Easy AI model compilation tools for Tachyon, please!

As a Kickstarter backer of Tachyon, I'm very excited about this product. One humble request:

Please provide some (at least nearly) idiotproof tools for converting very commonly used models from ONNX to whatever format runs on the Tachyon.

I've found this to be a real headache with other edge AI computers like the Coral TPU, Seeed Vision AI card, and even the Hailo AI Kit. (The Hailo AI Kit compiler, for example, only runs on x86 machines. And the cloud setup is a bit of a pain.)

I suspect this is tricky, but even some limited idiotproof options would be very welcome. For example, if you have a standard YOLOv8n model, you can just drop it into a compiler and get a file you can just drop onto the Tachyon.

thanks for considering!

This will be discussed during Spectra, which is tomorrow (Wednesday)!

This will be discussed during Spectra, which is tomorrow (Wednesday)!

I'm curious whether there's any transcript or details from Spectra which are available.

Also: there's some quantized models available at qualcomm (Qualcomm) - maybe some of these can target Tachyon's CPU or NPU?

It's expected that you will be able to use the models and features of the Qualcom AI hub with Tachyon.