Read_image doesn't work in pytroch 2.0 on google colab


I installed torch 2.0 on google colab using this comand
!pip3 install numpy --pre torch[dynamo] --force-reinstall --extra-index-url

and now doesn’t work. I get this warning when I’m importing it

/usr/local/lib/python3.9/dist-packages/torchvision/io/ UserWarning: Failed to load image Python extension: cannot open shared object file: No such file or directory warn(f"Failed to load image Python extension: {e}")

and this error when I try to load an image

RuntimeError                              Traceback (most recent call last)
/usr/local/lib/python3.9/dist-packages/torch/ in __getattr__(self, op_name)
    704         try:
--> 705             op, overload_names = torch._C._jit_get_operation(qualified_op_name)
    706         except RuntimeError as e:

RuntimeError: No such operator image::read_file

The above exception was the direct cause of the following exception:

AttributeError                            Traceback (most recent call last)
3 frames
<ipython-input-4-47407919c77a> in <module>
----> 1 image = read_image('/content/SSIW_single_node/Test_Minist/test_imgs/img1.png')

/usr/local/lib/python3.9/dist-packages/torchvision/io/ in read_image(path, mode)
    251     if not torch.jit.is_scripting() and not torch.jit.is_tracing():
    252         _log_api_usage_once(read_image)
--> 253     data = read_file(path)
    254     return decode_image(data, mode)

/usr/local/lib/python3.9/dist-packages/torchvision/io/ in read_file(path)
     45     if not torch.jit.is_scripting() and not torch.jit.is_tracing():
     46         _log_api_usage_once(read_file)
---> 47     data = torch.ops.image.read_file(path)
     48     return data

/usr/local/lib/python3.9/dist-packages/torch/ in __getattr__(self, op_name)
    707             # Turn this into AttributeError so getattr(obj, key, default)
    708             # works (this is called by TorchScript with __origin__)
--> 709             raise AttributeError(
    710                 f"'_OpNamespace' '{}' object has no attribute '{op_name}'"
    711             ) from e

AttributeError: '_OpNamespace' 'image' object has no attribute 'read_file'

This would install the nightly release of PyTorch, not 2.0.0, and might also not install torchvision unless it’s detected as a conflicting package.
I don’t know if the pip wheels ever supported custom backends or if source builds are needed so do you know if these ops were available in previous versions?