Hi, when I want to compile torch2.0 from source, I have met some problems.
The following is the bug information.
pytorch/c10/cuda/CUDACachingAllocator.cpp:1152:15: error: ‘nvmlProcessInfo_v1_t’ was not declared in this scope; did you mean ‘nvmlProcessInfo_t’?
1152 | std::vector<nvmlProcessInfo_v1_t> procs(8);
| ^~~~~~~~~~~~~~~~~~~~
| nvmlProcessInfo_t
/pytorch/c10/cuda/CUDACachingAllocator.cpp:1152:35: error: template argument 1 is invalid
1152 | std::vector<nvmlProcessInfo_v1_t> procs(8);
| ^
/pytorch/c10/cuda/CUDACachingAllocator.cpp:1152:35: error: template argument 2 is invalid
/pytorch/c10/cuda/CUDACachingAllocator.cpp:1153:29: error: request for member ‘size’ in ‘procs’, which is of non-class type ‘int’
1153 | unsigned int size = procs.size();
| ^~~~
/pytorch/c10/cuda/CUDACachingAllocator.cpp:1156:41: error: request for member ‘data’ in ‘procs’, which is of non-class type ‘int’
1156 | nvml_device, &size, procs.data())) ==
| ^~~~
/pytorch/c10/cuda/CUDACachingAllocator.cpp:1158:11: error: request for member ‘resize’ in ‘procs’, which is of non-class type ‘int’
1158 | procs.resize(size);
| ^~~~~~
/pytorch/c10/cuda/CUDACachingAllocator.cpp:1165:23: error: invalid types ‘int[unsigned int]’ for array subscript
1165 | auto& proc = procs[i];
| ^
Who can help me for this?