SlidingWindowInfererAdapt

[myron]

SlidingWindowInfererAdapt extends SlidingWindowInferer to automatically switch to buffered and then to CPU stitching, when OOM on GPU. It also records a size of such large images to automatically try CPU stitching for the next large image of a similar size. If the stitching 'device' input parameter is provided,
automatic adaptation won't be attempted, please keep the default option device = None for adaptive behavior.
Note: the output might be on CPU (even if the input was on GPU), if the GPU memory was not sufficient.

---

also fixes #6340 by adding one line to the resampling

[myron]

I'm not sure why it fails premerge/flake test. It's fine for me. can someone help plz @mingxin-zheng @wyli

[mingxin-zheng]

Hi @myron , as I replied in the slack channel, this could be caused by using a different version of isort.

I ran ./runtests --autofix and it shows the line 623 and 630 should be removed in monai/inferers/inferer.py

[myron]

@wyli I've updated this PR to include buffered case handling, please check and merge

also mypy complained on the super().call_() lines with error:

mypy
mypy 1.2.0 (compiled: yes)
monai/inferers/inferer.py:509:20: error: "__call__" of "SlidingWindowInferer" gets multiple values for keyword argument "inputs"  [misc]
monai/inferers/inferer.py:509:20: error: "__call__" of "SlidingWindowInferer" gets multiple values for keyword argument "network"  [misc]
monai/inferers/inferer.py:519:23: error: "__call__" of "SlidingWindowInferer" gets multiple values for keyword argument "inputs"  [misc]
monai/inferers/inferer.py:519:23: error: "__call__" of "SlidingWindowInferer" gets multiple values for keyword argument "network"  [misc]

even though, the code looks correct.
I added "type:ignore" , because I couldn't figure it out, may be there is better way.

[wyli]

thanks, I'll test and merge it soon

[wyli]

/build