Torch `rearrange` throws warning about incorrect division when running `torch.jit.trace`

Describe the bug When running torch.jit.trace on a nn.Module that contains a rearrange operation, the following warning is raised:

/home/shogg/.cache/pypoetry/virtualenvs/mldi-96vt4Weu-py3.8/lib/python3.8/site-packages/torch/_tensor.py:575: UserWarning: floor_divide is deprecated, and will be removed in a future version of pytorch. It currently rounds toward 0 (like the 'trunc' function NOT 'floor'). This results in incorrect rounding for negative values.
To keep the current behavior, use torch.div(a, b, rounding_mode='trunc'), or for actual floor division, use torch.div(a, b, rounding_mode='floor'). (Triggered internally at  /pytorch/aten/src/ATen/native/BinaryOps.cpp:467.)
  return torch.floor_divide(self, other)

Two questions arise from this:

1. Why is there a division step happening?
2. What is the potential impact?

Reproduction steps Code to reproduce this is as follows:

class TestModule(nn.Module):
    def forward(self, x: torch.Tensor) -> torch.Tensor:
        return einops.rearrange(x, '1 i f x y -> 1 i x y f')

test_tensor = torch.rand([1, 2, 376, 16, 16])
test_module = TestModule()
output_module = torch.jit.trace(test_module, test_tensor)

Expected behavior rearrange shouldn’t be throwing this warning

Your platform

python==3.8.0
torch==1.9.0
einops==0.3.0

[StephenHogg]

Ok, the pytorch issue is open - hope this helps

[arogozhnikov]

torch team decided to follow normal warn-change behavior policy.

Einops wasn’t affected and wouldn’t be affected by this torch transition. Warning are however still popping up during tracing (and will for a while).

Leaving this issue open for those who may encounter this in the future.