Hi
I am trying to use cuspatial to generate trajectories. I use the cuspatial.derive_trajectories()
method for the same. However I get the following error
objects, traj_offsets = cpp_derive_trajectories( File "trajectory.pyx", line 21, in cuspatial._lib.trajectory.derive_trajectories File "trajectory.pyx", line 29, in cuspatial._lib.trajectory.derive_trajectories RuntimeError: exclusive_scan failed to synchronize: cudaErrorIllegalAddress: an illegal memory access was encountered
As highlighted in https://github.com/rapidsai/cuspatial/issues/351
I am creating numpy arrays before passing them as params to cuspatial.derive_trajectories()
ids = data["DEVICEID"].to_numpy(np.uint32)
ts = data["DATETIME"].to_numpy(np.uint64)
xs = xys["x"].to_numpy(np.float64)
ys = xys["y"].to_numpy(np.float64)
The input data size is 24M probes
the machine type is :
- GPU type: NVIDIA Tesla P4
- Num GPUs: 1
- 1 vCPU, 3.75 GB memory
Can you please help us resolve this?