How building open3d for various CUDA architectures?

### Checklist

- [X] I have searched for [similar issues](https://github.com/isl-org/Open3D/issues).
- [X] For Python issues, I have tested with the [latest development wheel](http://www.open3d.org/docs/latest/getting_started.html#development-version-pip).
- [X] I have checked the [release documentation](http://www.open3d.org/docs/release/) and the [latest documentation](http://www.open3d.org/docs/latest/) (for `master` branch).

### My Question

Hello. I have a problem when running open3d built on the same GPU version with other GPU versions. I used the build script https://github.com/isl-org/Open3D/blob/master/docker/docker_build.sh cuda_wheel_py310_dev on my 1080TI GPU machine (sm_code 61). After successfully building the *whl file, I install open3d on a machine with a V100 GPU (sm_code 70). Installation succeeds, but an error occurs when using functions

```src/segmentator.py:91: in predict_mask
results = self.model(inputs["data"])
/root/.pyenv/versions/3.10.7/lib/python3.10/site-packages/torch/nn/modules/module.py:1130: in _call_impl
return forward_call(*input, **kwargs)
/opt/open3d-ml/ml3d/torch/models/sparseconvnet.py:86: forward
feat_list = self.sub_sparse_conv(feat_list, pos_list, voxel_size=1.0)
/root/.pyenv/versions/3.10.7/lib/python3.10/site-packages/torch/nn/modules/module.py:1130: in _call_impl
return forward_call(*input, **kwargs)
/opt/open3d-ml/ml3d/torch/models/sparseconvnet.py:390: forward
out_feat.append(self.net(feat, in_pos, out_pos, voxel_size))
/root/.pyenv/versions/3.10.7/lib/python3.10/site-packages/torch/nn/modules/module.py:1130: in _call_impl
return forward_call(*input, **kwargs)
/root/.pyenv/versions/3.10.7/lib/python3.10/site-packages/open3d/ml/torch/python/layers/convolutions.py:501: in forward
self.nns = self.fixed_radius_search(
/root/.pyenv/versions/3.10.7/lib/python3.10/site-packages/torch/nn/modules/module.py:1130: in _call_impl
return forward_call(*input, **kwargs)
/root/.pyenv/versions/3.10.7/lib/python3.10/site-packages/open3d/ml/torch/python/layers/neighbor_search.py:154: in forward
result = ops.fixed_radius_search(
/root/.pyenv/versions/3.10.7/lib/python3.10/site-packages/open3d/ml/torch/python/ops.py:877: in fixed_radius_search
*_torch.ops.open3d.fixed_radius_search(
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

self = , args = ()
kwargs = {'hash_table_cell_splits': tensor([ 0, 0, 0, -1215135857, -1247819812,
92484002...a:0', dtype=torch.int32), 'hash_table_splits': tensor([ 0, 384], dtype=torch.int32), 'ignore_query_point': False, ...}

def __call__(self, *args, **kwargs):
# overloading __call__ to ensure torch.ops.foo.bar()
# is still callable from JIT
# We save the function ptr as the `op` attribute on
# OpOverloadPacket to access it here.
> return self._op(*args, **kwargs or {})
E RuntimeError: Trying to create tensor with negative dimension -5394801434127068701: [-5394801434127068701]

/root/.pyenv/versions/3.10.7/lib/python3.10/site-packages/torch/_ops.py:143: RuntimeError
```

This error is eliminated if open3d building a machine with GPU v100.

I know cmake has options to set architectures. For example:
-DCMAKE_CUDA_ARCHITECTURES="60;61;70"

How to pass these flags to the build tool https://github.com/isl-org/Open3D/blob/master/docker/docker_build.sh and is it possible to build open3d for different CUDA architectures?