🐛 Describe the bug
Hello, I have encountered a problem while compiling ColossalAI within the Docker nvcr.io/nvidia/pytorch:23.07-py3. The error message is as follows:
nvcc warning : incompatible redefinition for option 'std', the last value of this option was used
/workspace/ColossalAI/colossalai/kernel/cuda_native/csrc/kernels/cuda_util.cu(148): error: namespace "thrust" has no member "transform_reduce"
res = thrust::transform_reduce(thrust::cuda::par.on(stream), data_ptr,
^
/workspace/ColossalAI/colossalai/kernel/cuda_native/csrc/kernels/cuda_util.cu(153): error: namespace "thrust" has no member "transform_reduce"
res = thrust::transform_reduce(thrust::cuda::par.on(stream), data_ptr,
^
2 errors detected in the compilation of "/workspace/ColossalAI/colossalai/kernel/cuda_native/csrc/kernels/cuda_util.cu".
From what I understand, it seems that the namespace "thrust" does not contain the member "transform_reduce". One potential solution I found is to include <thrust/transform_reduce.h> in the ColossalAI/colossalai/kernel/cuda_native/csrc/kernels/cuda_util.cu file.
Could you please help verify this solution or suggest any other potential fixes?
Environment
No response
🐛 Describe the bug
Hello, I have encountered a problem while compiling ColossalAI within the Docker nvcr.io/nvidia/pytorch:23.07-py3. The error message is as follows:
From what I understand, it seems that the namespace "thrust" does not contain the member "transform_reduce". One potential solution I found is to include <thrust/transform_reduce.h> in the ColossalAI/colossalai/kernel/cuda_native/csrc/kernels/cuda_util.cu file.
Could you please help verify this solution or suggest any other potential fixes?
Environment
No response