import apex always have AttributeError: module 'torch' has no attribute 'library'
always have
Traceback (most recent call last):
File "
no matter cuda10.1+torch1.6.0/1.7.1 python3.7 or cuda11.1+torch1.8.0 python3.8 or cuda11.3+torch1.10.1 python3.8, apex did nnot work!
Did you solve it?
Did you solve it? 你解决了吗?
NO, still bothers me
Did you solve it? 你解决了吗?
NO, still bothers me
you can try apex-22.02-parallel-state,I solve this problem just now cuda11.3+torch1.10.1+torchvision0.11.2+torchaudio0.10.1
Did you solve it? 你解决了吗?
NO, still bothers me 不,仍然困扰着我
you can try apex-22.02-parallel-state,I solve this problem just now你可以试试apex-22.02-parallel-state,我刚刚解决了这个问题 cuda11.3+torch1.10.1+torchvision0.11.2+torchaudio0.10.1
where to get apex-22.02-parallel-state? I can not find it. and python version? no 22.02
git clone https://github.com/NVIDIA/apex.git cd apex git tag # if not wanted version git branch -r git checkout origin/22.02-parallel-state # or another vesion python setup.py install --cuda_ext --cpp_ext pip3 install -v --no-cache-dir ./
git clone https://github.com/NVIDIA/apex.git cd apex git tag # if not wanted version git branch -r git checkout origin/22.02-parallel-state # or another vesion python setup.py install --cuda_ext --cpp_ext pip3 install -v --no-cache-dir ./
yes,you can try
git clone https://github.com/NVIDIA/apex.git cd apex git tag # if not wanted version git branch -r git checkout origin/22.02-parallel-state # or another vesion python setup.py install --cuda_ext --cpp_ext pip3 install -v --no-cache-dir ./
Thanks for your answer, I have solved it.
cuda11.3+torch1.10.1+torchvision0.11.2+torchaudio0.10.1
git clone https://github.com/NVIDIA/apex.git cd apex git tag # if not wanted version git branch -r git checkout origin/22.02-parallel-state # or another vesion python setup.py install --cuda_ext --cpp_ext pip3 install -v --no-cache-dir ./
It works in cuda 11.3, thank you a lot
我用这个版本成功解决了 git clone https://github.com/ptrblck/apex.git cd apex git checkout apex_no_distributed pip install -v --no-cache-dir ./
我用这个版本成功解决了 git clone https://github.com/ptrblck/apex.git cd apex git checkout apex_no_distributed pip install -v --no-cache-dir ./
thank you
我用这个版本成功解决了 git clone https://github.com/ptrblck/apex.git cd apex git checkout apex_no_distributed pip install -v --no-cache-dir ./
Thanks for sharing! This applies to my conda virtual environment on Ubuntu as well.
我用这个版本成功解决了 git clone https://github.com/ptrblck/apex.git cd apex git checkout apex_no_distributed pip install -v --no-cache-dir ./
Haha, legendary ptrblck to the rescue. :))
Thanks for sharing the fix, although I checked the differences in the no-distributed branch, it seems the latest version have those changes already. So, I think cloning the original repository (i.e. git clone https://github.com/NVIDIA/apex.git) should also work, imo.
Also, how do people find these fixes?!
git clone https://github.com/NVIDIA/apex.git cd apex git tag # if not wanted version git branch -r git checkout origin/22.02-parallel-state # or another vesion python setup.py install --cuda_ext --cpp_ext pip3 install -v --no-cache-dir ./
可以成功下载,但是print(apex.version)会显示AttributeError: module 'apex' has no attribute 'version', print(amp._amp_C.file)显示AttributeError: module apex.amp has no attribute _amp_C, 暂时忽略这些错误 ubuntu18.04+cuda10.2+torch1.7.1+torchvision0.8.2