How is pricing calculated?

On the new compute container page, you can view detailed configurations and prices for all compute types and recharge your balance.

What are the system version and external network bandwidth of HyperAI platform machines?

Our system is Ubuntu 18 LTS, which you can check in the container using lsb_release -a

Both uploading from local to HyperAI and downloading files from HyperAI to local computer have 300 Mbps bandwidth. Slow uploads are usually limited by insufficient user bandwidth. Installing dependencies within HyperAI containers (such as downloading models from Hugging Face) has 50 Mbps bandwidth.

What is the difference between "Workspace" and "Storage Space"?

• Workspace: The local disk of the compute container (e.g., RTX 5090 workspace is 50 GB). Users can upload code, files, datasets (not exceeding maximum capacity), etc. Uploaded content disappears after the container is closed. If you want to save content after closing the container, you need to save it in the specific directory /hyperai/home/, which will automatically sync to storage space after closing the container.
• Storage Space: Used to save user-uploaded datasets, models, container working directories, etc., which can be retained permanently. The platform currently provides 50 GB of permanently free storage space for each user. You can free up storage space by deleting execution records, datasets, models, etc. Containers will not stop running when storage space overflows, but it will affect starting new containers next time.

How to expand storage space?

This question needs to be discussed case by case:

1. Expand available storage space through data repositories

Since the "Workspace" capacity is currently fixed (e.g., 50GB workspace for RTX 5090), you can manage larger data through the following methods:

1. Create a data repository (dataset or model)
2. When creating a container, bind this data repository to /hyperai/input/input0 (or other input directories) in "read-write binding" mode
3. Write data directly to the bound directory, so the data will be stored directly in the data repository without occupying workspace

This method is particularly suitable for:

• Processing large datasets that exceed workspace capacity limits
• Sharing data between multiple containers
• Important data that needs to be preserved long-term

2. Expand permanent storage space

The platform provides each registered user with permanently valid 20 GB storage space for storing:

• User-uploaded datasets
• Model files
• Data in container working directories (/hyperai/home directory)

When the free 20GB storage space is insufficient, you need to expand it, otherwise you will not be able to start compute containers normally. The expansion method is as follows:

How to save virtual environments and avoid repeated installations?

HyperAI supports saving virtual environments and modifying Python scripts. Please refer to Runtime Environment (Image) Overview

How to save installed dependencies to avoid reinstalling them after every restart?

There are two ways to save your dependencies:

1. Use the pip install --user xxx command to install dependencies in the default conda environment. Dependencies installed this way will be saved in the user directory and will not be lost after restart.
2. Create a custom conda environment and place it in the user directory (/hyperai/home). For more information, see Dependency Management with Conda.

However, dependencies installed via apt cannot be saved and need to be reinstalled upon startup.

What happens to my running container if my balance is insufficient?

If your account balance is insufficient, running containers will be forcibly shut down. We recommend that you monitor your container status at all times and ensure your account has sufficient balance.

Why was my container deleted?

To prevent inactive project data from accumulating over time and affecting platform performance, HyperAI will permanently delete projects that have been inactive—without any task execution or container startup records—for one consecutive month. You can check the exact release date of a project by clicking the trash can icon in the upper-right corner of the container on the Model Training page. HyperAI will send email reminders 7 days, 3 days, and 1 day before the project is released. If you wish to retain a project, please regularly execute tasks and start the container, and ensure that there are no consecutive overdue payments.

How to check resource validity period?

You can check this in the console under Resource Usage - Resource Change Details

Why did my file upload fail?

File upload failures are usually caused by the following factors:

1. Unstable network - Uploading large files requires a stable network connection; any interruption may cause upload failure. When a computer enters sleep mode, the network connection is interrupted, causing upload failure.
2. Browser limitations - Uploading particularly large files through a browser may be subject to browser limitations
3. Insufficient storage space - If your storage space is full, files will not upload successfully

Solutions

1. Utilize resumable upload functionality

HyperAI supports resumable uploads via the web interface. Even if an upload fails, there's no need to worry too much—the system will remember the uploaded portion. As long as the file content and upload destination remain unchanged, the next upload will automatically resume from the breakpoint instead of starting from scratch, greatly saving your time and bandwidth.

2. Try alternative upload methods

   If upload fails through Jupyter workspace, you can try uploading via data warehouse. For details, see Data Upload Methods.

3. Check storage space

   Ensure your account has sufficient storage capacity. If storage space is insufficient, you can refer to the How to expand storage space section for solutions.