tensorflow - Using Google Coral mPCIe TPU in qemu VM

Question

I'm trying to use the Google Coral mPCIe EdgeTPU, which works perfectly fine in the host system, in a qemu VM and I'm unable to get any of the examples working.

Host System: Ubuntu 20.04

Guest System: Fedora 33 with Python 3.8

I have installed the necessary drivers and tools (edgetpu) after some trial and error, setup the PCI passthrough from Host to Guest system and the Edge TPU device is listed in the Guest system as expected (/dev/apex_0 exists and Coral Edge TPU is listed in lspci call).

But when I'm trying to start one of the examples, I get the following error message:

RuntimeError: Error in device opening (/dev/apex_0)!

and dmesg lists the following errors:

RAM did not enable within timeout (12000 ms)
Error in device open cb: -110

Any help would be appreciated.

Answer 1

等待大神答复