you can usually tell when things aren't running on gpu by looking at nvidia-smi though
-
-
-
Yeah, that's my method of choice for long running tasks. But when you're experimenting, developing and testing with short blocks of code that run in a couple of seconds they can easily not show up. And that's a really crude tool! :)
Kraj razgovora
Novi razgovor -
-
-
The best solution I've found is to use something like nvtop:https://github.com/Syllo/nvtop
-
As with nvidia-smi, that's great for long running processes but often you test and develop with small bits and pieces or you need to know if one particular bit ran on the GPU. Especially true for testing where you often want each test to be short so you can get through them all.
Kraj razgovora
Novi razgovor -
-
-
I just listen for the fan.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
You could use a library that parameterizes tensor types over compute devices, https://twitter.com/tscholak/status/1209260552247218177?s=21 …https://twitter.com/tscholak/status/1209260552247218177 …
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
YES! THIS BIT REALLY IS RUNNING ON THE GPU, NOT THE CPU YES! THIS BIT REALLY IS RUNNING ON THE GPU, NOT THE CPU YES! THIS BIT REALLY IS RUNNING ON THE GPU, NOT THE CPU [+] click to show 1048573 more lines
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Isn‘t this flag assumed to be given by the fans of the graphics card?
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Profiling tools?
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
...and vice versa
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.