My trouble with this, it's based on the Qualcomm QPM software and I'm not successful in downloading it. Somehow I simply cannot find the software on Qualcomm's website. I have a personal developer account.
So therefore I downloaded Ollama instead. However, when running AI models with Ollama, is Tachyon's AI Accelerator used by default? Or should I do something specific for that?
Here is the response from Qualcomm. It seems Qualcomm Package Manager 3 is deprecated in favor of Qualcomm Software Center.
Summary Step-by-Step:
Install Qualcomm Software Center (QSC).
Search for "AI Engine Direct" inside QSC.
If it still says "Restricted," you must request access permissions (QGroup membership) from your project lead or the partner support channel for Particle Tachyon.
After installing âQualcomm Software Centerâ on arm64 host, and then searching for âAI Engine Directâ inside QSC, I was able to activate the license but get âNo releases foundâ.
From the CLI, âqpm-cli âextract Qualcomm_ai_engine_directâ fails with error âDownload failed as releases are not available for the input product: qualcomm_ai_engine_direct.
The following commands worked fine:
qpm-cli --login
qpm-cli --license-activate qualcomm_ai_engine_direct
The Web version of QSC, lists downloads only for Linux x86 and Windows x86/arm64.
Do we need to use the Windows ARM64 to get the runtime libraries? Thanks.
Update: âQualcomm AI Runtime SDKâ has the required âAI Engine Directâ runtime libraries but no indication if it is for arm64 or x86.