4217 Update PyTorch docker to 22.04

[Nic-Ma]

Fixes #4217 .

Description

This PR updated the PyTorch base docker and CI tests docker to 22.04.

Status

Ready

Types of changes

• Non-breaking change (fix or new feature that would not break existing functionality).
• Breaking change (fix or new feature that would cause existing functionality to change).
• New tests added to cover the changes.
• Integration tests passed locally by running ./runtests.sh -f -u --net --coverage.
• Quick tests passed locally by running ./runtests.sh --quick --unittests --disttests.
• In-line docstrings updated.
• Documentation updated, tested make html command in the docs/ folder.

[Nic-Ma]

/black

[Nic-Ma]

/build

[Nic-Ma]

/integration-test

[Nic-Ma]

Hi @rijobro ,

When I try to run our CI tests on the latest PyTorch docker, got the same error from MetaTensor test cases, like:

======================================================================
ERROR: test_copy_9 (__main__.TestMetaTensor)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/opt/conda/lib/python3.8/site-packages/parameterized/parameterized.py", line 533, in standalone_func
    return func(*(a + p.args), **p.kwargs)
  File "tests/test_meta_tensor.py", line 172, in test_copy
    a = deepcopy(m)
  File "/opt/conda/lib/python3.8/copy.py", line 153, in deepcopy
    y = copier(memo)
  File "/opt/conda/lib/python3.8/site-packages/torch/_tensor.py", line 86, in __deepcopy__
    return handle_torch_function(Tensor.__deepcopy__, (self,), self, memo)
  File "/opt/conda/lib/python3.8/site-packages/torch/overrides.py", line 1394, in handle_torch_function
    result = torch_func_method(public_api, types, args, kwargs)
  File "/workspace/data/medical/MONAI/monai/data/meta_tensor.py", line 187, in __torch_function__
    ret = super().__torch_function__(func, types, args, kwargs)
  File "/opt/conda/lib/python3.8/site-packages/torch/_tensor.py", line 1178, in __torch_function__
    ret = func(*args, **kwargs)
  File "/opt/conda/lib/python3.8/site-packages/torch/_tensor.py", line 141, in __deepcopy__
    raise RuntimeError("The default implementation of __deepcopy__() for non-wrapper subclasses "
RuntimeError: The default implementation of __deepcopy__() for non-wrapper subclasses only works for subclass types that implement new_empty() and for which that function returns another instance of the same subclass. You should either properly implement new_empty() for your subclass or override __deepcopy__() if it is intended behavior for new_empty() to return an instance of a different type.

Do you have any idea?

Thanks in advance.

[Nic-Ma]

BTW, except for the MetaTensor tests, I manually executed all the integration tests on V100 locally, all passed OK.

Thanks.

[rijobro]

@wyli did any of your recent MetaTensor commits fix the issue @Nic-Ma is referring to, or do I still need to look into this? Thanks

[wyli]

@wyli did any of your recent MetaTensor commits fix the issue @Nic-Ma is referring to, or do I still need to look into this? Thanks

yes, I forgot to get back to the dev branch, I'll look into this.

[wyli]

@wyli did any of your recent MetaTensor commits fix the issue @Nic-Ma is referring to, or do I still need to look into this? Thanks

yes, I forgot to get back to the dev branch, I'll look into this.

this seems to address the issue c095f92 please let me know if there's a better way

[Nic-Ma]

/build

[Nic-Ma]

@wyli @rijobro Thanks for your fix! This PR is ready for review now.

[Nic-Ma]

/build

[Nic-Ma]

/build

[wyli]

looks good to me, there might be a more efficient way to implement new_empty

MONAI/monai/data/meta_tensor.py

Line 244 in c095f92

self.as_tensor().new_empty(size=size, dtype=dtype, device=device, requires_grad=requires_grad)