F.sphashquery() shows bug in cpu mode

mit-han-lab / torchsparse

[MICRO'23, MLSys'22] TorchSparse: Efficient Training and Inference Framework for Sparse Convolution on GPUs.

https://torchsparse.mit.edu

MIT License

1.16k stars 132 forks source link

F.sphashquery() shows bug in cpu mode #231

Closed Shadow581 closed 10 months ago

Shadow581 commented 11 months ago

Is there an existing issue for this?

[X] I have searched the existing issues

Current Behavior

Above the two codes, I take variable on cpu to F.sphashquery(), it shows bug. However, I take variable on gpu to F.sphashquery(), it is good.

Could you help me to solve the problem?

My torchsparse Library is torchsparse-2.1.0+torch112cu113-cp39-cp39-linux_x86_64.

Expected Behavior

fix it

Environment

- GCC: 9.4.0
- PyTorch: 1.12.1
- PyTorch CUDA: 11.3
- TorchSparse: 2.1.0+torch112cu113

Anything else?

None

zhijian-liu commented 10 months ago

@ys-2020, could you take a look at this issue?

Shadow581 commented 10 months ago

Nobody answer？

ys-2020 commented 10 months ago

Hi @Shadow581. Sorry for the late reply. We have temporarily deprecated the cpu sphashquery() in v2.1.0. And we are currently working on this problem. Hopefully we will have a cpu version of sphashquery() soon.

Shadow581 commented 10 months ago

@ys-2020 ok，thanks

ys-2020 commented 10 months ago

Hi @Shadow581 , we have solved this problem and update the wheels in our pypi sever. Could you please reinstall torchsparse v2.1.0 with python -c "$(curl -fsSL https://raw.githubusercontent.com/mit-han-lab/torchsparse/master/install.py)"? Thank you again for bringing this issue to our attention.