Notebooks and IDEs
Open a browser-based `Jupyter` or `VSCode` environment.
A web IDE is a browser-based `Jupyter` or `VSCode` environment running on zCLOUD compute. Open [**IDE**](https://console.zettabyte.com/task-ide/index) to see the page that lists your environments and lets you create new ones. Each row has a link to open the environment, plus these columns: `Name/ID`, `Status`, `Type`, `Resource Name`, `GPU`, `CPU`, `User`, `Creation Time`, and `Web IDE`.
Create a web IDE
Open [**IDE**](https://console.zettabyte.com/task-ide/index), select **Create IDE**, choose `Jupyter` or `VSCode` as the `Type`, and set up the environment. **Advanced Settings** turns the single form into a three-step wizard with extra options: `Basic Information`, `Instance Configuration` and an overview step.
Basic (single form)
- `Name`: Chinese characters, letters, numbers, and the symbols `- _ / ( )` are allowed.
- `Type`: `Jupyter` or `VSCode`.
- `Computing Config`: choose a `Resource Name` and set `GPU Count(Card)`, `CPU(Core)`, `Memory(GiB)`, and `Local Storage(GB)`. For a GPU-backed environment, select a launched resource (see GPU instances). The console note beside these fields reads "Setting GPU to 0 can create a free development machine".
Advanced Settings: Shared Memory
On the `Instance Configuration` step of the wizard, an optional `Shared Memory` field takes a value in GiB. The console shows the placeholder "Optional, defaults to 64 MiB".
Mount a `PVC` so your code and outputs persist beyond the life of the environment. See Storage and PVCs.
Free development machine
The console note on the create form describes a GPU Count of `0` as one way to create a free development machine. The number of free development machines is limited: a create request past the limit returns "Free development machine has reached the limit."
Dev environment vs. tasks
Package your code into an image for heavy work
The IDE is ideal for light iteration and testing. For heavy development, doing everything in the in-platform IDE is not realistic. The recommended pattern is to package your changes into a container image (ideally driven from CI or a version-controlled `Dockerfile`) and run it as a task. See Container images and How zCLOUD works.
You typically set up your environment (an IDE and/or an image) before creating a training task. A `PVC` lets your code and data persist between the two. See Run a training job.
Shell access
Native SSH into a task pod is not available today. Interactive shell access is through the console into the pod (and the zCLOUD CLI exec path). See zCLOUD CLI.
Next steps
- [Container images](https://docs.zettabyte.com/docs/zcloud/compute/container-images): package your code into an image and run it as a task, the recommended path beyond light iteration.
- [Run a training job](https://docs.zettabyte.com/docs/zcloud/tasks/run-a-training-job): run your script on GPUs as a task.