As I was wondering if it was possible to train a network using TAO toolkit on Jetson AGX Orin hardware, I came across a thread saying it was impossible because “TAO is designed on training on the platform with x86-based CPU and discrete GPUs.” (Can I try tao toolkit in AGX Orin?).
Actually Jetson AGX Orin has integrated graphics (Tegra Orin) so it answers the question, I need a separate GPU.
But is the ARM based CPU architecture a problem here ?
Do we really need an x86 CPU architecture instead to train with TAO ?
Thank you in advance.
Is the ARM based CPU architecture a problem for TAO toolkit training?
Yes, as of now, for training, it is only supported in x86-based CPU and discrete GPUs.
For training on ARM based CPU architecture, it is not supported yet.
Thank you for your quick answer !
Do you know if there is a Jetson hardware compatible with TAO toolkit training then ?
Jetson is ARM based. As mentioned above, we do not officially support running training on it.
Ok thank you for your help !
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.