Brauche ich eine GPU für ML?
Hallo,
wollte mal nach Eurer Erfahrung zu Cloud Services wie google colab fragen bzgl des Trainings von ML. Lohnt es sich mit dieser Möglichkeit überhaupt sich einen Rechner mit GPU von Nvidia zu kaufen, wenn man noch nie gezockt hat und es wahrscheinlich auch nicht tun wird?
Well. Cloud is so expensive (must make profit and for which is the hardware/power, etc. also not cheap), but also buy a GPU.
It’s all about what you’re doing. ML on the CPU is very slow, but also possible within certain limits.
So google colab also has a free version with up to 12h a day access to a K80. Isn’t that solid?
This 12h, as well as everything else is not guaranteed (especially if at the same time many others use the service or you read or write many data). See: https://research.google.com/colaboratory/faq.html
It is also not exceptionally ergonomic to use everything except Python or R and the data can always be uploaded and downloaded quickly.
If you’re okay with it and that Google can see your data, you can use it that way. More than toys can’t be developed with this, which is so thoughtful.
From none
I don’t have my own records. I make this from private practice and would therefore be affected by my data risk