torch-rechub 0.0.4__tar.gz → 0.0.5__tar.gz

{torch_rechub-0.0.4 → torch_rechub-0.0.5}/.github/release.yml

@@ -8,6 +8,7 @@ changelog:
       - ignore-for-release
       - duplicate
       - invalid
+      - wontfix
     authors:
       - dependabot
       - dependabot[bot]
@@ -16,30 +17,22 @@ changelog:
     - title: "✨ 新特性 / Features"
       labels:
         - enhancement
-        - feature
-        - feat
 
     - title: "🐛 Bug 修复 / Bug Fixes"
       labels:
         - bug
-        - fix
-        - bugfix
 
     - title: "⚡ 性能优化 / Performance"
       labels:
         - performance
-        - perf
 
     - title: "📝 文档更新 / Documentation"
       labels:
         - documentation
-        - docs
 
-    - title: "🔧 维护更新 / Maintenance"
+    - title: "🔧 模型更新 / Models"
       labels:
-        - maintenance
-        - chore
-        - refactor
+        - model
 
     - title: "📦 依赖更新 / Dependencies"
       labels:
@@ -47,5 +40,4 @@ changelog:
 
     - title: "🔄 其他变更 / Other Changes"
       labels:
-        - "*"
-
+        - "*"

{torch_rechub-0.0.4 → torch_rechub-0.0.5}/.github/workflows/ci.yml

@@ -1,8 +1,9 @@
 # ===================================================================
 # CI/CD 流程配置 - 代码质量检查、测试、构建、发布
 # ===================================================================
-# 这个workflow在代码文件变更时触发，运行完整的CI/CD流程
-# 排除docs目录和markdown文件的变更
+# 触发条件：
+# - push/pull_request: 运行完整 CI 检查（lint, test, security, build）
+# - release: 仅运行发布流程（跳过已执行的检查）
 
 name: CI/CD Pipeline
 
@@ -37,11 +38,13 @@ env:
 
 jobs:
   # ===================================================================
-  # 代码质量检查
+  # 代码质量检查 (仅在 push/PR 时运行，release 时跳过)
   # ===================================================================
   lint:
     name: Code Quality Checks
     runs-on: ubuntu-latest
+    # 跳过 release 事件，因为代码已在合并时检查过
+    if: github.event_name != 'release'
     
     steps:
       - name: Checkout code
@@ -88,11 +91,13 @@ jobs:
 
   # ===================================================================
   # 完整测试 (Python 3.9) - 运行所有测试和覆盖率报告
+  # (仅在 push/PR 时运行，release 时跳过)
   # ===================================================================
   test:
     name: Full Test Suite (Python 3.9)
     runs-on: ${{ matrix.os }}
     needs: lint
+    if: github.event_name != 'release'
 
     strategy:
       fail-fast: false
@@ -152,11 +157,13 @@ jobs:
 
   # ===================================================================
   # 依赖兼容性验证 (Python 3.10+) - 仅验证依赖安装成功
+  # (仅在 push/PR 时运行，release 时跳过)
   # ===================================================================
   compatibility:
     name: Dependency Check (Python ${{ matrix.python-version }})
     runs-on: ubuntu-latest
     needs: lint
+    if: github.event_name != 'release'
 
     strategy:
       fail-fast: false
@@ -186,12 +193,13 @@ jobs:
           python -c "import onnx; import onnxruntime; print('ONNX dependencies OK')"
 
   # ===================================================================
-  # 安全检查
+  # 安全检查 (仅在 push/PR 时运行，release 时跳过)
   # ===================================================================
   security:
     name: Security Scan
     runs-on: ubuntu-latest
     needs: lint
+    if: github.event_name != 'release'
 
     steps:
       - name: Checkout code
@@ -220,12 +228,13 @@ jobs:
           path: bandit-report.json
 
   # ===================================================================
-  # 构建检查
+  # 构建检查 (仅在 push/PR 时运行，release 时跳过)
   # ===================================================================
   build:
     name: Build Package
     runs-on: ubuntu-latest
     needs: [test, compatibility, security]
+    if: github.event_name != 'release'
 
     steps:
       - name: Checkout code
@@ -256,18 +265,23 @@ jobs:
           path: dist/
 
   # ===================================================================
-  # 自动发布到PyPI (使用 uv)
-  # 功能：从 GitHub Release 自动同步版本号、更新 CHANGELOG、发布到 PyPI
+  # 自动发布到 PyPI 和 GitHub Release (使用 uv)
+  # 功能：
+  # - 从 GitHub Release 自动同步版本号
+  # - 更新 CHANGELOG.md
+  # - 构建并发布到 PyPI
+  # - 上传构建产物到 GitHub Release 页面
+  # 注意：此 job 仅在 release 事件时运行，不依赖其他 job（代码已在合并时检查过）
   # ===================================================================
   publish:
-    name: Publish to PyPI
+    name: Publish to PyPI & GitHub Release
     runs-on: ubuntu-latest
-    needs: build
+    # 不再依赖 build job，直接运行（代码质量已在 PR 合并时验证）
     if: github.event_name == 'release' && github.event.action == 'published'
     environment: pypi
     permissions:
       id-token: write   # Required for trusted publishing
-      contents: write   # Required for pushing changes back to repo
+      contents: write   # Required for pushing changes and uploading release assets
 
     steps:
       - name: Checkout code
@@ -342,14 +356,28 @@ jobs:
         run: uv python install ${{ env.PYTHON_VERSION }}
 
       - name: Build package with uv
+        id: build
         run: |
           uv build
           echo "✅ Package built successfully"
           ls -la dist/
+          # 输出构建产物文件名供后续步骤使用
+          echo "WHEEL_FILE=$(ls dist/*.whl)" >> $GITHUB_OUTPUT
+          echo "SDIST_FILE=$(ls dist/*.tar.gz)" >> $GITHUB_OUTPUT
 
       - name: Publish to PyPI
         env:
           UV_PUBLISH_TOKEN: ${{ secrets.PYPI_API_TOKEN }}
         run: |
           uv publish
-          echo "🚀 Published to PyPI successfully!"
+          echo "🚀 Published to PyPI successfully!"
+
+      - name: Upload release assets to GitHub Release
+        uses: softprops/action-gh-release@v2
+        with:
+          files: |
+            dist/*.whl
+            dist/*.tar.gz
+          fail_on_unmatched_files: true
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

{torch_rechub-0.0.4 → torch_rechub-0.0.5}/CHANGELOG.md

@@ -7,6 +7,21 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ---
 
+## [0.0.5] - 2025-12-05
+
+<!-- Release notes generated using configuration in .github/release.yml at main -->
+
+## What's Changed
+### ✨ 新特性 / Features
+* Add torchview to Support Model Visualization && Update CI/CD and release workflows by @1985312383 in https://github.com/datawhalechina/torch-rechub/pull/141
+
+
+**Full Changelog**: https://github.com/datawhalechina/torch-rechub/compare/v0.0.4...v0.0.5
+
+---
+
+
+
 ## [0.0.4] - 2025-12-04
 
 <!-- Release notes generated using configuration in .github/release.yml at main -->

{torch_rechub-0.0.4 → torch_rechub-0.0.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: torch-rechub
-Version: 0.0.4
+Version: 0.0.5
 Summary: A Pytorch Toolbox for Recommendation Models, Easy-to-use and Easy-to-extend.
 Project-URL: Homepage, https://github.com/datawhalechina/torch-rechub
 Project-URL: Documentation, https://www.torch-rechub.com
@@ -41,6 +41,9 @@ Requires-Dist: yapf==0.43.0; extra == 'dev'
 Provides-Extra: onnx
 Requires-Dist: onnx>=1.12.0; extra == 'onnx'
 Requires-Dist: onnxruntime>=1.12.0; extra == 'onnx'
+Provides-Extra: visualization
+Requires-Dist: graphviz>=0.20; extra == 'visualization'
+Requires-Dist: torchview>=0.2.6; extra == 'visualization'
 Description-Content-Type: text/markdown
 
 # 🔥 Torch-RecHub - 轻量、高效、易用的 PyTorch 推荐系统框架
@@ -69,13 +72,13 @@ Description-Content-Type: text/markdown
 
 ## 🎯 为什么选择 Torch-RecHub？
 
-| 特性 | Torch-RecHub | 其他框架 |
-|------|-------------|---------|
-| 代码行数 | **10行** 完成训练+评估+部署 | 100+ 行 |
-| 模型覆盖 | **30+** 主流模型 | 有限 |
-| 生成式推荐 | ✅ HSTU/HLLM (Meta 2024) | ❌ |
-| ONNX 一键导出 | ✅ 内置支持 | 需手动适配 |
-| 学习曲线 | 极低 | 陡峭 |
+| 特性          | Torch-RecHub                | 其他框架   |
+| ------------- | --------------------------- | ---------- |
+| 代码行数      | **10行** 完成训练+评估+部署 | 100+ 行    |
+| 模型覆盖      | **30+** 主流模型            | 有限       |
+| 生成式推荐    | ✅ HSTU/HLLM (Meta 2024)     | ❌          |
+| ONNX 一键导出 | ✅ 内置支持                  | 需手动适配 |
+| 学习曲线      | 极低                        | 陡峭       |
 
 ## ✨ 特性
 
@@ -205,52 +208,52 @@ torch-rechub/             # 根目录
 
 ### 排序模型 (Ranking Models) - 13个
 
-| 模型 | 论文 | 简介 |
-|------|------|------|
-| **DeepFM** | [IJCAI 2017](https://arxiv.org/abs/1703.04247) | FM + Deep 联合训练 |
-| **Wide&Deep** | [DLRS 2016](https://arxiv.org/abs/1606.07792) | 记忆 + 泛化能力结合 |
-| **DCN** | [KDD 2017](https://arxiv.org/abs/1708.05123) | 显式特征交叉网络 |
-| **DCN-v2** | [WWW 2021](https://arxiv.org/abs/2008.13535) | 增强版交叉网络 |
-| **DIN** | [KDD 2018](https://arxiv.org/abs/1706.06978) | 注意力机制捕捉用户兴趣 |
-| **DIEN** | [AAAI 2019](https://arxiv.org/abs/1809.03672) | 兴趣演化建模 |
-| **BST** | [DLP-KDD 2019](https://arxiv.org/abs/1905.06874) | Transformer 序列建模 |
-| **AFM** | [IJCAI 2017](https://arxiv.org/abs/1708.04617) | 注意力因子分解机 |
-| **AutoInt** | [CIKM 2019](https://arxiv.org/abs/1810.11921) | 自动特征交互学习 |
-| **FiBiNET** | [RecSys 2019](https://arxiv.org/abs/1905.09433) | 特征重要性 + 双线性交互 |
-| **DeepFFM** | [RecSys 2019](https://arxiv.org/abs/1611.00144) | 场感知因子分解机 |
-| **EDCN** | [KDD 2021](https://arxiv.org/abs/2106.03032) | 增强型交叉网络 |
+| 模型          | 论文                                             | 简介                    |
+| ------------- | ------------------------------------------------ | ----------------------- |
+| **DeepFM**    | [IJCAI 2017](https://arxiv.org/abs/1703.04247)   | FM + Deep 联合训练      |
+| **Wide&Deep** | [DLRS 2016](https://arxiv.org/abs/1606.07792)    | 记忆 + 泛化能力结合     |
+| **DCN**       | [KDD 2017](https://arxiv.org/abs/1708.05123)     | 显式特征交叉网络        |
+| **DCN-v2**    | [WWW 2021](https://arxiv.org/abs/2008.13535)     | 增强版交叉网络          |
+| **DIN**       | [KDD 2018](https://arxiv.org/abs/1706.06978)     | 注意力机制捕捉用户兴趣  |
+| **DIEN**      | [AAAI 2019](https://arxiv.org/abs/1809.03672)    | 兴趣演化建模            |
+| **BST**       | [DLP-KDD 2019](https://arxiv.org/abs/1905.06874) | Transformer 序列建模    |
+| **AFM**       | [IJCAI 2017](https://arxiv.org/abs/1708.04617)   | 注意力因子分解机        |
+| **AutoInt**   | [CIKM 2019](https://arxiv.org/abs/1810.11921)    | 自动特征交互学习        |
+| **FiBiNET**   | [RecSys 2019](https://arxiv.org/abs/1905.09433)  | 特征重要性 + 双线性交互 |
+| **DeepFFM**   | [RecSys 2019](https://arxiv.org/abs/1611.00144)  | 场感知因子分解机        |
+| **EDCN**      | [KDD 2021](https://arxiv.org/abs/2106.03032)     | 增强型交叉网络          |
 
 ### 召回模型 (Matching Models) - 12个
 
-| 模型 | 论文 | 简介 |
-|------|------|------|
-| **DSSM** | [CIKM 2013](https://posenhuang.github.io/papers/cikm2013_DSSM_fullversion.pdf) | 经典双塔召回模型 |
-| **YoutubeDNN** | [RecSys 2016](https://dl.acm.org/doi/10.1145/2959100.2959190) | YouTube 深度召回 |
-| **YoutubeSBC** | [RecSys 2019](https://dl.acm.org/doi/10.1145/3298689.3346997) | 采样偏差校正版本 |
-| **MIND** | [CIKM 2019](https://arxiv.org/abs/1904.08030) | 多兴趣动态路由 |
-| **SINE** | [WSDM 2021](https://arxiv.org/abs/2103.06920) | 稀疏兴趣网络 |
-| **GRU4Rec** | [ICLR 2016](https://arxiv.org/abs/1511.06939) | GRU 序列推荐 |
-| **SASRec** | [ICDM 2018](https://arxiv.org/abs/1808.09781) | 自注意力序列推荐 |
-| **NARM** | [CIKM 2017](https://arxiv.org/abs/1711.04725) | 神经注意力会话推荐 |
-| **STAMP** | [KDD 2018](https://dl.acm.org/doi/10.1145/3219819.3219895) | 短期注意力记忆优先 |
-| **ComiRec** | [KDD 2020](https://arxiv.org/abs/2005.09347) | 可控多兴趣推荐 |
+| 模型           | 论文                                                                           | 简介               |
+| -------------- | ------------------------------------------------------------------------------ | ------------------ |
+| **DSSM**       | [CIKM 2013](https://posenhuang.github.io/papers/cikm2013_DSSM_fullversion.pdf) | 经典双塔召回模型   |
+| **YoutubeDNN** | [RecSys 2016](https://dl.acm.org/doi/10.1145/2959100.2959190)                  | YouTube 深度召回   |
+| **YoutubeSBC** | [RecSys 2019](https://dl.acm.org/doi/10.1145/3298689.3346997)                  | 采样偏差校正版本   |
+| **MIND**       | [CIKM 2019](https://arxiv.org/abs/1904.08030)                                  | 多兴趣动态路由     |
+| **SINE**       | [WSDM 2021](https://arxiv.org/abs/2103.06920)                                  | 稀疏兴趣网络       |
+| **GRU4Rec**    | [ICLR 2016](https://arxiv.org/abs/1511.06939)                                  | GRU 序列推荐       |
+| **SASRec**     | [ICDM 2018](https://arxiv.org/abs/1808.09781)                                  | 自注意力序列推荐   |
+| **NARM**       | [CIKM 2017](https://arxiv.org/abs/1711.04725)                                  | 神经注意力会话推荐 |
+| **STAMP**      | [KDD 2018](https://dl.acm.org/doi/10.1145/3219819.3219895)                     | 短期注意力记忆优先 |
+| **ComiRec**    | [KDD 2020](https://arxiv.org/abs/2005.09347)                                   | 可控多兴趣推荐     |
 
 ### 多任务模型 (Multi-Task Models) - 5个
 
-| 模型 | 论文 | 简介 |
-|------|------|------|
-| **ESMM** | [SIGIR 2018](https://arxiv.org/abs/1804.07931) | 全空间多任务建模 |
-| **MMoE** | [KDD 2018](https://dl.acm.org/doi/10.1145/3219819.3220007) | 多门控专家混合 |
-| **PLE** | [RecSys 2020](https://dl.acm.org/doi/10.1145/3383313.3412236) | 渐进式分层提取 |
-| **AITM** | [KDD 2021](https://arxiv.org/abs/2105.08489) | 自适应信息迁移 |
-| **SharedBottom** | - | 经典多任务共享底层 |
+| 模型             | 论文                                                          | 简介               |
+| ---------------- | ------------------------------------------------------------- | ------------------ |
+| **ESMM**         | [SIGIR 2018](https://arxiv.org/abs/1804.07931)                | 全空间多任务建模   |
+| **MMoE**         | [KDD 2018](https://dl.acm.org/doi/10.1145/3219819.3220007)    | 多门控专家混合     |
+| **PLE**          | [RecSys 2020](https://dl.acm.org/doi/10.1145/3383313.3412236) | 渐进式分层提取     |
+| **AITM**         | [KDD 2021](https://arxiv.org/abs/2105.08489)                  | 自适应信息迁移     |
+| **SharedBottom** | -                                                             | 经典多任务共享底层 |
 
 ### 生成式推荐 (Generative Recommendation) - 2个
 
-| 模型 | 论文 | 简介 |
-|------|------|------|
+| 模型     | 论文                                          | 简介                                         |
+| -------- | --------------------------------------------- | -------------------------------------------- |
 | **HSTU** | [Meta 2024](https://arxiv.org/abs/2402.17152) | 层级序列转换单元，支撑 Meta 万亿参数推荐系统 |
-| **HLLM** | [2024](https://arxiv.org/abs/2409.12740) | 层级大语言模型推荐，融合 LLM 语义理解能力 |
+| **HLLM** | [2024](https://arxiv.org/abs/2409.12740)      | 层级大语言模型推荐，融合 LLM 语义理解能力    |
 
 ## 📊 支持的数据集
 
@@ -338,11 +341,19 @@ model = DSSM(user_features, item_features, temperature=0.02,
 match_trainer = MatchTrainer(model)
 match_trainer.fit(train_dl)
 match_trainer.export_onnx("dssm.onnx")
-# 双塔模型可分别导出用户塔和物品塔: 
+# 双塔模型可分别导出用户塔和物品塔:
 # match_trainer.export_onnx("user_tower.onnx", mode="user")
 # match_trainer.export_onnx("dssm_item.onnx", tower="item")
 ```
 
+### 模型可视化
+
+```python
+# 可视化模型架构（需要安装: pip install torch-rechub[visualization]）
+graph = ctr_trainer.visualization(depth=4)  # 生成计算图
+ctr_trainer.visualization(save_path="model.pdf", dpi=300)  # 保存为高清 PDF
+```
+
 ## 👨‍💻‍ 贡献者
 
 感谢所有的贡献者！

{torch_rechub-0.0.4 → torch_rechub-0.0.5}/README.md

@@ -24,13 +24,13 @@
 
 ## 🎯 为什么选择 Torch-RecHub？
 
-| 特性 | Torch-RecHub | 其他框架 |
-|------|-------------|---------|
-| 代码行数 | **10行** 完成训练+评估+部署 | 100+ 行 |
-| 模型覆盖 | **30+** 主流模型 | 有限 |
-| 生成式推荐 | ✅ HSTU/HLLM (Meta 2024) | ❌ |
-| ONNX 一键导出 | ✅ 内置支持 | 需手动适配 |
-| 学习曲线 | 极低 | 陡峭 |
+| 特性          | Torch-RecHub                | 其他框架   |
+| ------------- | --------------------------- | ---------- |
+| 代码行数      | **10行** 完成训练+评估+部署 | 100+ 行    |
+| 模型覆盖      | **30+** 主流模型            | 有限       |
+| 生成式推荐    | ✅ HSTU/HLLM (Meta 2024)     | ❌          |
+| ONNX 一键导出 | ✅ 内置支持                  | 需手动适配 |
+| 学习曲线      | 极低                        | 陡峭       |
 
 ## ✨ 特性
 
@@ -160,52 +160,52 @@ torch-rechub/             # 根目录
 
 ### 排序模型 (Ranking Models) - 13个
 
-| 模型 | 论文 | 简介 |
-|------|------|------|
-| **DeepFM** | [IJCAI 2017](https://arxiv.org/abs/1703.04247) | FM + Deep 联合训练 |
-| **Wide&Deep** | [DLRS 2016](https://arxiv.org/abs/1606.07792) | 记忆 + 泛化能力结合 |
-| **DCN** | [KDD 2017](https://arxiv.org/abs/1708.05123) | 显式特征交叉网络 |
-| **DCN-v2** | [WWW 2021](https://arxiv.org/abs/2008.13535) | 增强版交叉网络 |
-| **DIN** | [KDD 2018](https://arxiv.org/abs/1706.06978) | 注意力机制捕捉用户兴趣 |
-| **DIEN** | [AAAI 2019](https://arxiv.org/abs/1809.03672) | 兴趣演化建模 |
-| **BST** | [DLP-KDD 2019](https://arxiv.org/abs/1905.06874) | Transformer 序列建模 |
-| **AFM** | [IJCAI 2017](https://arxiv.org/abs/1708.04617) | 注意力因子分解机 |
-| **AutoInt** | [CIKM 2019](https://arxiv.org/abs/1810.11921) | 自动特征交互学习 |
-| **FiBiNET** | [RecSys 2019](https://arxiv.org/abs/1905.09433) | 特征重要性 + 双线性交互 |
-| **DeepFFM** | [RecSys 2019](https://arxiv.org/abs/1611.00144) | 场感知因子分解机 |
-| **EDCN** | [KDD 2021](https://arxiv.org/abs/2106.03032) | 增强型交叉网络 |
+| 模型          | 论文                                             | 简介                    |
+| ------------- | ------------------------------------------------ | ----------------------- |
+| **DeepFM**    | [IJCAI 2017](https://arxiv.org/abs/1703.04247)   | FM + Deep 联合训练      |
+| **Wide&Deep** | [DLRS 2016](https://arxiv.org/abs/1606.07792)    | 记忆 + 泛化能力结合     |
+| **DCN**       | [KDD 2017](https://arxiv.org/abs/1708.05123)     | 显式特征交叉网络        |
+| **DCN-v2**    | [WWW 2021](https://arxiv.org/abs/2008.13535)     | 增强版交叉网络          |
+| **DIN**       | [KDD 2018](https://arxiv.org/abs/1706.06978)     | 注意力机制捕捉用户兴趣  |
+| **DIEN**      | [AAAI 2019](https://arxiv.org/abs/1809.03672)    | 兴趣演化建模            |
+| **BST**       | [DLP-KDD 2019](https://arxiv.org/abs/1905.06874) | Transformer 序列建模    |
+| **AFM**       | [IJCAI 2017](https://arxiv.org/abs/1708.04617)   | 注意力因子分解机        |
+| **AutoInt**   | [CIKM 2019](https://arxiv.org/abs/1810.11921)    | 自动特征交互学习        |
+| **FiBiNET**   | [RecSys 2019](https://arxiv.org/abs/1905.09433)  | 特征重要性 + 双线性交互 |
+| **DeepFFM**   | [RecSys 2019](https://arxiv.org/abs/1611.00144)  | 场感知因子分解机        |
+| **EDCN**      | [KDD 2021](https://arxiv.org/abs/2106.03032)     | 增强型交叉网络          |
 
 ### 召回模型 (Matching Models) - 12个
 
-| 模型 | 论文 | 简介 |
-|------|------|------|
-| **DSSM** | [CIKM 2013](https://posenhuang.github.io/papers/cikm2013_DSSM_fullversion.pdf) | 经典双塔召回模型 |
-| **YoutubeDNN** | [RecSys 2016](https://dl.acm.org/doi/10.1145/2959100.2959190) | YouTube 深度召回 |
-| **YoutubeSBC** | [RecSys 2019](https://dl.acm.org/doi/10.1145/3298689.3346997) | 采样偏差校正版本 |
-| **MIND** | [CIKM 2019](https://arxiv.org/abs/1904.08030) | 多兴趣动态路由 |
-| **SINE** | [WSDM 2021](https://arxiv.org/abs/2103.06920) | 稀疏兴趣网络 |
-| **GRU4Rec** | [ICLR 2016](https://arxiv.org/abs/1511.06939) | GRU 序列推荐 |
-| **SASRec** | [ICDM 2018](https://arxiv.org/abs/1808.09781) | 自注意力序列推荐 |
-| **NARM** | [CIKM 2017](https://arxiv.org/abs/1711.04725) | 神经注意力会话推荐 |
-| **STAMP** | [KDD 2018](https://dl.acm.org/doi/10.1145/3219819.3219895) | 短期注意力记忆优先 |
-| **ComiRec** | [KDD 2020](https://arxiv.org/abs/2005.09347) | 可控多兴趣推荐 |
+| 模型           | 论文                                                                           | 简介               |
+| -------------- | ------------------------------------------------------------------------------ | ------------------ |
+| **DSSM**       | [CIKM 2013](https://posenhuang.github.io/papers/cikm2013_DSSM_fullversion.pdf) | 经典双塔召回模型   |
+| **YoutubeDNN** | [RecSys 2016](https://dl.acm.org/doi/10.1145/2959100.2959190)                  | YouTube 深度召回   |
+| **YoutubeSBC** | [RecSys 2019](https://dl.acm.org/doi/10.1145/3298689.3346997)                  | 采样偏差校正版本   |
+| **MIND**       | [CIKM 2019](https://arxiv.org/abs/1904.08030)                                  | 多兴趣动态路由     |
+| **SINE**       | [WSDM 2021](https://arxiv.org/abs/2103.06920)                                  | 稀疏兴趣网络       |
+| **GRU4Rec**    | [ICLR 2016](https://arxiv.org/abs/1511.06939)                                  | GRU 序列推荐       |
+| **SASRec**     | [ICDM 2018](https://arxiv.org/abs/1808.09781)                                  | 自注意力序列推荐   |
+| **NARM**       | [CIKM 2017](https://arxiv.org/abs/1711.04725)                                  | 神经注意力会话推荐 |
+| **STAMP**      | [KDD 2018](https://dl.acm.org/doi/10.1145/3219819.3219895)                     | 短期注意力记忆优先 |
+| **ComiRec**    | [KDD 2020](https://arxiv.org/abs/2005.09347)                                   | 可控多兴趣推荐     |
 
 ### 多任务模型 (Multi-Task Models) - 5个
 
-| 模型 | 论文 | 简介 |
-|------|------|------|
-| **ESMM** | [SIGIR 2018](https://arxiv.org/abs/1804.07931) | 全空间多任务建模 |
-| **MMoE** | [KDD 2018](https://dl.acm.org/doi/10.1145/3219819.3220007) | 多门控专家混合 |
-| **PLE** | [RecSys 2020](https://dl.acm.org/doi/10.1145/3383313.3412236) | 渐进式分层提取 |
-| **AITM** | [KDD 2021](https://arxiv.org/abs/2105.08489) | 自适应信息迁移 |
-| **SharedBottom** | - | 经典多任务共享底层 |
+| 模型             | 论文                                                          | 简介               |
+| ---------------- | ------------------------------------------------------------- | ------------------ |
+| **ESMM**         | [SIGIR 2018](https://arxiv.org/abs/1804.07931)                | 全空间多任务建模   |
+| **MMoE**         | [KDD 2018](https://dl.acm.org/doi/10.1145/3219819.3220007)    | 多门控专家混合     |
+| **PLE**          | [RecSys 2020](https://dl.acm.org/doi/10.1145/3383313.3412236) | 渐进式分层提取     |
+| **AITM**         | [KDD 2021](https://arxiv.org/abs/2105.08489)                  | 自适应信息迁移     |
+| **SharedBottom** | -                                                             | 经典多任务共享底层 |
 
 ### 生成式推荐 (Generative Recommendation) - 2个
 
-| 模型 | 论文 | 简介 |
-|------|------|------|
+| 模型     | 论文                                          | 简介                                         |
+| -------- | --------------------------------------------- | -------------------------------------------- |
 | **HSTU** | [Meta 2024](https://arxiv.org/abs/2402.17152) | 层级序列转换单元，支撑 Meta 万亿参数推荐系统 |
-| **HLLM** | [2024](https://arxiv.org/abs/2409.12740) | 层级大语言模型推荐，融合 LLM 语义理解能力 |
+| **HLLM** | [2024](https://arxiv.org/abs/2409.12740)      | 层级大语言模型推荐，融合 LLM 语义理解能力    |
 
 ## 📊 支持的数据集
 
@@ -293,11 +293,19 @@ model = DSSM(user_features, item_features, temperature=0.02,
 match_trainer = MatchTrainer(model)
 match_trainer.fit(train_dl)
 match_trainer.export_onnx("dssm.onnx")
-# 双塔模型可分别导出用户塔和物品塔: 
+# 双塔模型可分别导出用户塔和物品塔:
 # match_trainer.export_onnx("user_tower.onnx", mode="user")
 # match_trainer.export_onnx("dssm_item.onnx", tower="item")
 ```
 
+### 模型可视化
+
+```python
+# 可视化模型架构（需要安装: pip install torch-rechub[visualization]）
+graph = ctr_trainer.visualization(depth=4)  # 生成计算图
+ctr_trainer.visualization(save_path="model.pdf", dpi=300)  # 保存为高清 PDF
+```
+
 ## 👨‍💻‍ 贡献者
 
 感谢所有的贡献者！

{torch_rechub-0.0.4 → torch_rechub-0.0.5}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "torch-rechub"
-version = "0.0.4"
+version = "0.0.5"
 description = "A Pytorch Toolbox for Recommendation Models, Easy-to-use and Easy-to-extend."
 readme = "README.md"
 license = { text = "MIT" }
@@ -40,6 +40,10 @@ onnx = [
     "onnx>=1.12.0",
     "onnxruntime>=1.12.0",
 ]
+visualization = [
+    "torchview>=0.2.6",
+    "graphviz>=0.20",
+]
 dev = [
     "pytest>=6.0",
     "pytest-cov>=2.0",

{torch_rechub-0.0.4 → torch_rechub-0.0.5}/torch_rechub/trainers/ctr_trainer.py

@@ -189,3 +189,100 @@ class CTRTrainer(object):
 
         exporter = ONNXExporter(model, device=export_device)
         return exporter.export(output_path=output_path, dummy_input=dummy_input, batch_size=batch_size, seq_length=seq_length, opset_version=opset_version, dynamic_batch=dynamic_batch, verbose=verbose)
+
+    def visualization(self, input_data=None, batch_size=2, seq_length=10, depth=3, show_shapes=True, expand_nested=True, save_path=None, graph_name="model", device=None, dpi=300, **kwargs):
+        """Visualize the model's computation graph.
+
+        This method generates a visual representation of the model architecture,
+        showing layer connections, tensor shapes, and nested module structures.
+        It automatically extracts feature information from the model.
+
+        Parameters
+        ----------
+        input_data : dict, optional
+            Example input dict {feature_name: tensor}.
+            If not provided, dummy inputs will be generated automatically.
+        batch_size : int, default=2
+            Batch size for auto-generated dummy input.
+        seq_length : int, default=10
+            Sequence length for SequenceFeature.
+        depth : int, default=3
+            Visualization depth, higher values show more detail.
+            Set to -1 to show all layers.
+        show_shapes : bool, default=True
+            Whether to display tensor shapes.
+        expand_nested : bool, default=True
+            Whether to expand nested modules.
+        save_path : str, optional
+            Path to save the graph image (.pdf, .svg, .png).
+            If None, displays in Jupyter or opens system viewer.
+        graph_name : str, default="model"
+            Name for the graph.
+        device : str, optional
+            Device for model execution. If None, defaults to 'cpu'.
+        dpi : int, default=300
+            Resolution in dots per inch for output image.
+            Higher values produce sharper images suitable for papers.
+        **kwargs : dict
+            Additional arguments passed to torchview.draw_graph().
+
+        Returns
+        -------
+        ComputationGraph
+            A torchview ComputationGraph object.
+
+        Raises
+        ------
+        ImportError
+            If torchview or graphviz is not installed.
+
+        Notes
+        -----
+        Default Display Behavior:
+            When `save_path` is None (default):
+            - In Jupyter/IPython: automatically displays the graph inline
+            - In Python script: opens the graph with system default viewer
+
+        Examples
+        --------
+        >>> trainer = CTRTrainer(model, ...)
+        >>> trainer.fit(train_dl, val_dl)
+        >>>
+        >>> # Auto-display in Jupyter (no save_path needed)
+        >>> trainer.visualization(depth=4)
+        >>>
+        >>> # Save to high-DPI PNG for papers
+        >>> trainer.visualization(save_path="model.png", dpi=300)
+        """
+        from ..utils.visualization import TORCHVIEW_AVAILABLE, visualize_model
+
+        if not TORCHVIEW_AVAILABLE:
+            raise ImportError(
+                "Visualization requires torchview. "
+                "Install with: pip install torch-rechub[visualization]\n"
+                "Also ensure graphviz is installed on your system:\n"
+                "  - Ubuntu/Debian: sudo apt-get install graphviz\n"
+                "  - macOS: brew install graphviz\n"
+                "  - Windows: choco install graphviz"
+            )
+
+        # Handle DataParallel wrapped model
+        model = self.model.module if hasattr(self.model, 'module') else self.model
+
+        # Use provided device or default to 'cpu'
+        viz_device = device if device is not None else 'cpu'
+
+        return visualize_model(
+            model,
+            input_data=input_data,
+            batch_size=batch_size,
+            seq_length=seq_length,
+            depth=depth,
+            show_shapes=show_shapes,
+            expand_nested=expand_nested,
+            save_path=save_path,
+            graph_name=graph_name,
+            device=viz_device,
+            dpi=dpi,
+            **kwargs
+        )