Setup - Embedl Hub

Create an Embedl Hub account

To get started with Embedl Hub, you’ll first need to create a free account. After you’ve signed up, we invite you to join our Slack community. Although joining the community isn’t required, we’d love to meet you and learn what excites you about efficient edge AI.

Install the Python library

With the Embedl Hub Python library, you can compile and profile models on real edge devices in the cloud. The library logs metrics and parameters, allowing you to analyze your results on the Embedl Hub website and to reproduce them later.

The library can be used via the embedl-hub CLI or programmatically through the Python API.

The recommended Python version is 3.11.

Base install (tracking only)

The base install covers the experiment tracking surface — CLI auth, init, log, and list-devices, plus the Python tracking API. It does not pull any backend execution toolchain, so it stays small:

pip install embedl-hub

Execution extras

To use the bundled compile, profile, and invoke components (Embedl’s ready-made wrappers for TFLite, ONNX Runtime, and TensorRT), install one or more toolchain extras:

Workflow	Install command
TFLite (local + device cloud + QAI Hub)	`pip install 'embedl-hub[tflite]'`
ONNX Runtime (embedl-ORT + QAI Hub)	`pip install 'embedl-hub[onnxruntime]'`
TensorRT (`trtexec` over SSH)	`pip install 'embedl-hub[tensorrt]'`
Everything	`pip install 'embedl-hub[all]'`

Linux aarch64 installs of [tflite] skip the onnx2tf-based local conversion path and the ai-edge-* quantization dependencies because the upstream TensorFlow package does not publish Linux aarch64 wheels. The QAI Hub TFLite provider still works on that platform.

Running a backend command without its extra installed prints a clear hint pointing at the right pip install command rather than a raw import traceback.

Configure an API key

The Embedl Hub Python library requires an API key for authentication. To start, create one under Personal API keys on your profile page.

Keep the key secret. Anyone with access to it will be able to read and write to your Embedl Hub projects.

We recommend configuring the API key using the embedl-hub CLI:

embedl-hub auth --api-key <your-key>

Your key will be stored in the plaintext configuration file at ~/.config/embedl-hub/config.yaml, and any existing key in the file will be overwritten.

For alternative ways to configure your API key, see the configuration guide.

Log your first run

After installing the library and configuring an API key, you can create a run directly from Python with the tracking client:

from pathlib import Path

from embedl_hub.tracking import Client

client = Client()

client.set_project("my-project")

with client.start_run("evaluate", name="Evaluate baseline"):

    client.log_param("model", "baseline")

    client.log_metric("accuracy", 0.923, step=1)

    report_path = Path("hub-example-report.txt")

    report_path.write_text("accuracy=0.923\n")

    client.log_artifact(report_path, name="report")

Once the script finishes, open the project in Embedl Hub to see the run and its contents.

(Optional) Set up a remote hardware cloud

Embedl Hub enables you to evaluate your models on real edge AI devices without needing physical access to the devices. Using the Embedl Hub device cloud requires no additional set up.

Follow the cloud-specific setup instructions to use remote devices from third-party device clouds:

Qualcomm AI Hub