Closed seanv507 closed 5 years ago
Hi @seanv507
Could you please clarify a little? I am not sure exactly what you mean. cuda
does start as False
in the args, but you can edit the args to change it to True
. It shouldn't automatically flip to True
if there is a GPU present, that would be too presumptuous of the runtime algorithm.
the success of nn's is because of gpu's. So I would expect the default to be use gpus unless there is no gpu, and obviously if for some reason you don't want to use a gpu even if you have one then you can set it to false. you might also want to compare with https://github.com/joosthub/PyTorchNLPBook/blob/db9fc8fe48a2416b36b21dde0dfce787c6ade2b2/chapters/chapter_3/3_5_Classifying_Yelp_Review_Sentiment.ipynb
hi @seanv507
Thanks for the response!
Part of the success of nn's is due to better compute, this is true. CPU as a default is just a design principle we meant to adhere to, and that other notebook should have been defaulting to False. It is easy for anyone to change that themselves, though :).
I will close this now but please file more issues if you have further concerns!
https://github.com/joosthub/PyTorchNLPBook/blob/db9fc8fe48a2416b36b21dde0dfce787c6ade2b2/chapters/chapter_4/4_4_cnn_surnames/4_4_Classifying_Surnames_with_a_CNN.ipynb#L552
code checks is not available to switch to false (but starts as false)