Error in trainmodel Function

[BARIS BATUHAN TOPAL]

Hi everyone,

I am getting an error in the trainmodel function in the line:

(dev=devloss, tst=tstloss, mem=Float32(CUDA.usage[]))

and especially in the part:

mem=Float32(CUDA.usage[])

After removing that statement the code works without any problem. Is there anyone who have the same problem but already fixed it?

Thanks.

Barış Batuhan Topal

M.Sc. Computer Science Student

Koç University

[Deniz Yuret]

In case the problem you run into is version dependent, let's try to include version information with these questions. I suggest an attachment with the output of three commands:

julia> versioninfo()

julia> CUDA.versioninfo()

julia> pkg"st -m"

best,

deniz

--
For information about this course, please see http://courses.ku.edu.tr/comp542
For questions about the course, please email com...@ku.edu.tr
---
You received this message because you are subscribed to the Google Groups "Comp542 Class" group.
To unsubscribe from this group and stop receiving emails from it, send an email to Comp542+u...@ku.edu.tr.
To view this discussion on the web visit https://groups.google.com/a/ku.edu.tr/d/msgid/Comp542/CA%2BNhokvozwOhjiVRyQ9xFXTboZkN0mStk%3DivXPVEVsBMg%2BOriQ%40mail.gmail.com.

[BARIS BATUHAN TOPAL]

Dear Deniz Hocam,

The suggested codes are run and their outputs can be viewed from the .txt file attached.

Regards.

Barış Batuhan Topal

M.Sc. Computer Science Student

Koç University

[Deniz Yuret]

It seems the interface for CUDA.usage has changed, please replace

   CUDA.usage[]

with 

   CUDA.usage[CUDA.device()][]

You can also remove it entirely like Baris, it is just there for debugging.

best,

deniz