Closed bastienvans closed 3 years ago
Do you have it running in docker? Did you install the edge TPU drivers?
Do you have it running in docker? Did you install the edge TPU drivers?
Yes, I installed docker on the edge tpu. It installed fine.
I believe the drivers are installed by default on the Coral Board.
Did you run docker with --device /dev/bus/usb
giving access to the TPU?
Did you run docker with
--device /dev/bus/usb
giving access to the TPU?
I haven't tried this because I am not connecting a USB Coral. I have the Coral Dev board and I've installed your library on the board itself. I am not using a raspberry pi or similar.
Cheers
Try running it with --device /dev/bus/usb
anyhow and see if that works. You have to figure out where the device is exposed under /dev
and pass that to the docker container.
This is the output.
mendel@coy-zebra:~/doods$ sudo docker run -it -v ~/doods/models:/opt/doods/models -v ~/doods/example.yaml:/opt/doods/config.yaml --device /dev/bus/usb -p 8080:8080 snowzach/doods:latest
2020-04-26T10:41:21.769Z INFO detector/detector.go:78 Configured Detector {"package": "detector", "name": "default", "type": "tflite", "model": "models/coco_ssd_mobilenet_v1_1.0_quant.tflite", "labels": 80, "width": 300, "height": 300}
2020-04-26T10:41:21.782Z ERROR detector/detector.go:73 Could not initialize detector edgetpu: no edgetpu devices detected {"package": "detector"}
2020-04-26T10:41:23.292Z INFO detector/detector.go:78 Configured Detector {"package": "detector", "name": "tensorflow", "type": "tensorflow", "model": "models/faster_rcnn_inception_v2_coco_2018_01_28.pb", "labels": 65, "width": -1, "height": -1}
This is /dev:
apex_0 cpu_dma_latency gpiochip0 i2c-1 loop0 loop-control mmcblk0p2 net pts ptyp7 ptypf stderr tty12 tty2 tty27 tty34 tty41 tty49 tty56 tty63 ttyp0 ttyp8 ttyS0 vcs1 vcsa1 vhci
autofs cuse gpiochip1 i2c-2 loop1 mapper mmcblk0p3 network_latency ptyp0 ptyp8 random stdin tty13 tty20 tty28 tty35 tty42 tty5 tty57 tty7 ttyp1 ttyp9 ttyS1 vcs2 vcsa2 video0
block disk gpiochip2 initctl loop2 mem mmcblk0rpmb network_throughput ptyp1 ptyp9 rtc stdout tty14 tty21 tty29 tty36 tty43 tty50 tty58 tty8 ttyp2 ttypa ttyS2 vcs3 vcsa3 watchdog
btrfs-control dri gpiochip3 input loop3 memory_bandwidth mmcblk1 null ptyp2 ptypa rtc0 tty tty15 tty22 tty3 tty37 tty44 tty51 tty59 tty9 ttyp3 ttypb ttyS3 vcs4 vcsa4 watchdog0
bus fd gpiochip4 ion loop4 mmcblk0 mmcblk1p1 port ptyp3 ptypb shm tty0 tty16 tty23 tty30 tty38 tty45 tty52 tty6 ttyGS0 ttyp4 ttypc ubi_ctrl vcs5 vcsa5 zero
cec0 full hugepages kmsg loop5 mmcblk0boot0 mmcblk1p2 pps0 ptyp4 ptypc snd tty1 tty17 tty24 tty31 tty39 tty46 tty53 tty60 ttymxc0 ttyp5 ttypd urandom vcs6 vcsa6
char fuse hwrng kvm loop6 mmcblk0boot1 mqueue ptmx ptyp5 ptypd spidev0.0 tty10 tty18 tty25 tty32 tty4 tty47 tty54 tty61 ttymxc1 ttyp6 ttype v4l vcs7 vcsa7
console galcore i2c-0 log loop7 mmcblk0p1 mxc_hantro ptp0 ptyp6 ptype spidev0.1 tty11 tty19 tty26 tty33 tty40 tty48 tty55 tty62 ttymxc2 ttyp7 ttypf vcs vcsa vga_arbiter
Does it say anything when you run lsusb
do you see the edgeTPU? I know it has issues with power sometimes.
I believe /dev/apex_0
is the PCIe based Coral. Try passing that into Docker.
I believe
/dev/apex_0
is the PCIe based Coral. Try passing that into Docker.
This worked, thank you so much! Consider adding it to the docs/wiki! 👍
How stable is the coral pcie device?
@runningman84 I have no issues. I had stability issues with the USB device when I would pass it into a VM. The Coral PCIe works great with VFIO passthrough into the VM.
Closing
Hello,
I've installed doods on my EdgeTPU dev board hoping it would make use of its HWAccel but it doesn't detect it.
ERROR detector/detector.go:73 Could not initialize detector edgetpu: no edgetpu devices detected {"package": "detector"}
Any help will be greatly appreciated!
Thanks