Apex: Install issue (nvcc fatal : Unsupported gpu architecture 'compute_75')

Created on 16 Jan 2020  Â·  1Comment  Â·  Source: NVIDIA/apex

Hi, I used the command line below to install apex.
pip install -v --no-cache-dir --global-option="--pyprof" --global-option="--cpp_ext" --global-option="--cuda_ext" ./

Then I got the following error message:

torch.__version__ = 1.4.0+cu92
/tmp/pip-req-build-gymhu27h/setup.py:43: UserWarning: Option --pyprof not specified. Not installing PyProf dependencies!
warnings.warn("Option --pyprof not specified. Not installing PyProf dependencies!")

Compiling cuda extensions with
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Tue_Jun_12_23:07:04_CDT_2018
Cuda compilation tools, release 9.2, V9.2.148
from /usr/local/cuda/bin

...

cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++
/usr/local/cuda/bin/nvcc -I/opt/conda/lib/python3.6/site-packages/torch/include -I/opt/conda/lib/python3.6/site-packages/torch/include/torch/csrc/api/include -I/opt/conda/lib/pytho
n3.6/site-packages/torch/include/TH -I/opt/conda/lib/python3.6/site-packages/torch/include/THC -I/usr/local/cuda/include -I/opt/conda/include/python3.6m -c csrc/multi_tensor_sgd_kernel
.cu -o build/temp.linux-x86_64-3.6/csrc/multi_tensor_sgd_kernel.o -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr --c
ompiler-options '-fPIC' -lineinfo -O3 --use_fast_math -DVERSION_GE_1_1 -DVERSION_GE_1_3 -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=amp_C -D_GLIBCXX_USE_CXX11_ABI=0 -gencode
=arch=compute_75,code=sm_75 -std=c++11
nvcc fatal : Unsupported gpu architecture 'compute_75'
error: command '/usr/local/cuda/bin/nvcc' failed with exit status 1
Running setup.py install for apex ... error

Do you know how to fix this problem? Thx

Most helpful comment

Solved by export TORCH_CUDA_ARCH_LIST="7.0"

For more details https://github.com/torch/torch7/issues/1190#issuecomment-498934400

>All comments

Solved by export TORCH_CUDA_ARCH_LIST="7.0"

For more details https://github.com/torch/torch7/issues/1190#issuecomment-498934400

Was this page helpful?
0 / 5 - 0 ratings