Describe the bug
When trying to run the ai-hub-models/Qwen3-8B model using the geniex CLI on Windows 11 ARM64 (Snapdragon X Elite), the command fails immediately after downloading. The process prints DSP_INFO UNSUPPORTED_KEY: 49 and DSP_INFO UNSUPPORTED_KEY: 50 repeatedly and then exits to the terminal prompt without starting the inference process.
To Reproduce
Steps to reproduce the behavior:
- Open PowerShell / Windows Terminal
- Run command:
geniex infer ai-hub-models/Qwen3-8B
- Wait for the model download to complete (reaches 100%)
- See error / immediate return to prompt
Expected behavior
The model should initialize successfully on the NPU (Hexagon DSP) and start the interactive inference session, or alternatively, fall back gracefully to the GPU or CPU runtime if NPU features are not supported.
Screenshots
If applicable, add screenshots to help explain your problem.
Desktop (please complete the following information):
- OS: Windows 11 (ARM64)
- Browser: N/A (CLI Tool)
- Version: N/A
- Hardware: Snapdragon X Elite (Qualcomm Oryon CPU)
- Memory: 32.0 GB
Smartphone (please complete the following information):
Additional context
Here is the full terminal output:
PS C:\Users\Acer> geniex infer ai-hub-models/Qwen3-8B
DSP_INFO UNSUPPORTED_KEY: 49
DSP_INFO UNSUPPORTED_KEY: 50
DSP_INFO UNSUPPORTED_KEY: 49
DSP_INFO UNSUPPORTED_KEY: 50
DSP_INFO UNSUPPORTED_KEY: 49
DSP_INFO UNSUPPORTED_KEY: 50
PS C:\Users\Acer>
It seems the GenieX runtime is trying to query DSP capabilities (Key 49 and 50) which are not supported by the current firmware/driver version on the Snapdragon X Elite platform, causing the NPU initialization to fail silently.
Describe the bug
When trying to run the
ai-hub-models/Qwen3-8Bmodel using thegeniexCLI on Windows 11 ARM64 (Snapdragon X Elite), the command fails immediately after downloading. The process printsDSP_INFO UNSUPPORTED_KEY: 49andDSP_INFO UNSUPPORTED_KEY: 50repeatedly and then exits to the terminal prompt without starting the inference process.To Reproduce
Steps to reproduce the behavior:
geniex infer ai-hub-models/Qwen3-8BExpected behavior
The model should initialize successfully on the NPU (Hexagon DSP) and start the interactive inference session, or alternatively, fall back gracefully to the GPU or CPU runtime if NPU features are not supported.
Screenshots
If applicable, add screenshots to help explain your problem.
Desktop (please complete the following information):
Smartphone (please complete the following information):
Additional context
Here is the full terminal output:
It seems the GenieX runtime is trying to query DSP capabilities (Key 49 and 50) which are not supported by the current firmware/driver version on the Snapdragon X Elite platform, causing the NPU initialization to fail silently.