This is an old post, but having the same issue!

How do you properly kill the process? I have been using 'ctrl+c' which 
stops the process, but this does not free the memory on the GPU. The only 
way I have been able to free up the memory has been to kill iPython and 
restart it. Is there another way to do this without having to restart 
iPython? 

On Tuesday, September 2, 2014 at 8:56:26 PM UTC+1, Muneeb Saleem wrote:
>
> I think I figured out my mistake. I wasn't properly killing the processes 
> after stopping them.
>
> On Tuesday, September 2, 2014 12:59:13 PM UTC-5, Arnaud Bergeron wrote:
>>
>> Which version of cuda is this?
>>
>>
>> 2014-09-01 0:36 GMT-04:00 Muneeb Saleem <[email protected]>:
>>
>>> Whenever I manually kill my program/process, the GPU memory under use by 
>>> it (using Theano shared variables with borrow=True) does not get freed 
>>> automatically. Since I'm training with large chunks of data, after a few 
>>> subsequent run-time terminations, the whole GPU memory gets full. The only 
>>> way to clear that then is to restart the system since my Titan Black on 
>>> Ubuntu does not have any memory reset functions. Is there a way to clear 
>>> this memory since I have to manually shut down my program often?
>>>
>>> -- 
>>>
>>> --- 
>>> You received this message because you are subscribed to the Google 
>>> Groups "theano-users" group.
>>> To unsubscribe from this group and stop receiving emails from it, send 
>>> an email to [email protected].
>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>
>>

-- 

--- 
You received this message because you are subscribed to the Google Groups 
"theano-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to