#pytorch 0.4.0 release
if this is pytorch 0.4.0 release version. i just check the latest release version of pytorch in github, the version is 0.3.1, so if there would be some 0.4.0 source file be released.
this is the latest version of pytorch
set cuda environment
torch.version = 0.3.1.post2
torch.version.cuda = 9.1.85
torch.backends.cudnn.version() = 7005
os[‘CUDA_VISIBLE_DEVICES’] = 0
torch.cuda.device_count() = 1
torch.cuda.current_device() = 0
All commits after 0.3.1 have that format. It shows the version number of next release, not current release. As I said, 0.4 is not finished and not released yet.
not sure if that would help in terms of avoiding confusion in future, but maybe switching to the more common “dev0” convention (at least more common in the scientific python community) might be worthwhile. E.g., instead of '0.4.0a0' using '0.4.dev0' and so forth
hi, I have solved this problem, in setup.py, the code assigns your pytorch version according to your machine statement. so you need to comment the version part.In the final step, you should install torch vision alone.
I don’t understand what the problem you are describing is. Weren’t you asking about 0.4? Why do you need to comment out part of the setup.py? Does it not build if you don’t do it?