Liang Xiao

@zhijian-liu, good jod! But I think the bug is still there. in the code: ```python # do upsample original_stride = int(cur_stride / stride) kernel_map = inputs.kernel_maps.get( 'k%s_os%d_s%d_d%d' % (ks, original_stride,...

Hi, @zhijian-liu take the modified minkunet in repo [e3d](https://github.com/mit-han-lab/torchsparse) as an example, ```python class DilatedMinkUNet(MinkUNet): def __init__(self, **kwargs): super().__init__(**kwargs) cr = kwargs.get('cr', 1.0) dilation = kwargs.get('dilation', [1, 1, 1, 2,...

Thanks @joel-mb . But I successfully run them on another pc with all the same setting (i.e., ubuntu18.04 and ros melodic with python3)

same question

Also, the number of training images provided for download is different from the paper, i.e., 2422 VS. 2002.

I didn't use conda to install boost and eigen. Instead, I use the library in the system path and cmake find them well. cmake . -- The C compiler identification...

still not work