CUDA kernel failed : invalid device function

[fuweifu-vtoo]

I ran train_cls.py, then encounter this error:

the error is:
CUDA kernel failed : invalid device function
void furthest_point_sampling_kernel_wrapper(int b, int n, int m,
const float *dataset, float *temp,
int *idxs) in /Pointnet2_PyTorch/pointnet2/_ext-src/src/sampling_gpu.cu

Do you know why?

[erikwijmans]

No idea what is causing that :/

[YimingCuiCuiCui]

I had the same problem and finally solved it. This may be caused by the version of Pytorch, try to recreate a virtual environment and follow the installment guide.

[frostinassiky]

@YimingCuiCuiCui 's solution works in my case. I met the same error with Python 3.7.6.
Then I created a new conda environment with Python 3.6.10 and the code was running successfully!

[w5688414]

I meet the same error:

Validation sanity check: 0it [00:00, ?it/s]CUDA kernel failed : invalid device function
void furthest_point_sampling_kernel_wrapper(int, int, int, const float*, float*, int*) at L:228 in pointnet2_ops/_ext-src/src/sampling_gpu.cu
[1]    28932 segmentation fault (core dumped)  python pointnet2/train.py task=cls

my python version is 3.6.4

[ghost]

did you manage to solve it? isn't it because of inadequate shared memory (probably caused by large
batch size)?