qai-hub-apps 0.34.0

CLI for browsing and downloading Qualcomm® AI Hub Apps.

qai-hub-apps

CLI for browsing and downloading Qualcomm® AI Hub Apps — sample apps for deploying ML models on Qualcomm® devices.

Installation

pip install qai-hub-apps

Quick Start

qai-hub-apps list                          # browse available apps
qai-hub-apps info whisper_windows_py       # inspect an app
qai-hub-apps fetch whisper_windows_py      # download source to current directory

Commands

list

List all available apps.

qai-hub-apps list

+----------------------------------------------------------------------+
|                        Qualcomm® AI Hub Apps                         |
+---------------------+------------------+---------------+-------------+
| ID                  | Name             | Domain        | Languages   |
+---------------------+------------------+---------------+-------------+
| whisper_windows_py  | Whisper Windows  | Audio         | Python      |
| stable_diffusion_py | Stable Diffusion | Generative AI | Python, C++ |
+---------------------+------------------+---------------+-------------+
Total: N apps

info

Show details for an app.

qai-hub-apps info <app_id>

+=============================================================================+
|                               Whisper Windows                               |
|  https://github.com/qualcomm/ai-hub-apps/tree/main/apps/whisper_windows_py  |
+=============================================================================+

+---------------------------+
|          Headline         |
+---------------------------+
| Speech to text on Windows |
+---------------------------+

+-----------------------------------+
|            Description            |
+-----------------------------------+
| Run Whisper on-device using ONNX. |
+-----------------------------------+

+--------------------------------------------+
|                  Metadata                  |
+-------------------+------------------------+
| ID                | whisper_windows_py     |
| Type              | windows                |
| Languages         | Python                 |
| Runtime           | onnx                   |
| Domain            | Audio                  |
| Use Case          | Speech Recognition     |
| Precision         | float                  |
| Models            | whisper_base           |
| Supported Devices | Snapdragon X Elite CRD |
| AI Hub Models     | 0.53.1                 |
+-------------------+------------------------+

fetch

Download and extract an app's source to a local directory.

# Download app source only
qai-hub-apps fetch <app_id>

# Download app source + a model to download for a specific chipset
qai-hub-apps fetch <app_id> --model <model_id> --chipset <chipset>

# Download app source + bundle a locally-exported model
qai-hub-apps fetch <app_id> --model <path/to/model>

Flag	Description
--output-dir PATH	Output directory (default: current directory)
--model MODEL_ID_OR_PATH	Model to bundle: a model ID to download (must be supported by the app), or a path to a locally-exported model (directory or .zip). Use --model-id/--model-path to be explicit
--model-id MODEL_ID	Model ID to download (must be supported by the app)
--model-path PATH	Path to a locally-exported model (directory or .zip)
--chipset CHIPSET	Target chipset for the model download. Only applies to a downloaded model (--model <id> or --model-id); not valid with a local model path
--device DEVICE	Target device for the model download. Only applies to a downloaded model (--model <id> or --model-id); not valid with a local model path

--model, --model-id, and --model-path are mutually exclusive.

On success, the path to the fetched app directory is printed.

Example — fetch app with model:

qai-hub-apps fetch stable_diffusion_windows_py --model stable_diffusion_v2_1 --chipset qualcomm-snapdragon-x-elite
# or
qai-hub-apps fetch stable_diffusion_windows_py --model stable_diffusion_v2_1 --device "Snapdragon X Elite CRD"

Logging

Control verbosity with a global flag or the QAI_HUB_APPS_LOG_LEVEL environment variable. The command-line flag takes precedence over the environment variable. The default level is info.

Flag	Description
--log-level {debug,info,error}	Set log verbosity
-v, --verbose	Shorthand for --log-level debug
-q, --quiet	Shorthand for --log-level error

qai-hub-apps -v list
QAI_HUB_APPS_LOG_LEVEL=debug qai-hub-apps list