按照官方文档修改

backend/app/services/dataset_service.py

@@ -104,21 +104,22 @@ async def download_dataset(req: DatasetDownloadRequest) -> DatasetDownloadRespon
 
 
 def _download_modelscope_dataset(dataset_id: str) -> tuple[Path, Path, int]:
-    """用 MsDataset 下载并转为 JSONL。"""
-    from modelscope.msdatasets import MsDataset
+    """按官方文档推荐方式：MsDataset 加载并转为 JSONL。"""
+    from modelscope import MsDataset
 
-    # 先尝试加载完整数据集
+    # 按官方文档推荐方式加载，优先使用 train split
     try:
-        ds = MsDataset.load(dataset_id)
+        ds = MsDataset.load(dataset_id, split='train')
     except Exception:
-        try:
-            ds = MsDataset.load(dataset_id, subset_name="default", split="train")
-        except Exception:
-            ds = MsDataset.load(dataset_id, split="train")
+        # 部分数据集可能没有 train split，尝试加载完整数据集
+        ds = MsDataset.load(dataset_id)
 
     ds_dir = settings.processed_dir / f"ms_{dataset_id.replace('/', '_')}"
     ds_dir.mkdir(parents=True, exist_ok=True)
 
+    # 如果是 DatasetDict（有多个 split），取第一个 split 的数据
+    split_data = ds if not hasattr(ds, "keys") else ds[list(ds.keys())[0]]
+
     # 如果是 DatasetDict，取第一个 split
     split_data = ds if not hasattr(ds, "keys") else ds[list(ds.keys())[0]]
 

backend/app/services/model_service.py

@@ -43,9 +43,10 @@ async def download_model(model_id: str, use_modelscope: bool = False) -> dict[st
     """从 HF 或 ModelScope 下载模型到本地缓存。"""
     try:
         if use_modelscope:
-            from modelscope import snapshot_download as ms_download
+            from modelscope.hub.snapshot_download import snapshot_download as ms_download
 
-            local_path = ms_download(model_id, cache_dir=str(settings.models_dir))
+            download_dir = str(settings.models_dir / model_id.replace("/", "_"))
+            local_path = ms_download(model_id, local_dir=download_dir)
         else:
             from huggingface_hub import snapshot_download
 

数据集的下载.ipynb

@@ -0,0 +1,392 @@
+{
+  "cells": [
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "在本篇文章中，我们将向您介绍如何在ModelScope平台下载数据集。无论您是想使用Python SDK、GIT还是相应的命令行工具，这里都有详细的步骤帮助您快速上手。",
+        "",
+        "# 快速使用",
+        "",
+        "## 下载指定的文件",
+        "1. 访问数据集页面。",
+        "2. 点击每个文件的`下载`按钮直接下载该文件。",
+        "",
+        "![image.png](https://resouces.modelscope.cn/document/docdata/2026-5-14_11-1-CN/dist/datasets/download/_resources/E9C0010F-A041-4770-9D6F-891531680E8C.png)",
+        "",
+        "## 下载整个数据集",
+        "点击`下载数据集`按钮，即可看到使用SDK、GIT或命令行下载数据集的具体命令。",
+        "",
+        "![image.png](https://resouces.modelscope.cn/document/docdata/2026-5-14_11-1-CN/dist/datasets/download/_resources/download_dataset1.png)",
+        "",
+        "-------",
+        "",
+        "下面对每种方法进行详细介绍：",
+        "",
+        "",
+        "#  1. 使用Python SDK下载数据集",
+        "您可以通过`modelscope`库轻松下载数据集，并进行相关的操作（创建、删除、更新和检索信息）。",
+        "",
+        "## 安装Python SDK",
+        "",
+        "首先，确保您已安装ModelScope的Python SDK，使用以下命令：",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "pip install modelscope[framework]"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "## 下载数据集示例",
+        "",
+        "使用以下代码下载数据集：",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "from modelscope import MsDataset",
+        "",
+        "# 加载数据集",
+        "ds = MsDataset.load('afqmc_small', split='train')"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "# 2. 使用GIT下载数据集",
+        "",
+        "如果您更喜欢使用GIT，可以通过以下方式下载数据集。",
+        "",
+        "",
+        "## 公开数据集下载",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "# 安装Git LFS（大文件存储）",
+        "git lfs install",
+        "",
+        "# 克隆公开数据集",
+        "git clone https://www.modelscope.cn/datasets/<namespace>/<dataset-name>.git",
+        "# 示例: git clone https://www.modelscope.cn/datasets/DAMO_NLP/jd.git"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "## 私有数据集下载",
+        "",
+        "**注意：** 您需要相应的数据集权限。",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "# 使用Git LFS",
+        "git lfs install",
+        "",
+        "# 克隆私有数据集",
+        "git clone http://oauth2:<your_access_token>@www.modelscope.cn/datasets/<namespace>/<dataset-name>.git"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "或者使用以下方法：",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "git clone http://<your_user_name>@www.modelscope.cn/datasets/<namespace>/<dataset-name>.git",
+        "# 输入您的访问令牌（Access Token）作为密码"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "### 获取访问令牌",
+        "",
+        "1. 登录 [ModelScope官网](https://www.modelscope.cn)。",
+        "2. 进入个人中心 -> 访问令牌，复制您的访问令牌。",
+        "",
+        "",
+        "# 3. 使用命令行工具下载数据集文件",
+        "",
+        "通过命令行，您可以灵活下载数据集文件。",
+        "",
+        "## 命令格式",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --dataset DATASET_NAME [options] [files ...]"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "## 参数说明",
+        "",
+        "| 参数 | 简写 | 类型 | 默认值 | 说明 |",
+        "|------|------|------|--------|------|",
+        "| `--dataset` | - | str | None | 数据集 ID（必需） |",
+        "| `files` | - | str | - | 位置参数，指定要下载的文件（支持多个） |",
+        "| `--revision` | - | str | None | 版本/分支/tag |",
+        "| `--cache_dir` | - | str | None | 缓存目录 |",
+        "| `--local_dir` | - | str | None | 本地目录（优先于 cache_dir） |",
+        "| `--include` | - | list | None | 包含的文件 glob 模式 |",
+        "| `--exclude` | - | list | None | 排除的文件 glob 模式 |",
+        "| `--token` | - | str | None | 访问令牌（私有数据集需要） |",
+        "| `--endpoint` | - | str | None | ModelScope 服务端点 |",
+        "| `--max-workers` | - | int | 默认 | 最大并发下载线程数 |",
+        "",
+        "",
+        "## 使用示例",
+        "",
+        "命令示例（以[SA1B-Dense-Caption](https://modelscope.cn/datasets/Tongyi-DataEngine/SA1B-Dense-Caption)）为例",
+        "",
+        "1. 指定下载单个文件",
+        "    ",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --dataset 'Tongyi-DataEngine/SA1B-Dense-Caption' dataset_infos.json"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "2. 指定下载多个个文件  ",
+        "    ",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --dataset 'Tongyi-DataEngine/SA1B-Dense-Caption' dataset_infos.json README.md"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "3. 指定下载某些文件 ",
+        "    ",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --dataset 'Tongyi-DataEngine/SA1B-Dense-Caption' --include 'data/train-000*'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "4. 过滤指定文件",
+        "    ",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --dataset 'Tongyi-DataEngine/SA1B-Dense-Caption' --exclude 'data/train-000*'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "5. 指定下载cache\\_dir ",
+        "    ",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --dataset 'Tongyi-DataEngine/SA1B-Dense-Caption' --include 'data/train-000*' --cache_dir './cache_dir'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "    模型文件将被下载在`./cache_dir`",
+        "",
+        "6. 指定下载local\\_dir    ",
+        "    ",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --dataset 'Tongyi-DataEngine/SA1B-Dense-Caption' --include 'data/train-000*' --local_dir './local_dir'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "    模型文件将被下载在`./local_dir`",
+        "",
+        "注意：若`cache_dir`和`local_dir`同时指定，`local_dir`优先级高，`cache_dir`将被忽略",
+        "",
+        "## 下载私有模型时需要访问令牌",
+        "",
+        "> 您可以在[我的页面](https://modelscope.cn/my/myaccesstoken)获取**访问令牌**",
+        "",
+        "### 方式1：先登录再下载",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope login --token YOUR_MODELSCOPE_ACCESS_TOKEN"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "### 方式二：指定token参数下载",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope --token 'YOUR_MODELSCOPE_ACCESS_TOKEN' download --dataset 'YOUR_DATASET_NAME'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        ""
+      ]
+    }
+  ],
+  "metadata": {
+    "kernelspec": {
+      "display_name": "Python 3",
+      "language": "python",
+      "name": "python3"
+    }
+  },
+  "nbformat": 4,
+  "nbformat_minor": 2
+}

模型的下载.ipynb

@@ -0,0 +1,489 @@
+{
+  "cells": [
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "本文介绍在 ModelScope 社区下载模型的三种方式：",
+        "",
+        "1. 使用命令行工具下载",
+        "2. 使用 SDK 下载",
+        "3. 通过 Git 下载",
+        "",
+        "# 模型下载默认存放地址",
+        "",
+        "无论是使用命令行还是ModelScope SDK，模型会下载到`~/.cache/modelscope/hub`默认路径下。如果需要修改 cache 目录，可以手动设置环境变量：`MODELSCOPE_CACHE`，完成设置后，模型将下载到该环境变量指定的目录中。",
+        "",
+        "# 使用命令行工具下载",
+        "",
+        "**`modelscope download` 参数说明**",
+        "",
+        "| 参数 | 简写 | 类型 | 默认值 | 说明 |",
+        "|------|------|------|--------|------|",
+        "| `repo_id` | - | str | - | 位置参数，仓库 ID（可选，也可通过 `--model` 指定） |",
+        "| `files` | - | str | - | 位置参数，指定要下载的文件（支持多个） |",
+        "| `--model` | - | str | None | 模型 ID（与 `--dataset` 互斥） |",
+        "| `--dataset` | - | str | None | 数据集 ID（与 `--model` 互斥） |",
+        "| `--repo-type` | - | choice | `model` | 仓库类型（model/dataset），与位置参数 repo_id 配合使用 |",
+        "| `--revision` | - | str | None | 版本/分支/tag |",
+        "| `--cache_dir` | - | str | None | 缓存目录 |",
+        "| `--local_dir` | - | str | None | 本地目录（优先于 cache_dir） |",
+        "| `--include` | - | list | None | 包含的文件 glob 模式 |",
+        "| `--exclude` | - | list | None | 排除的文件 glob 模式 |",
+        "| `--token` | - | str | None | 访问令牌（私有模型需要） |",
+        "| `--endpoint` | - | str | None | ModelScope 服务端点 |",
+        "| `--max-workers` | - | int | 默认 | 最大并发下载线程数 |",
+        "",
+        "## 使用示例",
+        "",
+        "命令示例（以[Qwen2-7B](https://www.modelscope.cn/models/Qwen/Qwen2-7b)）模型为例",
+        "",
+        "### 下载整个模型repo（到默认cache地址）",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --model 'Qwen/Qwen2-7b'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "### 下载整个模型repo到指定目录",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --model 'Qwen/Qwen2-7b' --local_dir 'path/to/dir'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "### 指定下载单个文件(以'tokenizer.json'文件为例)",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --model 'Qwen/Qwen2-7b' tokenizer.json"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "### 指定下载多个个文件",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --model 'Qwen/Qwen2-7b' tokenizer.json config.json"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "### 指定下载某些文件",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --model 'Qwen/Qwen2-7b' --include '*.safetensors'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "### 过滤指定文件",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --model 'Qwen/Qwen2-7b' --exclude '*.safetensors'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "### 指定下载cache\\_dir",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --model 'Qwen/Qwen2-7b' --include '*.json' --cache_dir './cache_dir'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "模型文件将被下载到`'cache_dir/Qwen/Qwen2-7b'`。",
+        "",
+        "### 指定下载local\\_dir",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "modelscope download --model 'Qwen/Qwen2-7b' --include '*.json' --local_dir './local_dir'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "模型文件将被下载到`'./local_dir'`。",
+        "",
+        "如果`cache_dir`和`local_dir`参数同时被指定，`local_dir`优先级高，`cache_dir`将被忽略。",
+        "",
+        "## 下载私有模型需要登录",
+        "",
+        "### 通过login命令",
+        "",
+        "当下载私有模型时，您需要先登陆。通过 CLI 方式登陆的命令为`modelscope login`，详细使用说明如下：",
+        "",
+        "    usage: modelscope <command> [<args>] login [-h] --token TOKEN",
+        "    ",
+        "    options:",
+        "      -h, --help     show this help message and exit",
+        "      --token TOKEN  The Access Token for modelscope.",
+        "",
+        "      modelscope login --token YOUR_MODELSCOPE_ACCESS_TOKEN",
+        "",
+        "您可以在 [我的访问令牌](https://modelscope.cn/my/myaccesstoken) 页面获取**访问令牌**。",
+        "",
+        "# 使用 ModelScope SDK 下载",
+        "",
+        "## 下载整个模型仓库",
+        "",
+        "您可以使用`snapshot_download`下载整个模型仓库，示例如下：",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "from modelscope.hub.snapshot_download import snapshot_download",
+        "",
+        "model_dir = snapshot_download('iic/nlp_xlmr_named-entity-recognition_viet-ecommerce-title')"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "**参数说明**",
+        "",
+        "|  **字段名**          |  **必填**  |  **类型**   | **描述**                 |",
+        "| ------------------- |  :---------: | :----------: |------------------------|",
+        "|  model_id           |  是        |  str       | 模型ID                   |",
+        "|  revision         |  否        |  str       | 模型的Git版本，分支名或tag |",
+        "|  cache_dir            |  否        |  str,Path | 指定模型本次下载缓存目录，给定后下载的具体模型文件将会被存储在cache_dir/model_id/THE_MODEL_FILES |",
+        "|  allow_patterns       |  否        |  str,List       | 指定要下载的文件模式，如文件名或文件扩展名 |",
+        "|  ignore_patterns       |  否        |  str,List       | 指定要忽略下载的文件模式，如文件名或文件扩展名|",
+        "|  local_dir       |  否        |  str       | 指定模型的下载存放目录，给定后本次下载的模型文件将会被存储在local_dir/THE_MODEL_FILES|",
+        "",
+        "如果`cache_dir`和`local_dir`参数同时被指定，local_dir优先级高，cache_dir将被忽略；更多参数使用说明可以参见开源代码的接口文档。如需指定下载或过滤下载某种/某类文件模式，可以使用 `allow_patterns`或`ignore_patterns`参数，示例如下：",
+        "",
+        "- **指定下载某些文件**",
+        "",
+        "以指定下载`Qwen/QwQ-32B-GGUF`中`q4_k_m`量化版本到`path/to/local/dir`目录下为例。",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "from modelscope.hub.snapshot_download import snapshot_download",
+        "",
+        "model_dir = snapshot_download('Qwen/QwQ-32B-GGUF',allow_patterns='qwq-32b-q4_k_m.gguf',local_dir='path/to/local/dir')"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "- **过滤指定文件**",
+        "",
+        "以将`deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B`仓库除`figures`子目录外的所有文件下载到指定的`path/to/local/dir`目录为例。",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "from modelscope.hub.snapshot_download import snapshot_download",
+        "",
+        "model_dir = snapshot_download('deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B',ignore_patterns='figures/',local_dir='path/to/local/dir')"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "## 下载模型指定文件",
+        "",
+        "您也可以使用`model_file_download`下载模型指定文件。示例如下：",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "from modelscope.hub.file_download import model_file_download",
+        "",
+        "model_dir = model_file_download(model_id='Qwen/QwQ-32B-GGUF',file_path='qwq-32b-q4_k_m.gguf')"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "**参数说明**",
+        "",
+        "|  **字段名**          |  **必填**  |  **类型**   | **描述**                 |",
+        "| ------------------- |  :---------: | :----------: |------------------------|",
+        "|  model_id           |  是        |  str       | 模型ID                   |",
+        "|  file_path       |  是        |  str      | 待下载文件在远程模型仓库的相对路径|",
+        "|  revision         |  否        |  str       | 模型的Git版本，分支名或tag |",
+        "|  cache_dir            |  否        |  str,Path | 指定模型本次下载缓存目录，给定后下载的具体模型文件将会被存储在cache_dir/models/model_id/THE_MODEL_FILES |",
+        "|  local_dir       |  否        |  str       | 指定模型的下载存放目录，给定后本次下载的模型文件将会被存储在local_dir/THE_MODEL_FILES|",
+        "",
+        "如果`cache_dir`和`local_dir`参数同时被指定，`local_dir`优先级高，`cache_dir`将被忽略；更多参数使用说明可以参见开源代码的接口文档。",
+        "",
+        "## 下载私有模型  ",
+        "",
+        "当下载非公开模型及申请制模型时，您需要先登陆，然后可参考前述文档步骤下载模型。此处以通过 snapshot_download 下载非公开模型为例：",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "from modelscope import HubApi",
+        "from modelscope import snapshot_download",
+        "",
+        "# login to ModelScope",
+        "api=HubApi()",
+        "api.login('YOUR_MODELSCOPE_ACCESS_TOKEN')",
+        "",
+        "# download your model, the model_path is downloaded model path.",
+        "model_path =snapshot_download(model_id='the_model_id')"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "ModelScope SDK提供了多种登陆方式，您也可以通过命令行等其他方式完成登陆。",
+        "",
+        "## 通过加载模型触发下载 ",
+        "",
+        "除了直接下载模型文件外，当使用ModelScope SDK加载模型时，也会自动触发模型下载。如果模型和ModelScope SDK绑定，则只需要几行代码即可加载模型，同时 ModelScope 还支持通过 AutoModel 等接口来加载模型。此处以使用 `Model` 加载模型为例：",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "from modelscope.models import Model",
+        "model = Model.from_pretrained('iic/nlp_xlmr_named-entity-recognition_viet-ecommerce-title', revision='v1.0.1')",
+        "# revision为可选参数，不指定版本会取模型默认版本，默认版本，默认版本为ModelScope library发布前最后一个版本",
+        "",
+        "# 如何得到发布时间",
+        "import modelscope",
+        "print(modelscope.version.__release_datetime__)"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "# 使用 GIT 下载模型",
+        "",
+        "ModelScope服务端的模型都是通过Git存储，所以也可以在安装Git LFS后，通过git clone的方式在本地下载模型。",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "# 公开模型下载",
+        "git lfs install",
+        "git clone https://www.modelscope.cn/<owner_name>/<model-name>.git",
+        "# 例如: git clone https://www.modelscope.cn/iic/ofa_image-caption_coco_large_en.git",
+        "",
+        "# 私有模型下载，前提是您有响应模型权限 方法1",
+        "git lfs install",
+        "git clone http://oauth2:your_access_token@www.modelscope.cn/<owner_name>/<model-name>.git",
+        "# 方法2",
+        "git clone http://your_user_name@www.modelscope.cn/<owner_name>/<model-name>.git",
+        "# Password for 'http://your_user_name@modelscope.cn':",
+        "# input access token"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "如果**希望跳过LFS大文件的下载**，可以在git clone命令前添加`GIT_LFS_SKIP_SMUDGE=1`，来只获取`LFS`指针，而不下载实际的大文件：",
+        "",
+        "",
+        ""
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "GIT_LFS_SKIP_SMUDGE=1 https://www.modelscope.cn/<namespace>/<model-name>.git"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "",
+        "",
+        "",
+        "# 如何获取访问令牌",
+        "",
+        "用您的账号登录<https://www.modelscope.cn> ，在个人中心->访问令牌，新建并拷贝访问令牌.",
+        "<!-- ![image.png](https://resouces.modelscope.cn/document/docdata/2026-5-14_11-1-CN/dist/models/download/_resources/1661399339161-32fe4a95-0ad0-47e0-a360-b2522762022d.png) -->",
+        "![image.png](https://resouces.modelscope.cn/document/docdata/2026-5-14_11-1-CN/dist/models/download/_resources/access_token.jpeg)",
+        "",
+        "",
+        "",
+        ""
+      ]
+    }
+  ],
+  "metadata": {
+    "kernelspec": {
+      "display_name": "Python 3",
+      "language": "python",
+      "name": "python3"
+    }
+  },
+  "nbformat": 4,
+  "nbformat_minor": 2
+}