Slightly improve performance in engines/utils.py

[benediktjohannes]

Description

This change probably improves performance slightly (unforetunately, I don't have a local instance of MONAI installed in order to test this, but I'm pretty sure from looking at the code that using [], {} and kwargs_[k] = get_data(v) should be semantically the same and probably slightly better in performance, see also #1423 (credits to @de-sh for the nice fix 👍 ) where at least [] and {} were changed as a refactoring and kwargs[k] = _get_data(v) just seems better to me becuase I think that this saves some time due to dictionary lookups and temporary objects, but I'm not quite sure because it wasn't tested, so please correct me if I'm mistaken).

A few sentences describing the changes proposed in this pull request.

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.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: a03c6105-60d8-400e-9630-fc56d787b3b3

📥 Commits

Reviewing files that changed from the base of the PR and between 540b662 and 94a465b.

📒 Files selected for processing (1)

• monai/engines/utils.py

✅ Files skipped from review due to trivial changes (1)

• monai/engines/utils.py

---

📝 Walkthrough

Walkthrough

Updated container initializations in monai/engines/utils.py within PrepareBatchExtraInput.__call__: replaced list() with [] and dict() with {}. The subsequent population logic (args_.append(...), kwargs_.update(...)/keyword extraction) and return values (tuple(args_), kwargs_) remain unchanged; no control-flow or error-handling changes.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	Title directly matches the main change: replacing list()/dict() calls with []/{}. Concise and specific.
Description check	✅ Passed	Description covers the change purpose and includes checklist. However, it lacks technical clarity about what specific optimization was attempted beyond general claims.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[ericspod]

Here this might make a slight improvement since this method is called a lot, but I would suggest the following would be more efficient still (You'd have to test this):

def __call__(
    self,
    batchdata: dict[str, torch.Tensor],
    device: str | torch.device | None = None,
    non_blocking: bool = False,
    **kwargs: Any,
) -> tuple[torch.Tensor, torch.Tensor, tuple, dict]:
    """
    Args `batchdata`, `device`, `non_blocking` refer to the ignite API:
    https://pytorch.org/ignite/v0.4.8/generated/ignite.engine.create_supervised_trainer.html.
    `kwargs` supports other args for `Tensor.to()` API.
    """
    image, label = default_prepare_batch(batchdata, device, non_blocking, **kwargs)
    args_ = ()
    kwargs_ = {}

    def _get_data(key: str) -> torch.Tensor:
        data = batchdata[key]

        if isinstance(data, torch.Tensor):
            data = data.to(device=device, non_blocking=non_blocking, **kwargs)

        return data

    if isinstance(self.extra_keys, (str, list, tuple)):
        args_ = tuple(_get_data(k) for k in ensure_tuple(self.extra_keys))
    elif isinstance(self.extra_keys, dict):
        kwargs_ = {k: _get_data(v) for k, v in self.extra_keys.items()}

    return cast(torch.Tensor, image), cast(torch.Tensor, label), args_, kwargs_

[benediktjohannes]

Hi, thanks for suggesting this! I'm not quite sure about this yet and I don't have a local instance of MONAI installed at the moment, so testing this would take a while for me because I'm relatively stressed at the moment and I guess it would take some time for me to download and build MONAI locally.

But maybe we can put these suggestions in another PR lateron and for now merge the suggested changes for at least a little improvement, see here https://towardsdatascience.com/no-and-list-are-different-in-python-8940530168b0/

[benediktjohannes]

@ericspod could you maybe open an extra PR for your additional suggestions and merge this for now, so that we can check on the more complex (and test requiring) changes lateron, but already have these safe slight improvements? Thanks!