I am trying to build pytorch for use with CUDA 12.
While attempting the command: python setup.py install I get 5 instances of the following error:
fatal error C1189: #error: This file was generated by an older version of protoc which is (compil
ing source file D:\Programming\pytorchSetup\build\caffe2\proto\caffe2.pb.cc) [D:\Programming\pytorchSetup\build\caffe2\proto\Caffe2_PROTO.vcxproj]
I then attempted installation using the pip install {path to pytorch local} command which ran for a long time then failed, this is the end of the log:
[5860/6367] Building CUDA object caffe2\CMakeFiles\cuda_vectorized_test.dir__\aten\src\ATen\test\cuda_vectorized_test.cu.obj
cuda_vectorized_test.cu
tmpxft_0000837c_00000000-7_cuda_vectorized_test.cudafe1.cpp
[5861/6367] Building CUDA object caffe2\CMakeFiles\torch_cuda.dir\__\aten\src\ATen\native\transformers\cuda\flash_attn\fmha_bwd_hdim64.cu.obj
fmha_bwd_hdim64.cu
tmpxft_000064a0_00000000-7_fmha_bwd_hdim64.cudafe1.cpp
[5862/6367] Building CUDA object caffe2\CMakeFiles\torch_cuda.dir\__\aten\src\ATen\native\cuda\Unique.cu.obj
Unique.cu
tmpxft_0000755c_00000000-7_Unique.cudafe1.cpp
ninja: build stopped: subcommand failed.
-- Building version 2.0.0a0+git0765dbc
cmake -GNinja -DBUILD_PYTHON=True -DBUILD_SPLIT_CUDA=1 -DBUILD_TEST=True -DCMAKE_BUILD_TYPE=Release -DCMAKE_INCLUDE_PATH=D:\Programming\mkl\include -DCMAKE_INSTALL_PREFIX=D:\Programming\pytorch\torch -DCMAKE_PREFIX_PATH=C:\Python310\Lib\site-packages -DJAVA_HOME=C:\Program Files\Java\jdk-14.0.1 -DNUMPY_INCLUDE_DIR=C:\Users\Connor\AppData\Local\Temp\pip-build-env-iut6kxu7\overlay\Lib\site-packages\numpy\core\include -DPYTHON_EXECUTABLE=C:\Python310\python.exe -DPYTHON_INCLUDE_DIR=C:\Python310\Include -DPYTHON_LIBRARY=C:\Python310/libs/python310.lib -DTORCH_BUILD_VERSION=2.0.0a0+git0765dbc -DUSE_NUMPY=True D:\Programming\pytorch
cmake --build . --target install --config Release
[end of output]
Thanks for your help!