- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hardware: Intel Joule 570x
System: Ubuntu 16.04
Software installed: caffe, Intel MKL, Intel OpenCL driver 4.1, Deep learning SDK
Issue: 'classification_sample -d GPU' returns failed to create engine, but 'classificaton_sample -d CPU' is ok.
Does anyone have any idea about this issue?
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
Joule is currently not supported in this release.
In addition, the supported OS for ubuntu is 14.04.
We will update once this changes.
Regards,
Stav
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I realize this is a really old thread, but I was able to get inference working on the Joule GPU.
I believe I used the same setup as the OP: Joule 570x / Ubuntu 16.04 / OpenCL driver 4.1 / Deep learning deployment toolkit / patched 4.7.6 kernel.
The original issue (for me) was that the i915 driver refused to load. So there was no OpenCL support for GPU. This can be overcome by using the experimental setting flag on the kernel command line in GRUB: i915.preliminary_hw_support=1
After rebooting, I was able to use clinfo to verify the OpenCL GPU device, and was able to run the inference samples on both CPU and GPU.
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page