cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Highlighted
Contributor
Contributor
727 Views
Registered: ‎03-14-2011

ml-suite v1.3 libxfdnn library

i try to run the latest ml-suite v1.3 with U250

however the classification jupyter notebook example fail

and the notebook cause kernel dead at the point of xdnn.computeFC

can anyone update the API of the xdnn

or provide the source code of xdnn library as in v1.2?

 

thanks

0 Kudos
4 Replies
Highlighted
Xilinx Employee
Xilinx Employee
689 Views
Registered: ‎09-11-2014

Hi,

Currently we are not publishing our C++ APIs. The current version of ml-suite uses C++ implementation of FC.

 

Can you post your usage of the API, and also the error message?

Thanks,

Bryan

 

0 Kudos
Highlighted
Contributor
Contributor
681 Views
Registered: ‎03-14-2011

thanks for your reply

i am running ml-suite v1.3 image_classification_caffe.ipynb with u250

it seems like the fpga part is completed as i can run the notebook up to Step 9

but when it comes to Step 10, Execute the FC Layers on the CPU, the kernel is reported dead

no further debug message

so i wonder maybe some of the argument is not valid / null. 

i compare to v1.2 and do spot some different in api argument list. 

but no further information can be found. 

would xilinx provide more information on xdnn apis or even source code like in v1.2?

or maybe just fix the out-of-the-box demo in github?

0 Kudos
Highlighted
Xilinx Employee
Xilinx Employee
676 Views
Registered: ‎09-11-2014

What AMI are you using?

The out of the box examples should of course work.

If you are not already, I would recommend that you start from the latest version of this AMI:

https://aws.amazon.com/marketplace/pp/B077FM2JNS

 

0 Kudos
Highlighted
Contributor
Contributor
672 Views
Registered: ‎03-14-2011

hi, i am running on premise u250 server

and the master github checkout is not working 

can you double check with the xdnn library?

0 Kudos