LGM icon indicating copy to clipboard operation
LGM copied to clipboard

Missing gridencoder source files

Open jclarkk opened this issue 1 year ago • 4 comments

Hey @ashawkey

Looks like another amazing project you've released :)

Having an issue when running the covert.py: FileNotFoundError: [Errno 2] No such file or directory: kiui/gridencoder/src/gridencoder.cu

jclarkk avatar Feb 08 '24 07:02 jclarkk

I am also running into this issue when running convert.py

zhizdev avatar Feb 08 '24 08:02 zhizdev

I've tried taking files from here: https://github.com/ashawkey/torch-ngp But then it seems the function is not called properly:

TypeError: grid_encode_forward(): incompatible function arguments. The following argument types are supported:
    1. (arg0: torch.Tensor, arg1: torch.Tensor, arg2: torch.Tensor, arg3: torch.Tensor, arg4: int, arg5: int, arg6: int, arg7: int, arg8: float, arg9: int, arg10: Optional[torch.Tensor], arg11: int, arg12: bool, arg13: int) -> None

Invoked with: tensor([[0.0135, 0.0089, 0.0110],
        [0.0032, 0.0090, 0.0301],
        [0.0132, 0.0054, 0.0452],
        ...,
        [0.9970, 0.9981, 0.9635],
        [0.9917, 0.9855, 0.9700],
        [0.9931, 0.9930, 0.9897]], device='cuda:0'), Parameter containing:
tensor([[ 5.6177e-05,  3.9003e-05],
        [-9.4321e-05,  6.2566e-05],
        [ 4.5953e-05, -4.0523e-05],
        ...,
        [-1.5921e-05,  5.3513e-05],
        [ 3.2847e-05, -1.3809e-05],
        [-4.4549e-05,  2.7390e-05]], device='cuda:0', requires_grad=True), tensor([      0,    4920,   40864,  315496,  839784, 1364072, 1888360, 2412648,
        2936936, 3461224, 3985512, 4509800, 5034088], device='cuda:0',
       dtype=torch.int32), tensor([[[0., 0.],
         [0., 0.],
         [0., 0.],
         ...,
         [0., 0.],
         [0., 0.],
         [0., 0.]],

        [[0., 0.],
         [0., 0.],
         [0., 0.],
         ...,
         [0., 0.],
         [0., 0.],
         [0., 0.]],

        [[0., 0.],
         [0., 0.],
         [0., 0.],
         ...,
         [0., 0.],
         [0., 0.],
         [0., 0.]],

        ...,

        [[0., 0.],
         [0., 0.],
         [0., 0.],
         ...,
         [0., 0.],
         [0., 0.],
         [0., 0.]],

        [[0., 0.],
         [0., 0.],
         [0., 0.],
         ...,
         [0., 0.],
         [0., 0.],
         [0., 0.]],

        [[0., 0.],
         [0., 0.],
         [0., 0.],
         ...,
         [0., 0.],
         [0., 0.],
         [0., 0.]]], device='cuda:0'), 262144, 3, 2, 12, 12, 1.0, 16, None, 0, False, 0```

jclarkk avatar Feb 08 '24 08:02 jclarkk

Thanks for reporting the bug! Could you try this and rerun convert.py?

pip install -U git+https://github.com/ashawkey/kiuikit.git

ashawkey avatar Feb 08 '24 08:02 ashawkey

Worked but then I encountered this: File "/home/ubuntu/.local/lib/python3.10/site-packages/nerfacc/cuda/init.py", line 13, in call_cuda return getattr(_C, name)(*args, **kwargs) AttributeError: 'NoneType' object has no attribute 'ray_aabb_intersect'

So I've tried installing nerfacc like this: pip install -U git+https://github.com/nerfstudio-project/nerfacc.git

And now everything works in convert.py

Thanks @ashawkey !

jclarkk avatar Feb 08 '24 08:02 jclarkk