PyTorch Recipes

译者：片刻小哥哥

项目地址：https://pytorch.apachecn.org/2.0/tutorials/recipes/recipes_index

原始地址：https://pytorch.org/tutorials/recipes/recipes_index.html

Recipes 是关于如何使用特定 PyTorch 功能的简短、可操作的示例，与我们的完整教程不同。

Basics

Example

在 PyTorch 中加载数据

了解如何使用 PyTorch 包为您的模型准备和加载通用数据集

:material-clock-fast:{ .lg .middle } Set up in 5 minutes

Install mkdocs-material with pip and get up and running in minutes

:octicons-arrow-right-24: Getting started
:fontawesome-brands-markdown:{ .lg .middle } It's just Markdown

Focus on your content and generate a responsive and searchable static site

:octicons-arrow-right-24: Reference
:material-format-font:{ .lg .middle } Made to measure

Change the colors, fonts, language, icons, logo and more with a few lines

:octicons-arrow-right-24: Customization
:material-scale-balance:{ .lg .middle } Open Source, MIT

Material for MkDocs is licensed under MIT and available on [GitHub]

:octicons-arrow-right-24: License

:material-brush-variant: Colors Change colors with an existing color palette or customize with CSS
:material-format-font: Fonts – Choose among 1,000 Google Fonts or load self-hosted fonts
:material-google-downasaur: Logo & Icons – Change the logo, use any of the 8,000+ icons, or add new ones
:material-cards-variant: Social Cards – Automatically create social media previews when sharing links

### 定义神经网络

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/defining-a-network.PNG" width=20% />

[了解如何使用 PyTorch 的 torch.nn 包为 MNIST 数据集创建和定义神经网络](https://pytorch.org/tutorials/recipes/recipes/defining_a_neural_network.html)

### PyTorch 中的 state_dict 是什么

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/what-is-a-state-dict.PNG" width=20% />

[了解如何使用 state_dict 对象和 Python 字典从 PyTorch 保存或加载模型](https://pytorch.org/tutorials/recipes/recipes/what_is_state_dict.html)

### 在 PyTorch 中保存和加载模型以进行推理

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/saving-and-loading-models-for-inference.PNG" width=20% />

[了解在 PyTorch 中保存和加载模型进行推理的两种方法 - 通过 state\_dict 和通过整个模型](https://pytorch.org/tutorials/recipes/recipes/saving_and_loading_models_for_inference.html)

### 在 PyTorch 中保存和加载常规检查点

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/saving-and-loading-general-checkpoint.PNG" width=20% />

[保存和加载用于推理或恢复训练的通用检查点模型有助于从上次停下的地方继续。在本节中，探索如何保存和加载多个检查点](https://pytorch.org/tutorials/recipes/recipes/saving_and_loading_a_general_checkpoint.html)

### 使用 PyTorch 在一个文件中保存和加载多个模型

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/saving-multiple-models.PNG" width=20% />

[在本节中，了解保存和加载多个模型如何有助于重用您之前训练过的模型](https://pytorch.org/tutorials/recipes/recipes/saving_multiple_models_in_one_file.html)


### 使用 PyTorch 中不同模型的参数热启动模型

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/warmstarting-models.PNG" width=20% />

[了解通过部分加载模型或加载部分模型来热启动训练过程如何帮助您的模型比从头开始训练更快地收敛](https://pytorch.org/tutorials/recipes/recipes/warmstarting_model_using_parameters_from_a_different_model.html)


### 在 PyTorch 中跨设备保存和加载模型

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/saving-and-loading-models-across-devices.PNG" width=20% />

[了解如何使用 PyTorch 相对简单地跨设备(CPU 和 GPU)保存和加载模型](https://pytorch.org/tutorials/recipes/recipes/save_load_across_devices.html)


### 在 PyTorch 中将梯度归零

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/zeroing-out-gradients.PNG" width=20% />

[了解何时应该将梯度归零以及这样做如何有助于提高模型的准确性](https://pytorch.org/tutorials/recipes/recipes/zeroing_out_gradients.html)


### PyTorch 基准测试

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/profiler.png" width=20% />

[了解如何使用 PyTorch 的基准测试模块来测量和比较代码的性能](https://pytorch.org/tutorials/recipes/recipes/benchmark.html)


### PyTorch 基准测试(快速入门)

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/profiler.png" width=20% />

[了解如何测量代码片段运行时间并收集指令](https://pytorch.org/tutorials/recipes/recipes/timer_quick_start.html)


### PyTorch Profiler

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/profiler.png" width=20% />

[了解如何使用 PyTorch 的分析器来测量操作员的时间和内存消耗](https://pytorch.org/tutorials/recipes/recipes/profiler_recipe.html)

### PyTorch Profiler 具有仪器和跟踪技术 API (ITT API) 支持

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/profiler.png" width=20% />

[了解如何将 PyTorch 的分析器与仪器和跟踪技术 API (ITT API) 结合使用，以可视化英特尔® VTune™ Profiler GUI 中的操作员标签](https://pytorch.org/tutorials/recipes/profile_with_itt.html)

### Torch 编译 IPEX 后端

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/profiler.png" width=20% />

[了解如何使用 torch.compile IPEX 后端](https://pytorch.org/tutorials/recipes/torch_compile_backend_ipex.html)

### 在 PyTorch 中推理形状

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/profiler.png" width=20% />

[了解如何使用元设备来推理模型中的形状](https://pytorch.org/tutorials/recipes/recipes/reasoning_about_shapes.html)

### 从检查点加载 nn.Module 的提示

<img src="https://pytorch.org/tutorials/_static/img/thumbnails/cropped/profiler.png" width=20% />

[了解从检查点加载 nn.Module 的技巧](https://pytorch.org/tutorials/recipes/recipes/module_load_state_dict_tips.html)

CaptumMobileModel Optimization