Installation fails when trying to build from source on MacOS

[skalaydzhiyski]

I am trying to build Pytorch on macOS Catalina 10.15.7 19H114 x86_64, but when I reach 94% I get the following error after running python3 setup.py install with python 3.8.2

[ 94%] Building CXX object caffe2/CMakeFiles/scalar_tensor_test.dir/__/aten/src/ATen/test/scalar_tensor_test.cpp.o Undefined symbols for architecture x86_64: [ 94%] Built target ssa_test "___kmpc_end_serialized_parallel", referenced from: TestParallel_NestedParallel_Test::TestBody() in test_parallel.cpp.o TestParallel_Exceptions_Test::TestBody() in test_parallel.cpp.o "___kmpc_fork_call", referenced from: TestParallel_NestedParallel_Test::TestBody() in test_parallel.cpp.o TestParallel_Exceptions_Test::TestBody() in test_parallel.cpp.o "___kmpc_global_thread_num", referenced from: TestParallel_NestedParallel_Test::TestBody() in test_parallel.cpp.o TestParallel_Exceptions_Test::TestBody() in test_parallel.cpp.o "___kmpc_serialized_parallel", referenced from: TestParallel_NestedParallel_Test::TestBody() in test_parallel.cpp.o TestParallel_Exceptions_Test::TestBody() in test_parallel.cpp.o "_omp_get_max_threads", referenced from: TestParallel_NestedParallel_Test::TestBody() in test_parallel.cpp.o TestParallel_Exceptions_Test::TestBody() in test_parallel.cpp.o "_omp_get_num_threads", referenced from: _.omp_outlined. in test_parallel.cpp.o _.omp_outlined..33 in test_parallel.cpp.o _.omp_outlined..35 in test_parallel.cpp.o "_omp_get_thread_num", referenced from: _.omp_outlined. in test_parallel.cpp.o _.omp_outlined..33 in test_parallel.cpp.o _.omp_outlined..35 in test_parallel.cpp.o "_omp_in_parallel", referenced from: TestParallel_NestedParallel_Test::TestBody() in test_parallel.cpp.o TestParallel_Exceptions_Test::TestBody() in test_parallel.cpp.o ld: symbol(s) not found for architecture x86_64 [ 94%] Building CXX object caffe2/CMakeFiles/scalar_test.dir/__/aten/src/ATen/test/scalar_test.cpp.o clang: error: linker command failed with exit code 1 (use -v to see invocation)

Has anyone seen this before ?

cc @malfet @seemethere @walterddr

[skalaydzhiyski]

Please close this issue, I have fixed it by using conda, as per the installation guide.
Something must have happened because my million different Python versions.

Thanks anyway.