Merge commit '6cead7123ef0a35d34fe755bea2e5a61cd275092' into feat/webui_0228

* commit '6cead7123ef0a35d34fe755bea2e5a61cd275092':
  Fix llama2 generation config (modelscope#468)
  swift export load model in cpu (modelscope#462)
  fix get_vllm_engine bug (modelscope#463)
  Fix llm quantization docs (modelscope#458)
  Support swift export (modelscope#455)

Author: tastelikefeet

README.md

@@ -62,6 +62,7 @@ Users can check the [documentation of SWIFT](docs/source/GetStarted/快速使用
 
 
 ## 🎉 News
+- 2024.02.25: Support `swift export` to export models for AWQ quantization and push to ModelScope Hub. For more details, please refer to the document: [LLM Quantization Document](https://github.com/modelscope/swift/blob/main/docs/source/LLM/LLM%E9%87%8F%E5%8C%96%E6%96%87%E6%A1%A3.md).
 - 2024.02.22: Support gemma series: gemma-2b, [gemma-2b-instruct](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/gemma_2b_instruct), gemma-7b, gemma-7b-instruct.
 - 2024.02.16: Support deepseek-math series: deepseek-math-7b, deepseek-math-7b-instruct, deepseek-math-7b-chat.
 - 🔥2024.02.05: Support **Qwen1.5** series, To view all supported Qwen1.5 models please check [Model List](https://github.com/modelscope/swift/blob/main/docs/source/LLM/%E6%94%AF%E6%8C%81%E7%9A%84%E6%A8%A1%E5%9E%8B%E5%92%8C%E6%95%B0%E6%8D%AE%E9%9B%86.md#%E6%A8%A1%E5%9E%8B). The [qwen1half-7b-chat](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/qwen1half_7b_chat), [qwen1half-7b-chat-int8](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/qwen1half_7b_chat_int8) fine-tuned scripts are provided.
@@ -71,9 +72,9 @@ Users can check the [documentation of SWIFT](docs/source/GetStarted/快速使用
 - 🔥2024.02.01: Support Agent training! Agent training algorithm comes from this [paper](https://arxiv.org/pdf/2309.00986.pdf). We also introduce the [ms-agent](https://www.modelscope.cn/datasets/iic/ms_agent/summary) dataset. Use [this script](https://github.com/modelscope/swift/blob/main/examples/pytorch/llm/scripts/qwen_7b_chat/lora/sft.sh) to begin an agent training!
 - 🔥2024.02.01: Support SFT loss to DPO training to reduce the repeat generation problem caused by the KL-divergence loss.
 - 2024.02.01: Support AdaLoRA and IA3 adapter in SFT.
-- 2024.02.01: Support `--merge_lora_and_save` in AnimateDiff training.
+- 2024.02.01: Support `--merge_lora` in AnimateDiff training.
 - 2024.01.30: Support [internlm-xcomposer2-7b-chat](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/internlm_xcomposer2_7b_chat).
-- 🔥2024.01.30: Support [ZeRO-3](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/qwen_14b_chat/full_ddp_zero3/), just need to specify `--deepspeed_config_path default-zero3`.
+- 🔥2024.01.30: Support [ZeRO-3](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/qwen_14b_chat/full_ddp_zero3/), just need to specify `--deepspeed default-zero3`.
 - 2024.01.29: Support internlm2-math series: internlm2-math-7b, internlm2-math-7b-chat, internlm2-math-20b, internlm2-math-20b-chat.
 - 🔥2024.01.26: Support [yi-vl-6b-chat](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/yi_vl_6b_chat), yi-vl-34b-chat.
 - 2024.01.24: Support codefuse-codegeex2-6b-chat, codefuse-qwen-14b-chat.
@@ -154,6 +155,7 @@ Here is a simple introduction of web-ui:
 - Rapidly **fine-tune** and perform inference on LLM, and build a Web-UI, see the [LLM Fine-tuning Documentation](https://github.com/modelscope/swift/blob/main/docs/source/LLM/LLM微调文档.md).
 - Using **interface** to fine-tuning and perform inference, see the [WEB-UI Documentation](https://github.com/modelscope/swift/blob/main/docs/source/GetStarted/%E7%95%8C%E9%9D%A2%E8%AE%AD%E7%BB%83%E6%8E%A8%E7%90%86.md).
 - **DPO training** supported, see the [DPO Documentation](https://github.com/modelscope/swift/blob/main/docs/source/LLM/LLM%E4%BA%BA%E7%B1%BB%E5%AF%B9%E9%BD%90%E8%AE%AD%E7%BB%83%E6%96%87%E6%A1%A3.md).
+- Export fine-tuned models, including: merge-lora, AWQ quantization, and push to ModelScope Hub. For more details, please refer to the [LLM Quantization Documentation](https://github.com/modelscope/swift/blob/main/docs/source/LLM/LLM%E9%87%8F%E5%8C%96%E6%A8%A1%E5%9E%8B.md).
 - Utilize VLLM for **inference acceleration** and **deployment(OpenAI API)**. Please refer to [VLLM Inference Acceleration and Deployment](https://github.com/modelscope/swift/blob/main/docs/source/LLM/VLLM推理加速与部署.md) for more information.
 - View the models and datasets supported by Swift. You can check [supported models and datasets](https://github.com/modelscope/swift/blob/main/docs/source/LLM/支持的模型和数据集.md).
 - Expand and customize models, datasets, and dialogue templates in Swift, see [Customization and Expansion](https://github.com/modelscope/swift/blob/main/docs/source/LLM/自定义与拓展.md).
@@ -266,6 +268,7 @@ app_ui_main(infer_args)
     - SQL: text2sql-en, 🔥sql-create-context-en.
     - Text Generation: 🔥advertise-gen-zh, 🔥dureader-robust-zh.
     - Classification: cmnli-zh, 🔥cmnli-mini-zh, 🔥jd-sentiment-zh, 🔥hc3-zh, 🔥hc3-en.
+    - AWQ: pileval.
     - Other: finance-en, poetry-zh, webnovel-zh, generated-chat-zh, cls-fudan-news-zh, ner-jave-zh.
   - Multi-Modal:
     - Vision: coco-en, 🔥coco-mini-en, coco-mini-en-2, capcha-images.

README_CN.md

@@ -60,6 +60,7 @@ SWIFT（Scalable lightWeight Infrastructure for Fine-Tuning）是一个可扩展
 用户可以查看 [SWIFT官方文档](docs/source/GetStarted/快速使用.md) 来了解详细信息。
 
 ## 🎉 新闻
+- 2024.02.25: 支持`swift export`, 对模型进行AWQ量化导出, 以及推送ModelScope Hub. 具体可以查看文档: [LLM量化文档](https://github.com/modelscope/swift/blob/main/docs/source/LLM/LLM%E9%87%8F%E5%8C%96%E6%96%87%E6%A1%A3.md).
 - 2024.02.22: 支持gemma系列: gemma-2b, [gemma-2b-instruct](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/gemma_2b_instruct), gemma-7b, gemma-7b-instruct.
 - 2024.02.16: 支持deepseek-math系列: deepseek-math-7b, deepseek-math-7b-instruct, deepseek-math-7b-chat.
 - 🔥2024.02.05: 支持**Qwen1.5**系列模型, 支持的所有Qwen1.5系列模型请查看[模型列表](https://github.com/modelscope/swift/blob/main/docs/source/LLM/%E6%94%AF%E6%8C%81%E7%9A%84%E6%A8%A1%E5%9E%8B%E5%92%8C%E6%95%B0%E6%8D%AE%E9%9B%86.md#%E6%A8%A1%E5%9E%8B). 提供了[qwen1half-7b-chat](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/qwen1half_7b_chat), [qwen1half-7b-chat-int8](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/qwen1half_7b_chat_int8)微调的脚本.
@@ -69,9 +70,9 @@ SWIFT（Scalable lightWeight Infrastructure for Fine-Tuning）是一个可扩展
 - 🔥2024.02.01: 支持Agent训练！Agent训练算法源自这篇[论文](https://arxiv.org/pdf/2309.00986.pdf). 我们也增加了[ms-agent](https://www.modelscope.cn/datasets/iic/ms_agent/summary)这个优质的agent数据集. 使用[这个脚本](https://github.com/modelscope/swift/blob/main/examples/pytorch/llm/scripts/qwen_7b_chat/lora/sft.sh)开启Agent训练!
 - 🔥2024.02.01: 支持在DPO训练中增加SFT loss来减少KL散度loss造成的生成重复问题.
 - 2024.02.01: 支持在训练中使用AdaLoRA和IA3两个adapter.
-- 2024.02.01: 支持在AnimateDiff训练中使用`--merge_lora_and_save`参数.
+- 2024.02.01: 支持在AnimateDiff训练中使用`--merge_lora`参数.
 - 2024.01.30: 支持[internlm-xcomposer2-7b-chat](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/internlm_xcomposer2_7b_chat).
-- 🔥2024.01.30: 支持[ZeRO-3](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/qwen_14b_chat/full_ddp_zero3/), 只需要指定`--deepspeed_config_path default-zero3`即可.
+- 🔥2024.01.30: 支持[ZeRO-3](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/qwen_14b_chat/full_ddp_zero3/), 只需要指定`--deepspeed default-zero3`即可.
 - 2024.01.29: 支持internlm2-math系列: internlm2-math-7b, internlm2-math-7b-chat, internlm2-math-20b, internlm2-math-20b-chat.
 - 🔥2024.01.26: 支持[yi-vl-6b-chat](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts/yi_vl_6b_chat), yi-vl-34b-chat.
 - 2024.01.24: 支持codefuse-codegeex2-6b-chat, codefuse-qwen-14b-chat.
@@ -154,6 +155,7 @@ swift web-ui
 - 快速对LLM进行**微调**, 推理并搭建Web-UI, 可以查看[LLM微调文档](https://github.com/modelscope/swift/blob/main/docs/source/LLM/LLM微调文档.md).
 - 使用**界面**方式进行微调和推理, 可以查看[WEB-UI文档](https://github.com/modelscope/swift/blob/main/docs/source/GetStarted/%E7%95%8C%E9%9D%A2%E8%AE%AD%E7%BB%83%E6%8E%A8%E7%90%86.md).
 - 支持**DPO训练**, 可以查看[DPO文档](https://github.com/modelscope/swift/blob/main/docs/source/LLM/LLM%E4%BA%BA%E7%B1%BB%E5%AF%B9%E9%BD%90%E8%AE%AD%E7%BB%83%E6%96%87%E6%A1%A3.md).
+- 对微调的模型进行导出, 包括: merge-lora, AWQ量化, 推送ModelScope Hub, 可以查看[LLM量化文档](https://github.com/modelscope/swift/blob/main/docs/source/LLM/LLM%E9%87%8F%E5%8C%96%E6%A8%A1%E5%9E%8B.md).
 - 使用VLLM进行**推理加速**和**部署(OpenAI API)**. 可以查看[VLLM推理加速与部署](https://github.com/modelscope/swift/blob/main/docs/source/LLM/VLLM推理加速与部署.md).
 - 查看swift支持的模型和数据集. 可以查看[支持的模型和数据集](https://github.com/modelscope/swift/blob/main/docs/source/LLM/支持的模型和数据集.md).
 - 对swift中的模型, 数据集, 对话模板进行**拓展**, 可以查看[自定义与拓展](https://github.com/modelscope/swift/blob/main/docs/source/LLM/自定义与拓展.md).
@@ -265,6 +267,7 @@ app_ui_main(infer_args)
     - SQL: text2sql-en, 🔥sql-create-context-en.
     - 文本生成: 🔥advertise-gen-zh, 🔥dureader-robust-zh.
     - 分类: cmnli-zh, 🔥cmnli-mini-zh, 🔥jd-sentiment-zh, 🔥hc3-zh, 🔥hc3-en.
+    - AWQ: pileval.
     - 其他: finance-en, poetry-zh, webnovel-zh, generated-chat-zh, cls-fudan-news-zh, ner-jave-zh.
   - 多模态:
     - 视觉: coco-en, 🔥coco-mini-en, coco-mini-en-2, capcha-images.

docs/source/AIGC/AnimateDiff微调推理文档.md

@@ -176,7 +176,7 @@ dataloader_num_workers: int = 1 # dataloader workers数量
 push_to_hub: bool = False # 是否推送到modelhub
 # 'user_name/repo_name' or 'repo_name'
 hub_model_id: Optional[str] = None # modelhub id
-hub_private_repo: bool = True
+hub_private_repo: bool = False
 push_hub_strategy: str = field( # 推送策略，推送最后一个还是每个都推送
     default='push_best',
     metadata={'choices': ['push_last', 'all_checkpoints']})
@@ -244,13 +244,13 @@ sft_type: str = field(
     default='lora', metadata={'choices': ['lora', 'full']}) # 训练方式，支持lora和全参数
 
 ckpt_dir: Optional[str] = field(
-    default=None, metadata={'help': '/path/to/your/vx_xxx/checkpoint-xxx'}) # 训练的输出文件夹
+    default=None, metadata={'help': '/path/to/your/vx-xxx/checkpoint-xxx'}) # 训练的输出文件夹
 eval_human: bool = False  # False: eval val_dataset # 是否使用人工输入评测
 
 seed: int = 42 # 随机种子
 
-merge_lora_and_save: bool = False # Merge lora into the MotionAdapter and save the model.
-replace_if_exists: bool = False # Replace the files if the output merged dir exists when `merge_lora_and_save` is True.
+merge_lora: bool = False # Merge lora into the MotionAdapter and save the model.
+replace_if_exists: bool = False # Replace the files if the output merged dir exists when `merge_lora` is True.
 
 # other
 ignore_args_error: bool = False  # True: notebook compatibility

docs/source/LLM/Agent微调最佳实践.md

@@ -11,7 +11,7 @@
 ## 环境安装
 
 ```bash
-# 设置pip全局镜像
+# 设置pip全局镜像 (加速下载)
 pip config set global.index-url https://mirrors.aliyun.com/pypi/simple/
 # 安装ms-swift
 git clone https://github.com/modelscope/swift.git

docs/source/LLM/LLM人类对齐训练文档.md

@@ -6,7 +6,7 @@
 ## 环境准备
 GPU设备: A10, 3090, V100, A100均可，如果是显存<=24G的GPU最少需要双卡环境。由于人类对齐训练在一张卡上加载两个模型，因此比微调的显存多占用一个推理模型的显存使用量。
 ```bash
-# 设置pip全局镜像
+# 设置pip全局镜像 (加速下载)
 pip config set global.index-url https://mirrors.aliyun.com/pypi/simple/
 # 安装ms-swift
 git clone https://github.com/modelscope/swift.git
@@ -84,7 +84,7 @@ cd examples/pytorch/llm
 - 我们默认在训练时设置`--gradient_checkpointing true`来**节约显存**, 这会略微降低训练速度.
 - 如果你使用的是**V100**等较老的GPU, 你需要设置`--dtype AUTO`或者`--dtype fp16`, 因为其不支持bf16.
 - 如果你的机器是A100等高性能显卡, 且使用的是qwen系列模型, 推荐你安装[**flash-attn**](https://github.com/Dao-AILab/flash-attention), 这将会加快训练和推理的速度以及显存占用(A10, 3090, V100等显卡不支持flash-attn进行训练). 支持flash-attn的模型可以查看[LLM支持的模型](./支持的模型和数据集.md#模型)
-- 如果你需要断网进行训练, 请使用`--model_cache_dir`和设置`--check_model_is_latest false`. 具体参数含义请查看[命令行参数](./命令行参数.md).
+- 如果你需要断网进行训练, 请使用`--model_id_or_path <model_dir>`和设置`--check_model_is_latest false`. 具体参数含义请查看[命令行参数](./命令行参数.md).
 - 如果你想在训练时, 将权重push到ModelScope Hub中, 你需要设置`--push_to_hub true`.
 
 ```bash

docs/source/LLM/LLM微调文档.md

@@ -4,13 +4,14 @@
 - [微调](#微调)
 - [DPO](#dpo)
 - [Merge LoRA](#merge-lora)
+- [量化](#量化)
 - [推理](#推理)
 - [Web-UI](#web-ui)
 
 ## 环境准备
 GPU设备: A10, 3090, V100, A100均可.
 ```bash
-# 设置pip全局镜像
+# 设置pip全局镜像 (加速下载)
 pip config set global.index-url https://mirrors.aliyun.com/pypi/simple/
 # 安装ms-swift
 git clone https://github.com/modelscope/swift.git
@@ -76,15 +77,13 @@ app_ui_main(infer_args)
 ```bash
 # Experimental environment: A10, 3090, V100, ...
 # 20GB GPU memory
-CUDA_VISIBLE_DEVICES=0 \
-swift sft \
+CUDA_VISIBLE_DEVICES=0 swift sft \
     --model_id_or_path qwen/Qwen-7B-Chat \
     --dataset blossom-math-zh \
     --output_dir output \
 
 # 使用自己的数据集
-CUDA_VISIBLE_DEVICES=0 \
-swift sft \
+CUDA_VISIBLE_DEVICES=0 swift sft \
     --model_id_or_path qwen/Qwen-7B-Chat \
     --custom_train_dataset_path chatml.jsonl \
     --output_dir output \
@@ -146,9 +145,9 @@ cd examples/pytorch/llm
 - 如果你使用的是**V100**等较老的GPU, 你需要设置`--dtype AUTO`或者`--dtype fp16`, 因为其不支持bf16.
 - 如果你的机器是A100等高性能显卡, 且使用的是qwen系列模型, 推荐你安装[**flash-attn**](https://github.com/Dao-AILab/flash-attention), 这将会加快训练和推理的速度以及显存占用(A10, 3090, V100等显卡不支持flash-attn进行训练). 支持flash-attn的模型可以查看[LLM支持的模型](./支持的模型和数据集.md#模型)
 - 如果你要进行**二次预训练**, **多轮对话**, 你可以参考[自定义与拓展](./自定义与拓展.md#注册数据集的方式)
-- 如果你需要**断网**进行训练, 请使用`--model_cache_dir`和设置`--check_model_is_latest false`. 具体参数含义请查看[命令行参数](./命令行参数.md).
+- 如果你需要**断网**进行训练, 请使用`--model_id_or_path <model_dir>`和设置`--check_model_is_latest false`. 具体参数含义请查看[命令行参数](./命令行参数.md).
 - 如果你想在训练时, 将权重push到ModelScope Hub中, 你需要设置`--push_to_hub true`.
-- 如何你想要在推理时, 合并LoRA权重并保存，你需要设置`--merge_lora_and_save true`. **不推荐对qlora训练的模型进行merge**, 这会存在精度损失.
+- 如何你想要在推理时, 合并LoRA权重并保存，你需要设置`--merge_lora true`. **不推荐对qlora训练的模型进行merge**, 这会存在精度损失.
 - 以下提供了可以直接运行的`qwen_7b_chat`的sh脚本(你只需要在推理时指定`--ckpt_dir`即可顺利执行). 更多模型的scripts脚本, 可以查看[scripts文件夹](https://github.com/modelscope/swift/tree/main/examples/pytorch/llm/scripts). 如果你想要**自定义sh脚本**, 推荐你参考`scripts/qwen_7b_chat`中的脚本进行书写.
 
 ```bash
@@ -225,9 +224,15 @@ bash scripts/qwen_7b_chat/qlora_ddp_ds/infer.sh
 ## Merge LoRA
 提示: **暂时**不支持bnb和auto_gptq量化模型的merge lora, 这会产生较大的精度损失.
 ```bash
-swift merge-lora --ckpt_dir 'xxx/vx_xxx/checkpoint-xxx'
+# 如果你需要量化, 可以指定`--quant_bits 4`.
+CUDA_VISIBLE_DEVICES=0 swift export \
+    --ckpt_dir 'xxx/vx-xxx/checkpoint-xxx' --merge_lora true
 ```
 
+## 量化
+
+对微调后模型进行量化可以查看[LLM量化文档](LLM量化文档.md#微调后模型)
+
 ## 推理
 如果你要使用VLLM进行推理加速, 可以查看[VLLM推理加速与部署](./VLLM推理加速与部署.md#微调后的模型)
 
@@ -251,13 +256,13 @@ from swift.llm import (
 )
 from swift.tuners import Swift
 
-model_dir = 'vx_xxx/checkpoint-100'
+ckpt_dir = 'vx-xxx/checkpoint-100'
 model_type = ModelType.qwen_7b_chat
 template_type = get_default_template_type(model_type)
 
 model, tokenizer = get_model_tokenizer(model_type, model_kwargs={'device_map': 'auto'})
 
-model = Swift.from_pretrained(model, model_dir, inference_mode=True)
+model = Swift.from_pretrained(model, ckpt_dir, inference_mode=True)
 template = get_template(template_type, tokenizer)
 query = 'xxxxxx'
 response, history = inference(model, template, query)
@@ -274,12 +279,12 @@ from swift.llm import (
     get_model_tokenizer, get_template, inference, ModelType, get_default_template_type
 )
 
-model_dir = 'vx_xxx/checkpoint-100-merged'
+ckpt_dir = 'vx-xxx/checkpoint-100-merged'
 model_type = ModelType.qwen_7b_chat
 template_type = get_default_template_type(model_type)
 
 model, tokenizer = get_model_tokenizer(model_type, model_kwargs={'device_map': 'auto'},
-                                       model_dir=model_dir)
+                                       model_id_or_path=ckpt_dir)
 
 template = get_template(template_type, tokenizer)
 query = 'xxxxxx'
@@ -291,27 +296,30 @@ print(f'history: {history}')
 使用**数据集**评估:
 ```bash
 # 直接推理
-CUDA_VISIBLE_DEVICES=0 \
-swift infer \
-    --ckpt_dir 'xxx/vx_xxx/checkpoint-xxx' \
+CUDA_VISIBLE_DEVICES=0 swift infer \
+    --ckpt_dir 'xxx/vx-xxx/checkpoint-xxx' \
     --load_dataset_config true \
 
 # Merge LoRA增量权重并推理
-swift merge-lora --ckpt_dir 'xxx/vx_xxx/checkpoint-xxx'
-CUDA_VISIBLE_DEVICES=0 \
-swift infer \
-    --ckpt_dir 'xxx/vx_xxx/checkpoint-xxx-merged' \
-    --load_dataset_config true \
+# 如果你需要量化, 可以指定`--quant_bits 4`.
+CUDA_VISIBLE_DEVICES=0 swift export \
+    --ckpt_dir 'xxx/vx-xxx/checkpoint-xxx' --merge_lora true
+
+CUDA_VISIBLE_DEVICES=0 swift infer \
+    --ckpt_dir 'xxx/vx-xxx/checkpoint-xxx-merged' --load_dataset_config true
 ```
 
 **人工**评估:
 ```bash
 # 直接推理
-CUDA_VISIBLE_DEVICES=0 swift infer --ckpt_dir 'xxx/vx_xxx/checkpoint-xxx'
+CUDA_VISIBLE_DEVICES=0 swift infer --ckpt_dir 'xxx/vx-xxx/checkpoint-xxx'
 
 # Merge LoRA增量权重并推理
-swift merge-lora --ckpt_dir 'xxx/vx_xxx/checkpoint-xxx'
-CUDA_VISIBLE_DEVICES=0 swift infer --ckpt_dir 'xxx/vx_xxx/checkpoint-xxx-merged'
+# 如果你需要量化, 可以指定`--quant_bits 4`.
+CUDA_VISIBLE_DEVICES=0 swift export \
+    --ckpt_dir 'xxx/vx-xxx/checkpoint-xxx' --merge_lora true
+
+CUDA_VISIBLE_DEVICES=0 swift infer --ckpt_dir 'xxx/vx-xxx/checkpoint-xxx-merged'
 ```
 
 ## Web-UI
@@ -323,9 +331,12 @@ CUDA_VISIBLE_DEVICES=0 swift infer --ckpt_dir 'xxx/vx_xxx/checkpoint-xxx-merged'
 ### 微调后模型
 ```bash
 # 直接使用app-ui
-CUDA_VISIBLE_DEVICES=0 swift app-ui --ckpt_dir 'xxx/vx_xxx/checkpoint-xxx'
+CUDA_VISIBLE_DEVICES=0 swift app-ui --ckpt_dir 'xxx/vx-xxx/checkpoint-xxx'
 
 # merge LoRA增量权重并使用app-ui
-swift merge-lora --ckpt_dir 'xxx/vx_xxx/checkpoint-xxx'
-CUDA_VISIBLE_DEVICES=0 swift app-ui --ckpt_dir 'xxx/vx_xxx/checkpoint-xxx-merged'
+# 如果你需要量化, 可以指定`--quant_bits 4`.
+CUDA_VISIBLE_DEVICES=0 swift export \
+    --ckpt_dir 'xxx/vx-xxx/checkpoint-xxx' --merge_lora true
+
+CUDA_VISIBLE_DEVICES=0 swift app-ui --ckpt_dir 'xxx/vx-xxx/checkpoint-xxx-merged'
 ```

docs/source/LLM/LLM推理文档.md

@@ -9,7 +9,7 @@
 ## 环境准备
 GPU设备: A10, 3090, V100, A100均可.
 ```bash
-# 设置pip全局镜像
+# 设置pip全局镜像 (加速下载)
 pip config set global.index-url https://mirrors.aliyun.com/pypi/simple/
 # 安装ms-swift
 git clone https://github.com/modelscope/swift.git