NetsPresso X Qualcomm AI Hub

How to sign up and get your Qualcomm AI Hub API Token

To use Qualcomm AI Hub features within NetsPresso, you need accounts for both services and an API token from Qualcomm AI Hub. Follow the steps below to get everything set up.

1. Create a NetsPresso Account

🔗 Sign up for NetsPresso

To access NetsPresso's features, you first need to create an account. Go to the link above and sign up with your email and password. After signing up, you can log in using the Python SDK like this:

from netspresso import NetsPresso

EMAIL = "<YOUR_EMAIL>"
PASSWORD = "<YOUR_PASSWORD>"

netspresso = NetsPresso(email=EMAIL, password=PASSWORD)

2. Sign Up for Qualcomm AI Hub

🔗 Sign up for Qualcomm AI Hub

Qualcomm AI Hub requires a separate registration. Visit the link above and create an account. You may need to verify your email before accessing the platform.

3. Where can I find my Qualcomm AI Hub API Token?

After signing in to Qualcomm AI Hub, you can find your API token on the Settings page:

1. Go to Qualcomm AI Hub and log in.
2. Click on your profile icon at the top right corner.
3. Select "Settings" from the dropdown menu.
4. In the “ACCOUNT” tab, under the “Your Account” section, your API Token will be displayed.
5. Copy the token string shown in the box.

This token is required to access Qualcomm AI Hub features via the Python SDK. Use it in your code like this:

from netspresso import NPQAI

QAI_HUB_API_TOKEN = "<YOUR_QAI_HUB_API_TOKEN>"
np_qai = NPQAI(api_token=QAI_HUB_API_TOKEN)

Where can I find the API documentation for Qualcomm AI Hub integration in NetsPresso?

To use Qualcomm AI Hub as an integrated part of your NetsPresso workflow, especially for conversion, quantization, and benchmarking, refer to the NPQAI module.

This module enables you to:

• Train and compress your model with NetsPresso’s Trainer and Compressor
• Then convert, quantize, and benchmark the model using Qualcomm AI Hub’s Converter, Quantizer, and Benchmarker

See the NPQAI API documentation to learn how to integrate and automate this workflow.

Can I Bring Your Own Model (BYOM) to NetsPresso and Qualcomm AI Hub?

NetsPresso

You can bring your own model for use with the Compressor.

• For PyTorch models, you need to convert them into either a GraphModule (via torch.fx) or ONNX format before uploading.
  See GraphModule conversion guide or ONNX conversion guide for details.

• Models compressed this way must be retrained using your own training code to ensure optimal performance.

• The compressed model is returned as a GraphModule, so retraining should be done in a way that supports this format, rather than using a regular PyTorch model directly.

Using your own model with the Trainer is not supported.

Qualcomm AI Hub

Bringing and using your own model is fully supported across all services.

Why am I seeing errors like “500”, “502”, or “504 Gateway Timeout”?

If you encounter messages like:

💥

Example error messages

• Internal server error occurred with status code 500
• Unexpected error occurred with status code 502
• 504 Gateway Timeout: The server did not receive a timely response with status code 504

These errors usually occur when the server is temporarily overloaded or experiencing high traffic.

⏳ Please try again after a short while.
In most cases, the issue resolves on its own without any additional action.

If the issue persists, please refer to the Contact Us section below to get in touch with us.

Contact Us

If you have questions beyond the FAQ or need further assistance, please don’t hesitate to email us—we’re happy to help! You can also reach out to NetsPresso directly, or contact us via email at [email protected] or by phone at +82 2-555-8659.