Compatibility with pytorch 1.12.0

[htkseason]

Hi, met some problem when using tinycudnn-1.6+pytorch1.12.0+cuda11.3 Works well with pytorch1.11.0, so I guess it was a compatibility problem with pytorch1.12.0. It happens with tcnn.Encoding using dense grid.

  File "***/anaconda3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
    return forward_call(*input, **kwargs)
  File "***/anaconda3/lib/python3.9/site-packages/tinycudann-1.6-py3.9-linux-x86_64.egg/tinycudann/modules.py", line 122, in forward
    self.params.to(_torch_precision(self.native_tcnn_module.param_precision())).contiguous(),
RuntimeError: Tensors of type TensorImpl do not have strides

[Tom94]

On my machine, PyTorch bindings compile and run on v1.12.0+cu116 -- unfortunately the runtime error sounds like some PyTorch-internal thing going wrong.

[Choconuts]

I have the same problem, too. How did you solve it?

[htkseason]

I have the same problem, too. How did you solve it?

I tried to start from a clean conda environment and it magically fixed...

[yyashpatel]

Hi,

I am also facing the same issue with pytorch 1.12.0 + cu11.6.

File "/home/yash/anaconda3/envs/NeRF_env/lib/python3.8/site-packages/tinycudann/modules.py", line 122, in forward self.params.to(_torch_precision(self.native_tcnn_module.param_precision())).contiguous(), RuntimeError: Tensors of type TensorImpl do not have strides

a new conda environment made the fix ?