I want to implement training with oneDNN,
and use mallo/free instead of vector.
I used the sample code at the URL below.
I changed the first code of simple_net().
CHECK(dnnl_engine_create(&engine, dnnl_cpu, 0)); // idx
CHECK(dnnl_engine_create(&engine, dnnl_gpu, 0)); // idx
When I build and run it, but I get an exception error.
What are you missing?
I want to know the correct way to implement oneDNN learning process in C API and run it on GPU.
OS: Windows 10 Pro
Intel oneAPI basetoolkit : 2021.2
CPU: Intel Core i7-1165G7 2.80GHz
Accelerator:Intel Iris Xe Graphics
driver ver.: 184.108.40.20695(2021/04/29)
Thank you for your answer.
>We are working on this. We will get back to you soon with updates.
Is there something wrong with oneDNN library?
Is it just a sample code issue?
(For GPU, many changes to the sample code are required.)
For your information, we have set of samples that can work both on cpu and gpu but few samples are only cpu supported.Since your sample is cpu supported you are getting this error and not because of any bug.There are some oneDNN C APIs that support gpu.
You can refer this link (https://oneapi-src.github.io/oneDNN/dev_guide_examples.html) to know about the samples and its supported engine. Also you can follow oneDNN developer guide(https://oneapi-src.github.io/oneDNN/dev_guide_build.html) to build samples.
Thanks for accepting the solution. If you need any additional information, please submit a new question as this thread will no longer be monitored.