[SOLVED] Problem Configuring AI CORE X mini-pcie with OpenVino

giov · July 2019

Hi all
I have an AI core X mPcie with 1 Myriad X.

I have installed OpenVino in ubuntu 16.04 system .

In lsub i have "Bus 004 Device 002: ID 03e7:2485 "

When i try some demos samples from OpenVino toolkit (latest release) and MobileNet or Yolo as network , i got the following error

using -d MYRIAD option

" E: [xLink] [ 673769] dispatcherEventSend:934 Write failed event -7

E: [xLink] [ 675144] dispatcherEventReceive:347 dispatcherEventReceive() Read failed -4 | event 0x7efdeedaec60 XLINK_WRITE_RESP

E: [xLink] [ 675144] eventReader:233 eventReader stopped
E: [watchdog] [ 675146] sendPingMessage:132 Failed send ping message: X_LINK_ERROR
E: [xLink] [ 675148] XLinkReadDataWithTimeOut:1343 Event data is invalid
E: [ncAPI] [ 675148] ncGraphAllocate:1822 Can't read input tensor descriptors of the graph, rc: X_LINK_ERROR
"

*** using -d HDDL as device**

" [17:40:52.3660][1710]ERROR[HddlMyriadXDevice.cpp:327] Error: ncGraphAllocate() failed, device=4.1 rc=-6
[17:40:52.3663][1710]W[HddlMyriadXDevice.cpp:351]
** Suggestion: This error is probably caused by the graph allocation requires more time than the set timeout (30000 ms)
due to the large graph size 36 MB.
You can try to increase the timeout with following steps:
(1) increasing the "alloc_graph_timeout" value in $HDDL_INSTALL_DIR/hddl_service.config;
(2) kill hddldaemon: [Linux] kill -9 $(pidof hddldaemon); [Windows] kill with Windows Task Manager;
(3) restart your application again.
** OR caused by the graph used doesn't match to firmware running on MyriadX.
In this case, you need to check your network to see whether it's supported by MyriadX. "

Can you help me ? Thank you

tjirsch · August 2019

Did you resolve your problem ? I seem to have similar issues.

DCleri · August 2019

Are you using the latest version of OpenVINO?

Please make sure you follow the steps to enable the Myriad-X:

using the MYRIAD plugin: https://docs.openvinotoolkit.org/latest/_docs_install_guides_installing_openvino_linux.html#additional-NCS-steps
using the HDDL plugin: https://docs.openvinotoolkit.org/latest/_docs_install_guides_installing_openvino_linux_ivad_vpu.html

tjirsch · August 2019

Thank you for your support, I really appreciate it!

I have tried all available 2019 Releases up to the latest one on a current ubuntu 16.04 image and followed the installation docs to replace the generic kernels.
If I get you right, there a two separate ways to enable the AI Core X module:
1. Use and configure it as if it were a NCS1/2 USB Stick and follow the instructions for them, mainly about the usb rules and the user enablement
2. Use it as a VPU and configure it according to the "vision accelerator design with Movidius"... part

If I try 2. I get compilation errors about missing files when building HDDL Drivers. I found another post about that error: https://forum.up-community.org/discussion/comment/10136#Comment_10136

I will try both ways from scratch and report back.

Could you move this post: https://forum.up-community.org/discussion/3879/openvino-installation-on-realsense-robotics-kit-fails-for-ai-core-x-myriad-step#latest

to the Intel Realsense Robotics Kit Category and maybe create an installation or setup subcategory ?
I don not seem to be able to do that.

Again, thank you.

tjirsch · August 2019

My AI Core X is now working. I installed Intel Openvino Distribution 2019 R2.01 and also did the steps for the "Steps for Intel® Movidius™ Neural Compute Stick and Intel® Neural Compute Stick 2".
After that, all demos worked with the -d MYRIAD option.
I thought that the AI Core X module could not be a usb device because of the mpcie Connector, so I never followed the steps in the above section. This could be clarified in the documentation.
I did not try to use the HDDL option / installation.

Thanks again for the support !

thiirane · March 2020

I have a question. What do I get for buying a AI core X mPcie with 1 Myriad X verses simply buying a second Neural Compute Stick 2? Thanks for taking my question.

DCleri · March 2020

The Neural Compute Stick 2 is an external accelerator while the AI Core X is a module that can be integrated inside your system.

That is the main difference so it depends on what are your needs.

When creating an embedded system you might prefer to have as much as possible integrated and not outside except cables which are mandatory (e.g. power, video)

[SOLVED] Problem Configuring AI CORE X mini-pcie with OpenVino

Comments

Tagged

Categories