ImportError: libcudart.so.10.2: cannot open shared object file: No such file or directory

I tried to run the CPU version following the installation but I have the error from the title.
`ImportError: libcudart.so.10.2: cannot open shared object file: No such file or directory`

## Environment:

1. `python -m pip install detectron2==0.6 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cpu/torch1.10/index.html`

2. I also installed the required torch version (with cpu)
```
pip install torch==1.10.0+cpu torchvision==0.11.0+cpu torchaudio==0.10.0 -f https://download.pytorch.org/whl/torch_stable.html
```

I was able to import detectron2 but importing other modules from inside of the package looks like need to have installed CUDA.
I saw previous issues resolved by renaming the repository and using the installed package. However, this is not an issue here because I don't even have the git repo cloned on my directory. I'm using it directly from the package.

## Instructions To Reproduce the Issue:
When I try to do
`from detectron2 import model_zoo`
```
---------------------------------------------------------------------------
ImportError Traceback (most recent call last)
Cell In[14], line 1
----> 1 from detectron2 import model_zoo

File ~/miniconda3/envs/py39/lib/python3.9/site-packages/detectron2/model_zoo/__init__.py:8
1 # Copyright (c) Facebook, Inc. and its affiliates.
2 """
3 Model Zoo API for Detectron2: a collection of functions to create common model architectures
4 listed in `MODEL_ZOO.md `_,
5 and optionally load their pre-trained weights.
6 """
----> 8 from .model_zoo import get, get_config_file, get_checkpoint_url, get_config
10 __all__ = ["get_checkpoint_url", "get", "get_config_file", "get_config"]

File ~/miniconda3/envs/py39/lib/python3.9/site-packages/detectron2/model_zoo/model_zoo.py:9
7 from detectron2.checkpoint import DetectionCheckpointer
8 from detectron2.config import CfgNode, LazyConfig, get_cfg, instantiate
----> 9 from detectron2.modeling import build_model
12 class _ModelZooUrls(object):
13 """
14 Mapping from names to officially released Detectron2 pre-trained models.
15 """

File ~/miniconda3/envs/py39/lib/python3.9/site-packages/detectron2/modeling/__init__.py:2
1 # Copyright (c) Facebook, Inc. and its affiliates.
----> 2 from detectron2.layers import ShapeSpec
4 from .anchor_generator import build_anchor_generator, ANCHOR_GENERATOR_REGISTRY
5 from .backbone import (
6 BACKBONE_REGISTRY,
7 FPN,
(...)
13 make_stage,
14 )

File ~/miniconda3/envs/py39/lib/python3.9/site-packages/detectron2/layers/__init__.py:3
1 # Copyright (c) Facebook, Inc. and its affiliates.
2 from .batch_norm import FrozenBatchNorm2d, get_norm, NaiveSyncBatchNorm, CycleBatchNormList
----> 3 from .deform_conv import DeformConv, ModulatedDeformConv
4 from .mask_ops import paste_masks_in_image
5 from .nms import batched_nms, batched_nms_rotated, nms, nms_rotated

File ~/miniconda3/envs/py39/lib/python3.9/site-packages/detectron2/layers/deform_conv.py:11
8 from torch.nn.modules.utils import _pair
9 from torchvision.ops import deform_conv2d
---> 11 from detectron2 import _C
13 from .wrappers import _NewEmptyTensorOp
16 class _DeformConv(Function):

ImportError: libcudart.so.10.2: cannot open shared object file: No such file or directory
```

Reproduce the import with the listed CPU installations, starting at detectron2/model_zoo/__init__.py and following the traceback through detectron2/modeling, detectron2/layers/deform_conv.py, and the detectron2._C import. Compare the CPU wheel behavior while importing model_zoo; done means the documented CPU installation imports these modules without requiring the missing CUDA runtime.

Written by the indexing model from the issue text.