search

sniklaus / softmax-splatting

an implementation of softmax splatting for differentiable forward warping using PyTorch
468 stars 58 forks source link

compile_with_cache deprecated #49

Closed JHLew closed 2 years ago

JHLew commented 2 years ago

Hi, thank you for sharing the code to the public. I tried running your implementation, but seems like cupy.cuda.compile_with_cache() of cupy is deprecated.

I felt that this is too minor for a pull request, so I am leaving this issue in case someone runs into the same problem.

The code seems to run properly once the function cupy_launch (lines 245-246 of softsplat.py) is fixed as below.

def cupy_launch(strFunction, strKernel):
    return cupy.RawModule(code=strKernel).get_function(strFunction)

Hope this helps anyone in need!

sniklaus commented 2 years ago

Thanks for sharing this! You were probably hitting this bug in CuPy: https://github.com/cupy/cupy/issues/6231

An alternative to changing the softmax splatting implementation would be to downgrade CuPy to an older version, or wait for the next version of CuPy.