs["attention_mask"] = self._pad_tensors_to_target_len(
-                        inputs["attention_mask"], inputs["labels"], pad_token_id=0
-                    )
-                if "position_ids" in inputs:
-                    inputs["position_ids"] = self._pad_tensors_to_target_len(
-                        inputs["position_ids"], inputs["labels"], pad_token_id=0
-                    )
+                inputs["labels"] = inputs["labels"][:, :prompt_len] # truncate the labels instead of padding the inputs

-        loss, generated_tokens, labels = super().prediction_step(
+        loss, generated_tokens, _ = super().prediction_step(
            model, inputs, prediction_loss_only=prediction_loss_only, ignore_keys=ignore_keys
        )
        if generated_tokens is not None and self.args.predict_with_generate:
-            generated_tokens[:, :max(prompt_len, label_len)] = self.tokenizer.pad_token_id
+            generated_tokens[:, :prompt_len] = self.tokenizer.pad_token_id
            generated_tokens = generated_tokens.contiguous()

        return loss, generated_tokens, labels
@@ -62,14 +54,13 @@ class CustomSeq2SeqTrainer(Seq2SeqTrainer):
    def _pad_tensors_to_target_len(
        self,
        src_tensor: torch.Tensor,
-        tgt_tensor: torch.Tensor,
-        pad_token_id: Optional[int] = None
+        tgt_tensor: torch.Tensor
    ) -> torch.Tensor:
        r"""
        Pads the tensor to the same length as the target tensor.
        """
-        pad_token_id = pad_token_id if pad_token_id is not None else self.tokenizer.pad_token_id
-        padded_tensor = pad_token_id * torch.ones_like(tgt_tensor)
+        assert self.tokenizer.pad_token_id is not None, "Pad token is required."
+        padded_tensor = self.tokenizer.pad_token_id * torch.ones_like(tgt_tensor)
        padded_tensor[:, -src_tensor.shape[-1]:] = src_tensor # adopt left-padding
        return padded_tensor.contiguous() # in contiguous memory

--- a/src/llmtuner/tuner/sft/workflow.py
+++ b/src/llmtuner/tuner/sft/workflow.py
@@ -1,4 +1,4 @@
-# Inspired by: https://github.com/huggingface/transformers/blob/v4.29.2/examples/pytorch/summarization/run_summarization.py
+# Inspired by: https://github.com/huggingface/transformers/blob/v4.34.1/examples/pytorch/summarization/run_summarization.py

 from typing import TYPE_CHECKING, Optional, List
 from transformers import DataCollatorForSeq2Seq, Seq2SeqTrainingArguments
@@ -7,7 +7,7 @@ from llmtuner.dsets import get_dataset, preprocess_dataset, split_dataset
 from llmtuner.extras.constants import IGNORE_INDEX
 from llmtuner.extras.misc import get_logits_processor
 from llmtuner.extras.ploting import plot_loss
-from llmtuner.tuner.core import load_model_and_tokenizer
+from llmtuner.tuner.core import generate_model_card, load_model_and_tokenizer
 from llmtuner.tuner.sft.metric import ComputeMetrics
 from llmtuner.tuner.sft.trainer import CustomSeq2SeqTrainer

@@ -33,7 +33,7 @@ def run_sft(

    data_collator = DataCollatorForSeq2Seq(
        tokenizer=tokenizer,
-        pad_to_multiple_of=4, # for shift short attention
+        pad_to_multiple_of=4 if tokenizer.padding_side == "right" else None, # for shift short attention
        label_pad_token_id=IGNORE_INDEX if data_args.ignore_pad_token_for_loss else tokenizer.pad_token_id
    )

@@ -65,10 +65,10 @@ def run_sft(
    # Training
    if training_args.do_train:
        train_result = trainer.train(resume_from_checkpoint=training_args.resume_from_checkpoint)
+        trainer.save_model()
        trainer.log_metrics("train", train_result.metrics)
        trainer.save_metrics("train", train_result.metrics)
        trainer.save_state()
-        trainer.save_model()
        if trainer.is_world_process_zero() and model_args.plot_loss:
            plot_loss(training_args.output_dir, keys=["loss", "eval_loss"])

@@ -88,3 +88,10 @@ def run_sft(
        trainer.log_metrics("predict", predict_results.metrics)
        trainer.save_metrics("predict", predict_results.metrics)
        trainer.save_predictions(predict_results)
+
+    # Create model card
+    if training_args.do_train:
+        if training_args.push_to_hub:
+            trainer.push_to_hub(**generate_model_card(model_args, data_args, finetuning_args))
+        else:
+            trainer.create_model_card(**generate_model_card(model_args, data_args, finetuning_args))
--- a/src/llmtuner/tuner/tune.py
+++ b/src/llmtuner/tuner/tune.py
@@ -38,10 +38,10 @@ def export_model(args: Optional[Dict[str, Any]] = None, max_shard_size: Optional
    model_args, _, finetuning_args, _ = get_infer_args(args)
    model, tokenizer = load_model_and_tokenizer(model_args, finetuning_args)
    model.config.use_cache = True
-    tokenizer.padding_side = "left" # restore padding side
-    tokenizer.init_kwargs["padding_side"] = "left"
    model.save_pretrained(model_args.export_dir, max_shard_size=max_shard_size)
    try:
+        tokenizer.padding_side = "left" # restore padding side
+        tokenizer.init_kwargs["padding_side"] = "left"
        tokenizer.save_pretrained(model_args.export_dir)
    except:
        logger.warning("Cannot save tokenizer, please copy the files manually.")
--- a/src/llmtuner/webui/chatter.py
+++ b/src/llmtuner/webui/chatter.py
@@ -1,3 +1,4 @@
+import gradio as gr
 from gradio.components import Component # cannot use TYPE_CHECKING here
 from typing import TYPE_CHECKING, Any, Dict, Generator, List, Optional, Tuple

@@ -26,19 +27,19 @@ class WebChatModel(ChatModel):
        return self.model is not None

    def load_model(self, data: Dict[Component, Any]) -> Generator[str, None, None]:
-        get = lambda name: data[self.manager.get_elem(name)]
+        get = lambda name: data[self.manager.get_elem_by_name(name)]
        lang = get("top.lang")
-
+        error = ""
        if self.loaded:
-            yield ALERTS["err_exists"][lang]
-            return
+            error = ALERTS["err_exists"][lang]
+        elif not get("top.model_name"):
+            error = ALERTS["err_no_model"][lang]
+        elif not get("top.model_path"):
+            error = ALERTS["err_no_path"][lang]

-        if not get("top.model_name"):
-            yield ALERTS["err_no_model"][lang]
-            return
-
-        if not get("top.model_path"):
-            yield ALERTS["err_no_path"][lang]
+        if error:
+            gr.Warning(error)
+            yield error
            return

        if get("top.checkpoints"):
@@ -65,9 +66,7 @@ class WebChatModel(ChatModel):
        yield ALERTS["info_loaded"][lang]

    def unload_model(self, data: Dict[Component, Any]) -> Generator[str, None, None]:
-        get = lambda name: data[self.manager.get_elem(name)]
-        lang = get("top.lang")
-
+        lang = data[self.manager.get_elem_by_name("top.lang")]
        yield ALERTS["info_unloading"][lang]
        self.model = None
        self.tokenizer = None
--- a/src/llmtuner/webui/common.py
+++ b/src/llmtuner/webui/common.py
@@ -61,26 +61,31 @@ def get_model_path(model_name: str) -> str:
    return user_config["path_dict"].get(model_name, None) or SUPPORTED_MODELS.get(model_name, "")


+def get_prefix(model_name: str) -> str:
+    return model_name.split("-")[0]
+
+
 def get_module(model_name: str) -> str:
-    return DEFAULT_MODULE.get(model_name.split("-")[0], "q_proj,v_proj")
+    return DEFAULT_MODULE.get(get_prefix(model_name), "q_proj,v_proj")


 def get_template(model_name: str) -> str:
-    if model_name.endswith("Chat") and model_name.split("-")[0] in DEFAULT_TEMPLATE:
-        return DEFAULT_TEMPLATE[model_name.split("-")[0]]
+    if model_name.endswith("Chat") and get_prefix(model_name) in DEFAULT_TEMPLATE:
+        return DEFAULT_TEMPLATE[get_prefix(model_name)]
    return "default"


 def list_checkpoint(model_name: str, finetuning_type: str) -> Dict[str, Any]:
    checkpoints = []
-    save_dir = get_save_dir(model_name, finetuning_type)
-    if save_dir and os.path.isdir(save_dir):
-        for checkpoint in os.listdir(save_dir):
-            if (
-                os.path.isdir(os.path.join(save_dir, checkpoint))
-                and any([os.path.isfile(os.path.join(save_dir, checkpoint, name)) for name in CKPT_NAMES])
-            ):
-                checkpoints.append(checkpoint)
+    if model_name:
+        save_dir = get_save_dir(model_name, finetuning_type)
+        if save_dir and os.path.isdir(save_dir):
+            for checkpoint in os.listdir(save_dir):
+                if (
+                    os.path.isdir(os.path.join(save_dir, checkpoint))
+                    and any([os.path.isfile(os.path.join(save_dir, checkpoint, name)) for name in CKPT_NAMES])
+                ):
+                    checkpoints.append(checkpoint)
    return gr.update(value=[], choices=checkpoints)


--- a/src/llmtuner/webui/components/chatbot.py
+++ b/src/llmtuner/webui/components/chatbot.py
@@ -11,11 +11,9 @@ def create_chat_box(
    engine: "Engine",
    visible: Optional[bool] = False
 ) -> Tuple["Block", "Component", "Component", Dict[str, "Component"]]:
-    elem_dict = dict()
-
    with gr.Box(visible=visible) as chat_box:
        chatbot = gr.Chatbot()
-
+        history = gr.State([])
        with gr.Row():
            with gr.Column(scale=4):
                system = gr.Textbox(show_label=False)
@@ -29,13 +27,6 @@ def create_chat_box(
                top_p = gr.Slider(0.01, 1, value=gen_kwargs.top_p, step=0.01)
                temperature = gr.Slider(0.01, 1.5, value=gen_kwargs.temperature, step=0.01)

-    elem_dict.update(dict(
-        system=system, query=query, submit_btn=submit_btn, clear_btn=clear_btn,
-        max_new_tokens=max_new_tokens, top_p=top_p, temperature=temperature
-    ))
-
-    history = gr.State([])
-
    submit_btn.click(
        engine.chatter.predict,
        [chatbot, query, history, system, max_new_tokens, top_p, temperature],
@@ -47,4 +38,12 @@ def create_chat_box(

    clear_btn.click(lambda: ([], []), outputs=[chatbot, history], show_progress=True)

-    return chat_box, chatbot, history, elem_dict
+    return chat_box, chatbot, history, dict(
+        system=system,
+        query=query,
+        submit_btn=submit_btn,
+        clear_btn=clear_btn,
+        max_new_tokens=max_new_tokens,
+        top_p=top_p,
+        temperature=temperature
+    )
--- a/src/llmtuner/webui/components/data.py
+++ b/src/llmtuner/webui/components/data.py
@@ -1,21 +1,103 @@
+import os
+import json
 import gradio as gr
-from typing import TYPE_CHECKING, Tuple
+from typing import TYPE_CHECKING, Any, Dict, Tuple
+
+from llmtuner.webui.common import DATA_CONFIG

 if TYPE_CHECKING:
-    from gradio.blocks import Block
    from gradio.components import Component


-def create_preview_box() -> Tuple["Block", "Component", "Component", "Component"]:
-    with gr.Box(visible=False, elem_classes="modal-box") as preview_box:
+PAGE_SIZE = 2
+
+
+def prev_page(page_index: int) -> int:
+    return page_index - 1 if page_index > 0 else page_index
+
+
+def next_page(page_index: int, total_num: int) -> int:
+    return page_index + 1 if (page_index + 1) * PAGE_SIZE < total_num else page_index
+
+
+def can_preview(dataset_dir: str, dataset: list) -> Dict[str, Any]:
+    with open(os.path.join(dataset_dir, DATA_CONFIG), "r", encoding="utf-8") as f:
+        dataset_info = json.load(f)
+
+    if (
+        len(dataset) > 0
+        and "file_name" in dataset_info[dataset[0]]
+        and os.path.isfile(os.path.join(dataset_dir, dataset_info[dataset[0]]["file_name"]))
+    ):
+        return gr.update(interactive=True)
+    else:
+        return gr.update(interactive=False)
+
+
+def get_preview(dataset_dir: str, dataset: list, page_index: int) -> Tuple[int, list, Dict[str, Any]]:
+    with open(os.path.join(dataset_dir, DATA_CONFIG), "r", encoding="utf-8") as f:
+        dataset_info = json.load(f)
+
+    data_file: str = dataset_info[dataset[0]]["file_name"]
+    with open(os.path.join(dataset_dir, data_file), "r", encoding="utf-8") as f:
+        if data_file.endswith(".json"):
+            data = json.load(f)
+        elif data_file.endswith(".jsonl"):
+            data = [json.loads(line) for line in f]
+        else:
+            data = [line for line in f]
+    return len(data), data[PAGE_SIZE * page_index : PAGE_SIZE * (page_index + 1)], gr.update(visible=True)
+
+
+def create_preview_box(dataset_dir: "gr.Textbox", dataset: "gr.Dropdown") -> Dict[str, "Component"]:
+    data_preview_btn = gr.Button(interactive=False, scale=1)
+    with gr.Column(visible=False, elem_classes="modal-box") as preview_box:
        with gr.Row():
-            preview_count = gr.Number(interactive=False)
+            preview_count = gr.Number(value=0, interactive=False, precision=0)
+            page_index = gr.Number(value=0, interactive=False, precision=0)
+
+        with gr.Row():
+            prev_btn = gr.Button()
+            next_btn = gr.Button()
+            close_btn = gr.Button()

        with gr.Row():
            preview_samples = gr.JSON(interactive=False)

-        close_btn = gr.Button()
-
+    dataset.change(
+        can_preview, [dataset_dir, dataset], [data_preview_btn], queue=False
+    ).then(
+        lambda: 0, outputs=[page_index], queue=False
+    )
+    data_preview_btn.click(
+        get_preview,
+        [dataset_dir, dataset, page_index],
+        [preview_count, preview_samples, preview_box],
+        queue=False
+    )
+    prev_btn.click(
+        prev_page, [page_index], [page_index], queue=False
+    ).then(
+        get_preview,
+        [dataset_dir, dataset, page_index],
+        [preview_count, preview_samples, preview_box],
+        queue=False
+    )
+    next_btn.click(
+        next_page, [page_index, preview_count], [page_index], queue=False
+    ).then(
+        get_preview,
+        [dataset_dir, dataset, page_index],
+        [preview_count, preview_samples, preview_box],
+        queue=False
+    )
    close_btn.click(lambda: gr.update(visible=False), outputs=[preview_box], queue=False)
-
-    return preview_box, preview_count, preview_samples, close_btn
+    return dict(
+        data_preview_btn=data_preview_btn,
+        preview_count=preview_count,
+        page_index=page_index,
+        prev_btn=prev_btn,
+        next_btn=next_btn,
+        close_btn=close_btn,
+        preview_samples=preview_samples
+    )
--- a/src/llmtuner/webui/components/eval.py
+++ b/src/llmtuner/webui/components/eval.py
@@ -3,7 +3,6 @@ from typing import TYPE_CHECKING, Dict

 from llmtuner.webui.common import list_dataset, DEFAULT_DATA_DIR
 from llmtuner.webui.components.data import create_preview_box
-from llmtuner.webui.utils import can_preview, get_preview

 if TYPE_CHECKING:
    from gradio.components import Component
@@ -17,28 +16,12 @@ def create_eval_tab(engine: "Engine") -> Dict[str, "Component"]:
    with gr.Row():
        dataset_dir = gr.Textbox(value=DEFAULT_DATA_DIR, scale=2)
        dataset = gr.Dropdown(multiselect=True, scale=4)
-        data_preview_btn = gr.Button(interactive=False, scale=1)
+        preview_elems = create_preview_box(dataset_dir, dataset)

    dataset_dir.change(list_dataset, [dataset_dir], [dataset], queue=False)
-    dataset.change(can_preview, [dataset_dir, dataset], [data_preview_btn], queue=False)

    input_elems.update({dataset_dir, dataset})
-    elem_dict.update(dict(
-        dataset_dir=dataset_dir, dataset=dataset, data_preview_btn=data_preview_btn
-    ))
-
-    preview_box, preview_count, preview_samples, close_btn = create_preview_box()
-
-    data_preview_btn.click(
-        get_preview,
-        [dataset_dir, dataset],
-        [preview_count, preview_samples, preview_box],
-        queue=False
-    )
-
-    elem_dict.update(dict(
-        preview_count=preview_count, preview_samples=preview_samples, close_btn=close_btn
-    ))
+    elem_dict.update(dict(dataset_dir=dataset_dir, dataset=dataset, **preview_elems))

    with gr.Row():
        cutoff_len = gr.Slider(value=1024, minimum=4, maximum=8192, step=1)
--- a/src/llmtuner/webui/components/export.py
+++ b/src/llmtuner/webui/components/export.py
@@ -1,16 +1,54 @@
 import gradio as gr
-from typing import TYPE_CHECKING, Dict
+from typing import TYPE_CHECKING, Dict, Generator, List

-from llmtuner.webui.utils import save_model
+from llmtuner.tuner import export_model
+from llmtuner.webui.common import get_save_dir
+from llmtuner.webui.locales import ALERTS

 if TYPE_CHECKING:
    from gradio.components import Component
    from llmtuner.webui.engine import Engine


-def create_export_tab(engine: "Engine") -> Dict[str, "Component"]:
-    elem_dict = dict()
+def save_model(
+    lang: str,
+    model_name: str,
+    model_path: str,
+    checkpoints: List[str],
+    finetuning_type: str,
+    template: str,
+    max_shard_size: int,
+    export_dir: str
+) -> Generator[str, None, None]:
+    error = ""
+    if not model_name:
+        error = ALERTS["err_no_model"][lang]
+    elif not model_path:
+        error = ALERTS["err_no_path"][lang]
+    elif not checkpoints:
+        error = ALERTS["err_no_checkpoint"][lang]
+    elif not export_dir:
+        error = ALERTS["err_no_export_dir"][lang]

+    if error:
+        gr.Warning(error)
+        yield error
+        return
+
+    args = dict(
+        model_name_or_path=model_path,
+        checkpoint_dir=",".join([get_save_dir(model_name, finetuning_type, ckpt) for ckpt in checkpoints]),
+        finetuning_type=finetuning_type,
+        template=template,
+        export_dir=export_dir
+    )
+
+    yield ALERTS["info_exporting"][lang]
+    export_model(args, max_shard_size="{}GB".format(max_shard_size))
+    yield ALERTS["info_exported"][lang]
+
+
+def create_export_tab(engine: "Engine") -> Dict[str, "Component"]:
    with gr.Row():
        export_dir = gr.Textbox()
        max_shard_size = gr.Slider(value=10, minimum=1, maximum=100)
@@ -21,23 +59,21 @@ def create_export_tab(engine: "Engine") -> Dict[str, "Component"]:
    export_btn.click(
        save_model,
        [
-            engine.manager.get_elem("top.lang"),
-            engine.manager.get_elem("top.model_name"),
-            engine.manager.get_elem("top.model_path"),
-            engine.manager.get_elem("top.checkpoints"),
-            engine.manager.get_elem("top.finetuning_type"),
-            engine.manager.get_elem("top.template"),
+            engine.manager.get_elem_by_name("top.lang"),
+            engine.manager.get_elem_by_name("top.model_name"),
+            engine.manager.get_elem_by_name("top.model_path"),
+            engine.manager.get_elem_by_name("top.checkpoints"),
+            engine.manager.get_elem_by_name("top.finetuning_type"),
+            engine.manager.get_elem_by_name("top.template"),
            max_shard_size,
            export_dir
        ],
        [info_box]
    )

-    elem_dict.update(dict(
+    return dict(
        export_dir=export_dir,
        max_shard_size=max_shard_size,
        export_btn=export_btn,
        info_box=info_box
-    ))
-
-    return elem_dict
+    )
--- a/src/llmtuner/webui/components/top.py
+++ b/src/llmtuner/webui/components/top.py
@@ -31,9 +31,10 @@ def create_top() -> Dict[str, "Component"]:

    with gr.Accordion(label="Model config (LLaMA only)", open=False) as llama_tab:
        with gr.Row():
-            flash_attn = gr.Checkbox(value=False)
-            shift_attn = gr.Checkbox(value=False)
-            rope_scaling = gr.Dropdown(choices=["none", "linear", "dynamic"], value="none")
+            with gr.Column():
+                flash_attn = gr.Checkbox(value=False)
+                shift_attn = gr.Checkbox(value=False)
+            rope_scaling = gr.Radio(choices=["none", "linear", "dynamic"], value="none")

    model_name.change(
        list_checkpoint, [model_name, finetuning_type], [checkpoints], queue=False
--- a/src/llmtuner/webui/components/train.py
+++ b/src/llmtuner/webui/components/train.py
@@ -5,7 +5,7 @@ from transformers.trainer_utils import SchedulerType
 from llmtuner.extras.constants import TRAINING_STAGES
 from llmtuner.webui.common import list_checkpoint, list_dataset, DEFAULT_DATA_DIR
 from llmtuner.webui.components.data import create_preview_box
-from llmtuner.webui.utils import can_preview, get_preview, gen_plot
+from llmtuner.webui.utils import gen_plot

 if TYPE_CHECKING:
    from gradio.components import Component
@@ -22,28 +22,14 @@ def create_train_tab(engine: "Engine") -> Dict[str, "Component"]:
        )
        dataset_dir = gr.Textbox(value=DEFAULT_DATA_DIR, scale=2)
        dataset = gr.Dropdown(multiselect=True, scale=4)
-        data_preview_btn = gr.Button(interactive=False, scale=1)
+        preview_elems = create_preview_box(dataset_dir, dataset)

    training_stage.change(list_dataset, [dataset_dir, training_stage], [dataset], queue=False)
    dataset_dir.change(list_dataset, [dataset_dir, training_stage], [dataset], queue=False)
-    dataset.change(can_preview, [dataset_dir, dataset], [data_preview_btn], queue=False)

    input_elems.update({training_stage, dataset_dir, dataset})
    elem_dict.update(dict(
-        training_stage=training_stage, dataset_dir=dataset_dir, dataset=dataset, data_preview_btn=data_preview_btn
-    ))
-
-    preview_box, preview_count, preview_samples, close_btn = create_preview_box()
-
-    data_preview_btn.click(
-        get_preview,
-        [dataset_dir, dataset],
-        [preview_count, preview_samples, preview_box],
-        queue=False
-    )
-
-    elem_dict.update(dict(
-        preview_count=preview_count, preview_samples=preview_samples, close_btn=close_btn
+        training_stage=training_stage, dataset_dir=dataset_dir, dataset=dataset, **preview_elems
    ))

    with gr.Row():
@@ -79,26 +65,30 @@ def create_train_tab(engine: "Engine") -> Dict[str, "Component"]:
            logging_steps = gr.Slider(value=5, minimum=5, maximum=1000, step=5)
            save_steps = gr.Slider(value=100, minimum=10, maximum=5000, step=10)
            warmup_steps = gr.Slider(value=0, minimum=0, maximum=5000, step=1)
+            neft_alpha = gr.Slider(value=0, minimum=0, maximum=10, step=0.1)

-    input_elems.update({logging_steps, save_steps, warmup_steps})
+            with gr.Column():
+                train_on_prompt = gr.Checkbox(value=False)
+                upcast_layernorm = gr.Checkbox(value=False)
+
+    input_elems.update({logging_steps, save_steps, warmup_steps, neft_alpha, train_on_prompt, upcast_layernorm})
    elem_dict.update(dict(
-        advanced_tab=advanced_tab, logging_steps=logging_steps, save_steps=save_steps, warmup_steps=warmup_steps
+        advanced_tab=advanced_tab, logging_steps=logging_steps, save_steps=save_steps, warmup_steps=warmup_steps,
+        neft_alpha=neft_alpha, train_on_prompt=train_on_prompt, upcast_layernorm=upcast_layernorm
    ))

    with gr.Accordion(label="LoRA config", open=False) as lora_tab:
        with gr.Row():
            lora_rank = gr.Slider(value=8, minimum=1, maximum=1024, step=1, scale=1)
            lora_dropout = gr.Slider(value=0.1, minimum=0, maximum=1, step=0.01, scale=1)
-            lora_target = gr.Textbox(scale=2)
+            lora_target = gr.Textbox(scale=1)
+            additional_target = gr.Textbox(scale=1)
            resume_lora_training = gr.Checkbox(value=True, scale=1)

-    input_elems.update({lora_rank, lora_dropout, lora_target, resume_lora_training})
+    input_elems.update({lora_rank, lora_dropout, lora_target, additional_target, resume_lora_training})
    elem_dict.update(dict(
-        lora_tab=lora_tab,
-        lora_rank=lora_rank,
-        lora_dropout=lora_dropout,
-        lora_target=lora_target,
-        resume_lora_training=resume_lora_training,
+        lora_tab=lora_tab, lora_rank=lora_rank, lora_dropout=lora_dropout, lora_target=lora_target,
+        additional_target=additional_target, resume_lora_training=resume_lora_training,
    ))

    with gr.Accordion(label="RLHF config", open=False) as rlhf_tab:
@@ -109,7 +99,7 @@ def create_train_tab(engine: "Engine") -> Dict[str, "Component"]:

    refresh_btn.click(
        list_checkpoint,
-        [engine.manager.get_elem("top.model_name"), engine.manager.get_elem("top.finetuning_type")],
+        [engine.manager.get_elem_by_name("top.model_name"), engine.manager.get_elem_by_name("top.finetuning_type")],
        [reward_model],
        queue=False
    )
@@ -139,19 +129,24 @@ def create_train_tab(engine: "Engine") -> Dict[str, "Component"]:

    input_elems.add(output_dir)
    output_elems = [output_box, process_bar]
-    elem_dict.update(dict(
-        cmd_preview_btn=cmd_preview_btn, start_btn=start_btn, stop_btn=stop_btn, output_dir=output_dir,
-        resume_btn=resume_btn, process_bar=process_bar, output_box=output_box, loss_viewer=loss_viewer
-    ))

    cmd_preview_btn.click(engine.runner.preview_train, input_elems, output_elems)
    start_btn.click(engine.runner.run_train, input_elems, output_elems)
    stop_btn.click(engine.runner.set_abort, queue=False)
    resume_btn.change(engine.runner.monitor, outputs=output_elems)

+    elem_dict.update(dict(
+        cmd_preview_btn=cmd_preview_btn, start_btn=start_btn, stop_btn=stop_btn, output_dir=output_dir,
+        resume_btn=resume_btn, process_bar=process_bar, output_box=output_box, loss_viewer=loss_viewer
+    ))
+
    output_box.change(
        gen_plot,
-        [engine.manager.get_elem("top.model_name"), engine.manager.get_elem("top.finetuning_type"), output_dir],
+        [
+            engine.manager.get_elem_by_name("top.model_name"),
+            engine.manager.get_elem_by_name("top.finetuning_type"),
+            output_dir
+        ],
        loss_viewer,
        queue=False
    )
--- a/src/llmtuner/webui/css.py
+++ b/src/llmtuner/webui/css.py
@@ -6,10 +6,12 @@ CSS = r"""
  transform: translate(-50%, -50%); /* center horizontally */
  max-width: 1000px;
  max-height: 750px;
-  overflow-y: scroll !important;
+  overflow-y: auto;
  background-color: var(--input-background-fill);
+  flex-wrap: nowrap !important;
  border: 2px solid black !important;
  z-index: 1000;
+  padding: 10px;
 }

 .dark .modal-box {
--- a/src/llmtuner/webui/engine.py
+++ b/src/llmtuner/webui/engine.py
@@ -19,36 +19,36 @@ class Engine:
        self.chatter: "WebChatModel" = WebChatModel(manager=self.manager, lazy_init=(not pure_chat))

    def _form_dict(self, resume_dict: Dict[str, Dict[str, Any]]):
-        return {self.manager.get_elem(k): gr.update(**v) for k, v in resume_dict.items()}
+        return {self.manager.get_elem_by_name(k): gr.update(**v) for k, v in resume_dict.items()}

    def resume(self) -> Generator[Dict[Component, Dict[str, Any]], None, None]:
        user_config = load_config()
        lang = user_config.get("lang", None) or "en"

-        resume_dict = {
+        init_dict = {
            "top.lang": {"value": lang},
            "infer.chat_box": {"visible": self.chatter.loaded}
        }

        if not self.pure_chat:
-            resume_dict["train.dataset"] = {"choices": list_dataset()["choices"]}
-            resume_dict["eval.dataset"] = {"choices": list_dataset()["choices"]}
+            init_dict["train.dataset"] = {"choices": list_dataset()["choices"]}
+            init_dict["eval.dataset"] = {"choices": list_dataset()["choices"]}

            if user_config.get("last_model", None):
-                resume_dict["top.model_name"] = {"value": user_config["last_model"]}
-                resume_dict["top.model_path"] = {"value": get_model_path(user_config["last_model"])}
+                init_dict["top.model_name"] = {"value": user_config["last_model"]}
+                init_dict["top.model_path"] = {"value": get_model_path(user_config["last_model"])}

-        yield self._form_dict(resume_dict)
+        yield self._form_dict(init_dict)

-        if self.runner.alive:
-            yield {elem: gr.update(value=value) for elem, value in self.runner.data.items()}
-            if self.runner.do_train:
-                resume_dict = {"train.resume_btn": {"value": True}}
+        if not self.pure_chat:
+            if self.runner.alive:
+                yield {elem: gr.update(value=value) for elem, value in self.runner.running_data.items()}
+                if self.runner.do_train:
+                    yield self._form_dict({"train.resume_btn": {"value": True}})
+                else:
+                    yield self._form_dict({"eval.resume_btn": {"value": True}})
            else:
-                resume_dict = {"eval.resume_btn": {"value": True}}
-        else:
-            resume_dict = {"train.output_dir": {"value": get_time()}}
-        yield self._form_dict(resume_dict)
+                yield self._form_dict({"train.output_dir": {"value": get_time()}})

    def change_lang(self, lang: str) -> Dict[Component, Dict[str, Any]]:
        return {
--- a/src/llmtuner/webui/interface.py
+++ b/src/llmtuner/webui/interface.py
@@ -14,15 +14,15 @@ from llmtuner.webui.css import CSS
 from llmtuner.webui.engine import Engine


-require_version("gradio==3.38.0", "To fix: pip install gradio==3.38.0")
+require_version("gradio>=3.38.0,<4.0.0", "To fix: pip install \"gradio>=3.38.0,<4.0.0\"")


 def create_ui() -> gr.Blocks:
    engine = Engine(pure_chat=False)

-    with gr.Blocks(title="Web Tuner", css=CSS) as demo:
+    with gr.Blocks(title="LLaMA Board", css=CSS) as demo:
        engine.manager.all_elems["top"] = create_top()
-        lang: "gr.Dropdown" = engine.manager.get_elem("top.lang")
+        lang: "gr.Dropdown" = engine.manager.get_elem_by_name("top.lang")

        with gr.Tab("Train"):
            engine.manager.all_elems["train"] = create_train_tab(engine)
--- a/src/llmtuner/webui/locales.py
+++ b/src/llmtuner/webui/locales.py
@@ -163,12 +163,28 @@ LOCALES = {
            "label": "数量"
        }
    },
-    "preview_samples": {
+    "page_index": {
        "en": {
-            "label": "Samples"
+            "label": "Page"
        },
        "zh": {
-            "label": "样例"
+            "label": "页数"
+        }
+    },
+    "prev_btn": {
+        "en": {
+            "value": "Prev"
+        },
+        "zh": {
+            "value": "上一页"
+        }
+    },
+    "next_btn": {
+        "en": {
+            "value": "Next"
+        },
+        "zh": {
+            "value": "下一页"
        }
    },
    "close_btn": {
@@ -179,6 +195,14 @@ LOCALES = {
            "value": "关闭"
        }
    },
+    "preview_samples": {
+        "en": {
+            "label": "Samples"
+        },
+        "zh": {
+            "label": "样例"
+        }
+    },
    "cutoff_len": {
        "en": {
            "label": "Cutoff length",
@@ -309,6 +333,36 @@ LOCALES = {
            "info": "学习率预热采用的步数。"
        }
    },
+    "neft_alpha": {
+        "en": {
+            "label": "NEFTune Alpha",
+            "info": "Magnitude of noise adding to embedding vectors."
+        },
+        "zh": {
+            "label": "NEFTune 噪声参数",
+            "info": "嵌入向量所添加的噪声大小。"
+        }
+    },
+    "train_on_prompt": {
+        "en": {
+            "label": "Train on prompt",
+            "info": "Compute loss on the prompt tokens in supervised fine-tuning."
+        },
+        "zh": {
+            "label": "计算输入损失",
+            "info": "在监督微调时候计算输入序列的损失。"
+        }
+    },
+    "upcast_layernorm": {
+        "en": {
+            "label": "Upcast LayerNorm",
+            "info": "Upcast weights of layernorm in float32."
+        },
+        "zh": {
+            "label": "缩放归一化层",
+            "info": "将归一化层权重缩放至 32 位浮点数。"
+        }
+    },
    "lora_tab": {
        "en": {
            "label": "LoRA configurations"
@@ -340,11 +394,21 @@ LOCALES = {
    "lora_target": {
        "en": {
            "label": "LoRA modules (optional)",
-            "info": "The name(s) of target modules to apply LoRA. Use commas to separate multiple modules."
+            "info": "Name(s) of target modules to apply LoRA. Use commas to separate multiple modules."
        },
        "zh": {
-            "label": "LoRA 作用层（非必填）",
-            "info": "应用 LoRA 的线性层名称。使用英文逗号分隔多个名称。"
+            "label": "LoRA 作用模块（非必填）",
+            "info": "应用 LoRA 的目标模块名称。使用英文逗号分隔多个名称。"
+        }
+    },
+    "additional_target": {
+        "en": {
+            "label": "Additional modules (optional)",
+            "info": "Name(s) of modules apart from LoRA layers to be set as trainable. Use commas to separate multiple modules."
+        },
+        "zh": {
+            "label": "附加模块（非必填）",
+            "info": "除 LoRA 层以外的可训练模块名称。使用英文逗号分隔多个名称。"
        }
    },
    "resume_lora_training": {
--- a/src/llmtuner/webui/manager.py
+++ b/src/llmtuner/webui/manager.py
@@ -1,4 +1,4 @@
-from typing import TYPE_CHECKING, Dict, List
+from typing import TYPE_CHECKING, Dict, List, Set

 if TYPE_CHECKING:
    from gradio.components import Component
@@ -9,14 +9,14 @@ class Manager:
    def __init__(self) -> None:
        self.all_elems: Dict[str, Dict[str, "Component"]] = {}

-    def get_elem(self, name: str) -> "Component":
+    def get_elem_by_name(self, name: str) -> "Component":
        r"""
        Example: top.lang, train.dataset
        """
        tab_name, elem_name = name.split(".")
        return self.all_elems[tab_name][elem_name]

-    def get_base_elems(self):
+    def get_base_elems(self) -> Set["Component"]:
        return {
            self.all_elems["top"]["lang"],
            self.all_elems["top"]["model_name"],
--- a/src/llmtuner/webui/runner.py
+++ b/src/llmtuner/webui/runner.py
@@ -26,12 +26,15 @@ class Runner:

    def __init__(self, manager: "Manager") -> None:
        self.manager = manager
+        """ Resume """
        self.thread: "Thread" = None
-        self.data: Dict["Component", Any] = None
        self.do_train = True
+        self.running_data: Dict["Component", Any] = None
        self.monitor_inputs: Dict[str, str] = None
+        """ State """
        self.aborted = False
        self.running = False
+        """ Handler """
        self.logger_handler = LoggerHandler()
        self.logger_handler.setLevel(logging.INFO)
        logging.root.addHandler(self.logger_handler)
@@ -45,7 +48,11 @@ class Runner:
        self.aborted = True
        self.running = False

-    def _initialize(self, lang: str, model_name: str, model_path: str, dataset: List[str]) -> str:
+    def _initialize(self, data: Dict[Component, Any], do_train: bool) -> str:
+        get = lambda name: data[self.manager.get_elem_by_name(name)]
+        lang, model_name, model_path = get("top.lang"), get("top.model_name"), get("top.model_path")
+        dataset = get("train.dataset") if do_train else get("eval.dataset")
+
        if self.running:
            return ALERTS["err_conflict"][lang]

@@ -72,8 +79,8 @@ class Runner:
        else:
            return finish_info

-    def _parse_train_args(self, data: Dict[Component, Any]) -> Tuple[str, str, str, List[str], str, Dict[str, Any]]:
-        get = lambda name: data[self.manager.get_elem(name)]
+    def _parse_train_args(self, data: Dict[Component, Any]) -> Dict[str, Any]:
+        get = lambda name: data[self.manager.get_elem_by_name(name)]
        user_config = load_config()

        if get("top.checkpoints"):
@@ -83,13 +90,10 @@ class Runner:
        else:
            checkpoint_dir = None

-        output_dir = get_save_dir(get("top.model_name"), get("top.finetuning_type"), get("train.output_dir"))
-
        args = dict(
            stage=TRAINING_STAGES[get("train.training_stage")],
            model_name_or_path=get("top.model_path"),
            do_train=True,
-            overwrite_cache=False,
            cache_dir=user_config.get("cache_dir", None),
            checkpoint_dir=checkpoint_dir,
            finetuning_type=get("top.finetuning_type"),
@@ -112,11 +116,15 @@ class Runner:
            logging_steps=get("train.logging_steps"),
            save_steps=get("train.save_steps"),
            warmup_steps=get("train.warmup_steps"),
+            neft_alpha=get("train.neft_alpha"),
+            train_on_prompt=get("train.train_on_prompt"),
+            upcast_layernorm=get("train.upcast_layernorm"),
            lora_rank=get("train.lora_rank"),
            lora_dropout=get("train.lora_dropout"),
            lora_target=get("train.lora_target") or get_module(get("top.model_name")),
+            additional_target=get("train.additional_target") if get("train.additional_target") else None,
            resume_lora_training=get("train.resume_lora_training"),
-            output_dir=output_dir
+            output_dir=get_save_dir(get("top.model_name"), get("top.finetuning_type"), get("train.output_dir"))
        )
        args[get("train.compute_type")] = True
        args["disable_tqdm"] = True
@@ -128,7 +136,7 @@ class Runner:
            args["upcast_layernorm"] = True

        if args["stage"] == "ppo":
-            args["reward_model"] = get("train.reward_model")
+            args["reward_model"] = get_save_dir(get("top.model_name"), get("top.finetuning_type"), get("train.reward_model"))

        if args["stage"] == "dpo":
            args["dpo_beta"] = get("train.dpo_beta")
@@ -139,10 +147,10 @@ class Runner:
            args["eval_steps"] = get("train.save_steps")
            args["load_best_model_at_end"] = True

-        return get("top.lang"), get("top.model_name"), get("top.model_path"), get("train.dataset"), output_dir, args
+        return args

-    def _parse_eval_args(self, data: Dict[Component, Any]) -> Tuple[str, str, str, List[str], str, Dict[str, Any]]:
-        get = lambda name: data[self.manager.get_elem(name)]
+    def _parse_eval_args(self, data: Dict[Component, Any]) -> Dict[str, Any]:
+        get = lambda name: data[self.manager.get_elem_by_name(name)]
        user_config = load_config()

        if get("top.checkpoints"):
@@ -160,7 +168,6 @@ class Runner:
            stage="sft",
            model_name_or_path=get("top.model_path"),
            do_eval=True,
-            overwrite_cache=False,
            predict_with_generate=True,
            cache_dir=user_config.get("cache_dir", None),
            checkpoint_dir=checkpoint_dir,
@@ -179,34 +186,35 @@ class Runner:
            max_new_tokens=get("eval.max_new_tokens"),
            top_p=get("eval.top_p"),
            temperature=get("eval.temperature"),
-            output_dir=get("eval.output_dir")
+            output_dir=output_dir
        )

        if get("eval.predict"):
            args.pop("do_eval", None)
            args["do_predict"] = True

-        return get("top.lang"), get("top.model_name"), get("top.model_path"), get("eval.dataset"), output_dir, args
+        return args

    def _preview(self, data: Dict[Component, Any], do_train: bool) -> Generator[Tuple[str, Dict[str, Any]], None, None]:
-        parse_func = self._parse_train_args if do_train else self._parse_eval_args
-        lang, model_name, model_path, dataset, _, args = parse_func(data)
-        error = self._initialize(lang, model_name, model_path, dataset)
+        error = self._initialize(data, do_train)
        if error:
+            gr.Warning(error)
            yield error, gr.update(visible=False)
        else:
+            args = self._parse_train_args(data) if do_train else self._parse_eval_args(data)
            yield gen_cmd(args), gr.update(visible=False)

    def _launch(self, data: Dict[Component, Any], do_train: bool) -> Generator[Tuple[str, Dict[str, Any]], None, None]:
-        parse_func = self._parse_train_args if do_train else self._parse_eval_args
-        lang, model_name, model_path, dataset, output_dir, args = parse_func(data)
-        self.data, self.do_train, self.monitor_inputs = data, do_train, dict(lang=lang, output_dir=output_dir)
-        error = self._initialize(lang, model_name, model_path, dataset)
+        error = self._initialize(data, do_train)
        if error:
+            gr.Warning(error)
            yield error, gr.update(visible=False)
        else:
-            self.running = True
+            args = self._parse_train_args(data) if do_train else self._parse_eval_args(data)
            run_kwargs = dict(args=args, callbacks=[self.trainer_callback])
+            self.running = True
+            self.do_train, self.running_data = do_train, data
+            self.monitor_inputs = dict(lang=data[self.manager.get_elem_by_name("top.lang")], output_dir=args["output_dir"])
            self.thread = Thread(target=run_exp, kwargs=run_kwargs)
            self.thread.start()
            yield from self.monitor()
--- a/src/llmtuner/webui/utils.py
+++ b/src/llmtuner/webui/utils.py
@@ -3,13 +3,11 @@ import json
 import gradio as gr
 import matplotlib.figure
 import matplotlib.pyplot as plt
-from typing import TYPE_CHECKING, Any, Dict, Generator, List, Optional, Tuple
+from typing import TYPE_CHECKING, Any, Dict
 from datetime import datetime

 from llmtuner.extras.ploting import smooth
-from llmtuner.tuner import export_model
-from llmtuner.webui.common import get_save_dir, DATA_CONFIG
-from llmtuner.webui.locales import ALERTS
+from llmtuner.webui.common import get_save_dir

 if TYPE_CHECKING:
    from llmtuner.extras.callbacks import LogCallback
@@ -33,37 +31,6 @@ def get_time() -> str:
    return datetime.now().strftime('%Y-%m-%d-%H-%M-%S')


-def can_preview(dataset_dir: str, dataset: list) -> Dict[str, Any]:
-    with open(os.path.join(dataset_dir, DATA_CONFIG), "r", encoding="utf-8") as f:
-        dataset_info = json.load(f)
-
-    if (
-        len(dataset) > 0
-        and "file_name" in dataset_info[dataset[0]]
-        and os.path.isfile(os.path.join(dataset_dir, dataset_info[dataset[0]]["file_name"]))
-    ):
-        return gr.update(interactive=True)
-    else:
-        return gr.update(interactive=False)
-
-
-def get_preview(
-    dataset_dir: str, dataset: list, start: Optional[int] = 0, end: Optional[int] = 2
-) -> Tuple[int, list, Dict[str, Any]]:
-    with open(os.path.join(dataset_dir, DATA_CONFIG), "r", encoding="utf-8") as f:
-        dataset_info = json.load(f)
-
-    data_file: str = dataset_info[dataset[0]]["file_name"]
-    with open(os.path.join(dataset_dir, data_file), "r", encoding="utf-8") as f:
-        if data_file.endswith(".json"):
-            data = json.load(f)
-        elif data_file.endswith(".jsonl"):
-            data = [json.loads(line) for line in f]
-        else:
-            data = [line for line in f]
-    return len(data), data[start:end], gr.update(visible=True)
-
-
 def can_quantize(finetuning_type: str) -> Dict[str, Any]:
    if finetuning_type != "lora":
        return gr.update(value="None", interactive=False)
@@ -90,9 +57,11 @@ def get_eval_results(path: os.PathLike) -> str:


 def gen_plot(base_model: str, finetuning_type: str, output_dir: str) -> matplotlib.figure.Figure:
+    if not base_model:
+        return
    log_file = get_save_dir(base_model, finetuning_type, output_dir, "trainer_log.jsonl")
    if not os.path.isfile(log_file):
-        return None
+        return

    plt.close("all")
    fig = plt.figure()
@@ -114,42 +83,3 @@ def gen_plot(base_model: str, finetuning_type: str, output_dir: str) -> matplotl
    ax.set_xlabel("step")
    ax.set_ylabel("loss")
    return fig
-
-
-def save_model(
-    lang: str,
-    model_name: str,
-    model_path: str,
-    checkpoints: List[str],
-    finetuning_type: str,
-    template: str,
-    max_shard_size: int,
-    export_dir: str
-) -> Generator[str, None, None]:
-    if not model_name:
-        yield ALERTS["err_no_model"][lang]
-        return
-
-    if not model_path:
-        yield ALERTS["err_no_path"][lang]
-        return
-
-    if not checkpoints:
-        yield ALERTS["err_no_checkpoint"][lang]
-        return
-
-    if not export_dir:
-        yield ALERTS["err_no_export_dir"][lang]
-        return
-
-    args = dict(
-        model_name_or_path=model_path,
-        checkpoint_dir=",".join([get_save_dir(model_name, finetuning_type, ckpt) for ckpt in checkpoints]),
-        finetuning_type=finetuning_type,
-        template=template,
-        export_dir=export_dir
-    )
-
-    yield ALERTS["info_exporting"][lang]
-    export_model(args, max_shard_size="{}GB".format(max_shard_size))
-    yield ALERTS["info_exported"][lang]
Author	SHA1	Message	Date
hiyouga	ec334f5891	release v0.2.2, fix #1478 #1466 Former-commit-id: c9534c411716e1dceb54c5eb35fe845c93ee2973	2023-11-13 23:09:05 +08:00
hiyouga	885efe772e	fix #424 Former-commit-id: ca24d445f825e120e659f5cd080a954c2243b8f2	2023-11-13 22:42:23 +08:00
hiyouga	64fc9ba678	refactor evaluation, upgrade trl to 074 Former-commit-id: ed09ebe2c1926ffdb0520b3866f7fd03a9aed046	2023-11-13 22:20:35 +08:00
hiyouga	989eccd286	fix flashattn warning Former-commit-id: 6eb095d39bd82fdbdb729a0ea57fc7246e3a60d6	2023-11-10 18:34:54 +08:00
hiyouga	f0766a2ab0	add todo Former-commit-id: 0bd884feb11736d0ab24ca19885151cb47d9dcd3	2023-11-10 14:38:18 +08:00
hiyouga	178b85ff9a	refactor constants Former-commit-id: a4d4c3fd35276f20e3b354e9d13ea971029c8775	2023-11-10 14:16:10 +08:00
hiyouga	68dd1ef121	tiny fix Former-commit-id: 97ba2027bb1ddc01a3c824c40d5a180828810c2c	2023-11-09 17:20:49 +08:00
hoshi-hiyouga	b222cffe98	Merge pull request #1454 from yyq/main Update finetuning_args.py Former-commit-id: e67d8b93705383a8590f99e26e9fe8f663712aef	2023-11-09 17:12:18 +08:00
Yanqing	b4f1ab93d1	Update finetuning_args.py 更新 chatglm/falcon/bloom 的 lora_target 的名称 Former-commit-id: 06606739af035a80ae9ddba9d12c965ed289305d	2023-11-09 17:04:40 +08:00
hiyouga	f2e139f5cd	fix #1452 Former-commit-id: 4d16214467715df458e24d03bb7d303d62b8bdcd	2023-11-09 16:41:32 +08:00
hiyouga	a9cbca1604	update readme Former-commit-id: f7ead54042868550a3e8a6928ea3c0e2673f15b3	2023-11-09 16:00:24 +08:00
hiyouga	3a30ce6c16	release v0.2.1 Former-commit-id: 1c30f2be0140f5ab47c2bc811170d0271a0cdad6	2023-11-09 15:54:16 +08:00
hiyouga	48ec5355f9	add template, modify datasets Former-commit-id: 81e54beb4d0f792f4fd7f450643caaf10f2f0b7d	2023-11-09 15:53:23 +08:00
hoshi-hiyouga	11859bc322	Merge pull request #1436 from lvzii/main fix tokenizer config changed after pretrain Former-commit-id: f485c3983e413fd3a3a57b451800705b072869a7	2023-11-09 14:30:50 +08:00
hiyouga	28c67a5be8	support parquet format #1446 Former-commit-id: 44a3b9ac9f10d2012b8ad3d8c48123db9a0da2f1	2023-11-09 14:17:40 +08:00
hiyouga	44fe93e9b0	fix #1438 #1439 Former-commit-id: 84260d58dda22adc32c26bc943ed2a36fd01341d	2023-11-09 13:45:10 +08:00
lvzi	09a1681b63	fix tokenizer config changed after pretrain Changing tokenizer's attribute at preprocessing stage will result in saving a wrong tokenizer. for example, baichuan2 Former-commit-id: 19942b5314b84267691f0a5657d0679f2ddbe58b	2023-11-08 15:50:46 +08:00
hiyouga	f5ba2190fb	fix ppo train and dpo eval Former-commit-id: ced863031836632cb5920e22ae6991f251372118	2023-11-07 22:48:51 +08:00
hiyouga	14a38b5069	fix #1422 Former-commit-id: 25d7bbd0a5142f001bd2ff498df07b24137050a9	2023-11-07 19:42:01 +08:00
hiyouga	f23e5b602a	fix reward model loading Former-commit-id: 9709ca501180a1afce32e9043aedb359762b437d	2023-11-07 17:20:51 +08:00
hiyouga	857696ed9c	fix args Former-commit-id: 44d0fa2ac6a6423c7ddaf91eb8998c1b9248c04e	2023-11-07 16:36:06 +08:00
hiyouga	2084133058	update info Former-commit-id: 89643b8ac1e3fa8d2f29f1c88e4d4503410c0d05	2023-11-07 16:28:21 +08:00
hiyouga	f7f0c3070e	delete file Former-commit-id: 7d6355db0fd5809b99f3fa42753cf4dffd251fd1	2023-11-07 16:20:12 +08:00
hiyouga	46235aa514	fix #1418 Former-commit-id: 9bfecc72c53cf95fea4a9ff02ec40a65da6d4f54	2023-11-07 16:17:22 +08:00
hiyouga	2eb65d21ac	upgrade peft, fix #1088 #1411 Former-commit-id: aa7d104f8e050d12cb8f585bc8a52c850995500f	2023-11-07 16:13:36 +08:00
hiyouga	37a0d62a82	update requirements Former-commit-id: 82ebbbbb80b3f3f616274210970738d0f44b5a0a	2023-11-06 19:01:21 +08:00
hiyouga	21ac46e439	use seed in evaluate.py Former-commit-id: ab5cac1dfa681933f3266827f80068ce798b4c56	2023-11-06 18:17:51 +08:00
hiyouga	ba3e8ba20c	update readme (list in alphabetical order) Former-commit-id: e6a67b5477ee095bd92764581cfe6af57e799a69	2023-11-06 17:18:12 +08:00
hiyouga	2c48e798ca	update templates Former-commit-id: 85be2e242b062283f192c4c4d0715dc1e8a68589	2023-11-06 12:25:47 +08:00
hiyouga	4e40f5b62b	fix #1383 Former-commit-id: 9b8a782aa80f27c3e2a2e2621f9be17cae1a27e8	2023-11-06 11:42:23 +08:00
hiyouga	2a8892b785	fix deepseek template Former-commit-id: 1fdbcdad9a1cdb20299350efd87a8e5cb8c625a3	2023-11-05 13:08:46 +08:00
hiyouga	ee3b33ff03	support deepseek coder #1378 Former-commit-id: ae0c829917b9de10e71199c85c77a52cdcd2b7b3	2023-11-05 12:51:03 +08:00
hiyouga	b2c3001f8e	fix #1365 Former-commit-id: 0277d120e62164bb7fa1d6043b8fcc52c881fe96	2023-11-05 12:21:07 +08:00
hiyouga	6cfe1e1ac2	tiny fix Former-commit-id: 594c510a20d6c2782d7b7ffff18931e3003e6c22	2023-11-03 01:26:06 +08:00
hiyouga	52326870e4	fix #1290 Former-commit-id: ad911d258c4cea16f54d09bc192e076c21d26394	2023-11-03 00:44:53 +08:00
hiyouga	217fde0918	fix bug in data loader, support dpo eval Former-commit-id: f4f3dcff990468a2fa864b7176adcebbcf16dac9	2023-11-03 00:34:26 +08:00
hiyouga	065021d82a	update data readme Former-commit-id: 6a65ef44ed58714c611da60b5af96b85352e8735	2023-11-03 00:15:23 +08:00
hiyouga	4bb643e685	update data readme (zh) Former-commit-id: b32fb3a984c681732b82f6544d6c05a98c34cf4c	2023-11-02 23:42:49 +08:00
hiyouga	b77c745b1a	support sharegpt format, add datasets Former-commit-id: 202daf8987ccb7523be03ca535b572b5c9e65994	2023-11-02 23:10:04 +08:00
hiyouga	7d13501b94	support pagination in webui preview Former-commit-id: f2307e26b9c2ce5d60917cce5a9638466ea676c8	2023-11-02 21:21:45 +08:00
hiyouga	ac74639b32	fix webui Former-commit-id: 9192948fa221c0275ddfa579ef6b3442d45b8962	2023-11-02 18:03:14 +08:00
hiyouga	12fa56ae68	support warning in webui Former-commit-id: 9903b523fad2f0ec0e66c3d313823bd4674bfa2b	2023-11-02 17:57:04 +08:00
hiyouga	f11b863f4b	fix #1349 Former-commit-id: 556c023eab2a68560b26a7d5318a79410fb0c700	2023-11-02 17:02:44 +08:00
hiyouga	f3e4b72957	fix #1356 Former-commit-id: d2ed436108a339d405dad1be1ca15baca3d6d3e4	2023-11-02 16:51:52 +08:00
hiyouga	8d52fb46ca	fix #1325 Former-commit-id: 59f2cbbd52d4646fbd1ba83032bf522ecc49a50f	2023-11-01 23:38:49 +08:00
hiyouga	dab8f45033	fix chat Former-commit-id: 68f2b3df09c4c8638b9e225fd5b8aed3541e97a0	2023-11-01 23:07:58 +08:00
hiyouga	bff8b02543	update gradio, support multiple resp in api Former-commit-id: a34263e7c0e07a080276d164cdab9f12f1d767d2	2023-11-01 23:02:16 +08:00
hiyouga	2406200914	fix SFT trainer Former-commit-id: bf09b6a6cd75cc2738d9af6b8c30bcbba77fa9b5	2023-10-31 21:52:52 +08:00
hiyouga	db06fcfc84	fix #1316 Former-commit-id: 88a753fe80e277007bac2264aee24024e18f2314	2023-10-31 11:32:08 +08:00
hiyouga	93b9f74e9f	update projects Former-commit-id: 33d58e9171ad2693b9d54715eb61a6f4326c59f4	2023-10-29 22:53:47 +08:00
hiyouga	33ec844f76	add projects Former-commit-id: 495a68cd5962dd3b3af7e4a920d91ac25531a862	2023-10-29 22:07:13 +08:00
hiyouga	0f727b393e	update constants Former-commit-id: ebacbb1072045924a7e335cc9dda488d6f0be8b3	2023-10-29 13:30:20 +08:00
hiyouga	7da2aad6ee	fix vicuna template Former-commit-id: a98eda0803e4b73a24f12d848e14161451921e98	2023-10-27 22:15:25 +08:00
hiyouga	6f09f50d02	fix chatglm3 template Former-commit-id: 69bcbc9f6c98e4f4ad97ec0306b33ab21923d311	2023-10-27 21:12:06 +08:00
hiyouga	5919832059	update readme Former-commit-id: 6fb92c7088316c56ce8656e540fc47b0a5a1bf18	2023-10-27 19:19:03 +08:00
hiyouga	f7635c1afc	support chatglm3 Former-commit-id: ba82e13bbeed3b262d301196b1860d73f319401d	2023-10-27 19:16:28 +08:00
hiyouga	c762168ed0	support dataset cache Former-commit-id: f79ee62eb4a2a4a01cb4e2a6aa2d07158cf8eb59	2023-10-26 21:48:45 +08:00
hiyouga	67a46e553f	fix #1287 Former-commit-id: d885aca472c6448bbf9a9e8d16bead92038825e3	2023-10-26 17:49:41 +08:00
hiyouga	e406f37b54	fix #1285 Former-commit-id: 2f8fe4439506e844b147fe38b5eb878c5748c31c	2023-10-26 16:34:52 +08:00
hiyouga	62fe877124	remove filter in preprocess Former-commit-id: 9eac08b35fec47129a29c401ca265343f8388ab0	2023-10-23 23:46:02 +08:00
hiyouga	a0e682ba79	update neftune logic Former-commit-id: bb4f0589ed23bf0236d3e918272ad64f0a05ef39	2023-10-22 17:42:13 +08:00
hiyouga	49e8a87383	fix webui Former-commit-id: a5a5a7bc1f53d36e1b26e418999465903cb7d9ed	2023-10-22 17:24:56 +08:00
hiyouga	b2764b49ca	add new options in webui Former-commit-id: 6698b832dd9cc2d7d60be4fa5ab90e34a7e9d8e0	2023-10-22 17:17:58 +08:00
hiyouga	06b810de8f	fix recursion error Former-commit-id: c7938188c36a71a878bca982b7dd151195164986	2023-10-22 16:28:37 +08:00
hiyouga	6da51565f5	reimplement neftune Former-commit-id: efe9e5a194d3a9f052701d904715238816e4c09e	2023-10-22 16:15:08 +08:00
hoshi-hiyouga	1f69965239	Merge pull request #1252 from anvie/neftune add NEFTune optimization Former-commit-id: 85d5c5fbe731f486c3e83812227fa05edc131487	2023-10-22 15:59:20 +08:00
anvie	af2d61178d	add NEFTune optimization Former-commit-id: 603e0298af64116ac07130fe6661a9ba823c186c	2023-10-21 13:24:10 +07:00
hiyouga	6a955ccf4f	fix openchat template Former-commit-id: 88b9b657bc50495ac4c42f64195fc652fe4ca3df	2023-10-21 01:25:42 +08:00
hiyouga	c0658711ca	fix tokenizer padding side in evaluate.py Former-commit-id: bcb43ff8ba1946c1f7e7865c9d0fb47ba276935d	2023-10-21 00:30:04 +08:00
hiyouga	d602f06882	fix #1232 Former-commit-id: 49975755d47344e362145c52548fdda8783f2c0c	2023-10-20 23:28:52 +08:00
hiyouga	1cb9a38ac2	fix #1215 Former-commit-id: d91b43a8afbea4859357f2224e3d9b9d71160e6d	2023-10-19 16:19:21 +08:00
hiyouga	47a1f73d0f	fix #1218 Former-commit-id: b301f35bd4a3bf368159c8f5fb4e2736f922115b	2023-10-19 16:17:41 +08:00
hiyouga	142dd63b47	fix #1228 Former-commit-id: e4e0cae3f55da2f1b566c97dbfdd7fc5b7b728a4	2023-10-19 15:54:10 +08:00
hiyouga	b1bd8370c2	fix #1217 Former-commit-id: 065fc0a6f3f005bb87e1c5c126c8b6bb470ce700	2023-10-19 15:52:24 +08:00
hiyouga	215660c8da	rename webui Former-commit-id: 26feaf80fff6177d9eb4e28ad18feb6d34d3ea27	2023-10-16 15:16:24 +08:00
hiyouga	0cafe67efe	fix #1197 Former-commit-id: 00100e23fcfef9587fda4cf01c62599d996e1176	2023-10-16 15:13:46 +08:00
hoshi-hiyouga	ea83b3222b	Update README_zh.md Former-commit-id: 3450404bb9a33c3bd4b45ac4afcf51062f8c7d1d	2023-10-16 00:28:27 +08:00
hoshi-hiyouga	725087a04f	Update README.md Former-commit-id: d84896597eded79f78224faed81cc9f2df222978	2023-10-16 00:23:37 +08:00