Python 2.7, 3.5, and 3.6

Dear All,

I am trying to start using PyTorch.
I have a stupid question at the stage of installing PyTorch…
Is there any functional or performance differences on PyTorch over the different Python versions 2.7, 3.5, and 3.6?


py3 versions (3.5 and 3.6) also have CUDA multiprocessing available.

Apart from that feature difference, they are within epsilon of performance with 2.7.

1 Like

Thanks a lot!
I will try to use py3 version.

1 Like

I changed from 2.7 to 3.6 but am having problems with the model zoo. The Imagenet example will download pre-trained weights using a hickle file. Unfortunately it seems that the hickle module is not compatible with Python V 3.* due to the way strings are handled, which makes it difficult to use the pre-trained examples.

Does anybody know a way around this, I don’t want to have to go back to 2.7 if I can avoid it