SwinUnet deprecation error and when correct TypeError over img_size

[joaomamede]

Describe the bug
Img_size argument is required, but not allowed at the same time

To Reproduce
print_config()

MONAI version: 1.5.dev2450
Numpy version: 1.26.4
Pytorch version: 2.5.1
MONAI flags: HAS_EXT = False, USE_COMPILED = False,

device = torch.device("cuda:0")
model = SwinUNETR(
    img_size=(512,512,64),#this is not an argument anymore
    spatial_dims=3,
    in_channels=2,
    out_channels=3,
    feature_size=48,
    drop_rate=0.0,
    attn_drop_rate=0.0,
    dropout_path_rate=0.0,
    use_checkpoint=True
    
).to(device)

DeprecatedError: Argument img_size was removed in version 1.5. The img_size argument is not required anymore and checks on the input size are run during forward().

But

model = SwinUNETR(
    # img_size=(512,512,64),#this is not an argument anymore
    spatial_dims=3,
    in_channels=2,
    out_channels=3,
    feature_size=48,
    drop_rate=0.0,
    attn_drop_rate=0.0,
    dropout_path_rate=0.0,
    use_checkpoint=True
    
).to(device)

TypeError: missing a required argument: 'img_size'

Expected behavior
Img_size is no longer needed but we have type error when not present and DeprecationError when present.

Environment

anaconda linux ubuntu 24.04

python -c "import monai; monai.config.print_debug_info()"

================================
Printing MONAI config...

MONAI version: 1.5.dev2450
Numpy version: 1.26.4
Pytorch version: 2.5.1
MONAI flags: HAS_EXT = False, USE_COMPILED = False, USE_META_DICT = False
MONAI rev id: 0726cee7ba6c1b31f34d54fc26fe6b8db5f5c79c
MONAI file: /home//anaconda3/envs/monai/lib/python3.9/site-packages/monai/init.py

Optional dependencies:
Pytorch Ignite version: 0.4.11
ITK version: 5.4.0
Nibabel version: 5.3.2
scikit-image version: 0.22.0
scipy version: 1.13.1
Pillow version: 11.0.0
Tensorboard version: 2.18.0
gdown version: 5.2.0
TorchVision version: 0.20.1
tqdm version: 4.67.1
lmdb version: 1.5.1
psutil version: 6.1.0
pandas version: 2.2.3
einops version: 0.8.0
transformers version: 4.40.2
mlflow version: 2.19.0
pynrrd version: 1.1.1
clearml version: 1.17.0rc0

For details about installing the optional dependencies, please visit:
https://docs.monai.io/en/latest/installation.html#installing-the-recommended-dependencies

================================
Printing system config...

System: Linux
Linux version: Ubuntu 24.04.1 LTS
Platform: Linux-6.8.0-51-generic-x86_64-with-glibc2.39
Processor: x86_64
Machine: x86_64
Python version: 3.9.21
Process name: python
Command: ['/home/jmamede/anaconda3/envs/monai/bin/python', '-m', 'ipykernel_launcher', '--f=/home/jmamede/.local/share/jupyter/runtime/kernel-v3a91b9ef6ea6c88b8a84ed6c43240878d988e2a4a.json']
Open files: [popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/exthost.log', fd=43, position=5315, mode='a', flags=33793), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/extensionTelemetry.log', fd=44, position=490, mode='a', flags=33793), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/vscode.github/GitHub.log', fd=45, position=219, mode='a', flags=33793), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/vscode.github-authentication/GitHub Authentication.log', fd=46, position=723, mode='a', flags=33793), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/vscode.git/Git.log', fd=49, position=876031, mode='a', flags=33793), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/output_logging_20250112T171631/1-Jupyter.log', fd=50, position=503828, mode='a', flags=33793), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/ms-python.debugpy/Python Debugger.log', fd=52, position=0, mode='a', flags=33793), popenfile(path='/home/jmamede/.ipython/profile_default/history.sqlite', fd=53, position=0, mode='r+', flags=688130), popenfile(path='/home/jmamede/.ipython/profile_default/history.sqlite', fd=54, position=0, mode='r+', flags=688130), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/output_logging_20250112T171631/2-Jupyter Server Console.log', fd=62, position=0, mode='a', flags=33793), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/output_logging_20250112T171631/3-Python Test Log.log', fd=64, position=0, mode='a', flags=33793), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/ms-python.python/Python Language Server.log', fd=70, position=14458, mode='a', flags=33793), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/ms-python.python/Python.log', fd=81, position=28422, mode='a', flags=33793), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/output_logging_20250112T171631/4-Tensorboard.log', fd=93, position=5868, mode='a', flags=33793), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/ms-python.python/Python Locator.log', fd=95, position=38210, mode='a', flags=33793), popenfile(path='/usr/share/code/v8_context_snapshot.bin', fd=103, position=0, mode='r', flags=32768), popenfile(path='/home/jmamede/.config/Code/logs/20250112T171623/window1/exthost/vscode.json-language-features/JSON Language Server.log', fd=114, position=0, mode='a', flags=33793)]
Num physical CPUs: 24
Num logical CPUs: 48
Num usable CPUs: 48
CPU usage (%): [2.8, 3.0, 3.0, 5.0, 3.5, 4.3, 3.6, 3.5, 3.1, 3.3, 3.5, 3.1, 3.2, 2.7, 4.5, 4.3, 3.5, 3.7, 4.9, 3.5, 3.1, 3.5, 3.1, 2.9, 3.6, 2.9, 2.9, 6.4, 3.0, 5.5, 3.1, 4.3, 3.1, 3.5, 3.2, 3.3, 3.1, 2.6, 3.6, 3.4, 3.8, 6.9, 3.7, 3.5, 3.3, 2.8, 3.0, 3.8]
CPU freq. (MHz): 2279
Load avg. in last 1, 5, 15 mins (%): [4.7, 5.3, 7.3]
Disk usage (%): 78.8
Avg. sensor temp. (Celsius): UNKNOWN for given OS
Total physical memory (GB): 125.6
Available memory (GB): 7.5
Used memory (GB): 115.7

================================
Printing GPU config...

Num GPUs: 1
Has CUDA: True
CUDA version: 12.4
cuDNN enabled: True
NVIDIA_TF32_OVERRIDE: None
TORCH_ALLOW_TF32_CUBLAS_OVERRIDE: None
cuDNN version: 90100
Current device: 0
Library compiled for CUDA architectures: ['sm_50', 'sm_60', 'sm_61', 'sm_70', 'sm_75', 'sm_80', 'sm_86', 'sm_90']
GPU 0 Name: NVIDIA RTX A4000
GPU 0 Is integrated: False
GPU 0 Is multi GPU board: False
GPU 0 Multi processor count: 48
GPU 0 Total memory (GB): 15.7
GPU 0 CUDA capability (maj.min): 8.6

[KumoLiu]

Hi @joaomamede, may I ask how did you install the monai?

[joaomamede]

Hi @joaomamede, may I ask how did you install the monai?

Good question I think I did pip install. It was a while back.

If I remove the deprecation type error in the code the swin runs.

[KumoLiu]

Good question I think I did pip install. It was a while back.

If I remove the deprecation type error in the code the swin runs.

Hi @joaomamede, yes. I know it's a bit of a mismatch, because we want to tell the user something through the deprecate message, but because the img_size doesn't have a default value, if we add a default value for it and change the position of the parameter, it may cause some users get error who specifying the parameters in order. Theoretically, if your monai version doesn't start with 1.5, there won't be a problem.

[joaomamede]

Hi @joaomamede, may I ask how did you install the monai?

I did pip install monai-weekly[all] with a fresh anaconda environment with python 3.9.21.

if I remove the deprecation type error in the code it runs, but we need to add a value to img_size (I believe it does nothing) because I added (512,512,64) as option in images that were 256,256,64 and it ran fine.

[KumoLiu]

Hi @joaomamede, yes. I get the point.
It may not easy to add a default value since if we add a default value for it, then we need to change the position of the parameter, it may cause some users get error who specifying the parameters in order.

Do you have any opinion? cc @ericspod

[ericspod]

Hi @joaomamede, yes. I get the point. It may not easy to add a default value since if we add a default value for it, then we need to change the position of the parameter, it may cause some users get error who specifying the parameters in order.

Do you have any opinion? cc @ericspod

We're prepping the code for the 1.5 release so now is the time to remove things that are marked deprecated. The decorator says that the it is to be removed in 1.5 so we can do that now, I would move the argument to the end of the argument list for now, provide it a default value of None, and remove any code which uses it. That will leave the deprecation error in place for anyone using weekly releases but will allow you to instantiate the class correctly. If this breaks things in tutorials we'll find out about it quickly.

[joaomamede]

If you want to remove the argument (as it is deprecated) isn't it simpler to simply remove the argument from the "needed arguments list"?

The problem is that it does not run because it outputs a "argument required" error.

I fixed it the other way around, to remove the deprecated argument error. because it was simply commenting those 3 or 4 lines. But as it is, the swinunetR is simply not running.

As in, if you want to add the argument as optional it could be passed in kw_args?