CUDA kernel failed : invalid device function

[binlidaily]

When I try to train the model, I get CUDA kernel failed:

File "mmdet/core/bbox3d/geometry.py", line 240:
def points_in_convex_polygon_jit(points, polygon, clockwise=True):
    <source elided>
    # first convert polygon to directed lines
    num_points_of_polygon = polygon.shape[1]
    ^

  state.func_ir.loc))
CUDA kernel failed : invalid device function
Segmentation fault (core dumped)

I google the solution, which is said I should change my GPU setting, I use RTX 2080 Ti and CUDA 10.0, so I should set:

-gencode=arch=compute_75,code=sm_75

The problem is where should I do this setting?

Thank you very much!

[nmll]

you can use gdb

[hughlee815]

i also have the same problem and the same question. why does everyone not tell where to do the setting.

[mc171819]

i also have the same problem and the same question. why does everyone not tell where to do the setting.

did you solve this?