Deep Learning with PyTorch Tutorial

Quick Start explained how to run your first JupyterLab with HyperAI. Here we will specifically introduce how to clone a public tutorial using HyperAI's Jupyter workspace.

Deep Learning with PyTorch: A 60 Minute Blitz is PyTorch's official learning tutorial, which introduces some basic PyTorch syntax and concepts, and demonstrates the process of training an image classification model using the cifar10 dataset. It provides execution methods via Google Colab and also allows downloading Jupyter .ipynb files. OpenBayes also provides an organized and translated version that can be directly used in Jupyter workspaces under HyperAI.

Clone Public Tutorial

After logging into HyperAI, select "Public Resources" - "Public Tutorials" on the left side, then search for "PyTorch Official Tutorial - Deep Learning with PyTorch".

You can see that this tutorial uses PyTorch 1.9 and employs a vGPU type container. Click "Clone" in the upper right corner to clone this container as a new container.

As shown above, after cloning the container, you can independently select "Computing Resources" and "Runtime Environment".

:::info If the public tutorial also has other public models or datasets bound to it, they will all be bound by default when "cloning". :::

After waiting a moment, you can see a started container:

Click "Open Jupyter Workspace" to use the JupyterLab workspace.

:::caution Note If you see the following prompt, please click allow. This feature will support browser notifications when JupyterLab execution is complete.

:::

If you don't know how to use JupyterLab yet, you can refer to our Jupyter Workspace documentation, JupyterLab official documentation, or related Chinese translation materials.

Execute Tutorial

In the left directory of the opened Jupyter workspace, double-click the file Deep Learning with PyTorch.ipynb to open it.

Then you can browse the tutorial and learn chapter by chapter.

If you need to restart the tutorial, you can select "Kernel" - "Restart Kernel" from the top navigation bar.

If you want to execute all code in the ipynb file at once, you can select "Kernel" - "Restart Kernel and Run All Cells".

Download Files

After performing a series of editing tasks, you need to download the updated .ipynb file to your local machine. There are multiple ways to download.

Download Directly from Jupyter Workspace

For a running Jupyter workspace, you can right-click on the file you want to download and click download.

If you want to download an entire directory, you can right-click in the blank area of the left navigation and select "Download Current Folder as an Archive", which will automatically package and download the current directory:

Download from Working Directory Tab

If the Jupyter workspace has been closed, you can download through the "Output" tab on the page. You can select "Download Current Directory" on the right side.

You can also directly download individual files using the download button on the right side of the file list.

:::note The files in the "Output" tab of a running container are periodically synchronized from the container, so compared to the files in the container, the update time of the files here will be later than those in the container. It is recommended to download files from here after the container has finished executing to avoid obtaining incomplete file content. :::