FindDefinition

@InfiniteSamele you can try spconv 2.2, this bug is related to thrust, i can't reproduce this in my environment. I need a minimal reproduce script to debug it.

@InfiniteSamele I can't reproduce this problem in V100 32G/16G with spconv-cu102/cu114/cu117 (spconv 2.2), test command: ```python -m spconv.benchmark bench_basic f16```, test nvidia driver: 515. That's why I need a minimal...

Sparse is better. You may need classifiers after each sparse transpose conv stage to remove empty voxels .

when you run conv in first time, the tuner object will iterate all available algo and choose fastest one, then cache result in tuner. this will cost much time. subsequence...

currently not, but looks like a good feature to use new cutlass feature quickly instead of re-implement them. I will consider this feature.

could you provide full traceback of this error?

fixed in spconv 2.2.3, you need to wait for this [build task](https://github.com/traveller59/spconv/actions/runs/3141187111) finish and update spconv.

This is a bug, these code should be covered by ```TV_CUDA``` macro, will be fixed soon. Do you use ```export CUMM_CUDA_VERSION=""```? If you haven't set this env, the cuda build...

~~Your problem [shouldn't happen](https://github.com/FindDefinition/cumm/blob/main/cumm/common.py#L258) if you haven't set ```CUMM_CUDA_VERSION```. you can print that env to check whether TV_CUDA is set.~~ The problem come from [this](https://github.com/FindDefinition/cumm/blob/main/cumm/tensorview_bind.py#L33), I need to change this...

1. you should use ```tv::from_blob``` to create ```tv::Tensor``` from tensorrt pointers. ```C++ // assume you record tensor metas in configure function auto out1_ten = tv::from_blob(out1_shape, out1_dtype, 0 /*GPU*/); ``` You...