ValueError while training the model for Brain Tumor Segmentation

[project-monai]

ValueError while training the model for Brain Tumor Segmentation

[Nic-Ma]

Hi @pfdamasceno ,

@yiheng-wang-nv is updating all the examples and tutorials for 0.6 release: #246.

Thanks.

[Nic-Ma]

Hi @sekh77 ,

Thanks for your bug report, this is compatible issue with latest MONAI code, I will fix it tomorrow.
Please install monai==0.3.0 as it's the stable version.

Thanks.

[Nic-Ma]

Hi @yiheng-wang-nv ,

I am afraid maybe your previous mdification to not_nans also breaks other tutorials, could you please help double confirm all the tutorials and examples which used DiceMetric?
Fo these affected tutorials, you need to add the logic to install MONAI from the master code instead of the release version.
You can refer to the second cell of this tutorial:
https://github.com/Project-MONAI/tutorials/blob/master/2d_classification/mednist_tutorial.ipynb

Thanks.

[sekh77]

Hi @Nic-Ma -

This is the config report, and I see that I'm already on 0.3.0.

MONAI version: 0.3.0+29.g56dd27a
Python version: 3.7.7 (default, May 7 2020, 21:25:33) [GCC 7.3.0]
OS version: Linux (5.4.0-1031-azure)
Numpy version: 1.18.1
Pytorch version: 1.6.0
MONAI flags: HAS_EXT = True, USE_COMPILED = False

Optional dependencies:
Pytorch Ignite version: 0.4.2
Nibabel version: 3.2.0
scikit-image version: 0.16.2
Pillow version: 7.2.0
Tensorboard version: NOT INSTALLED or UNKNOWN VERSION.
gdown version: 3.11.0
TorchVision version: 0.8.1
ITK version: 5.1.1
tqdm version: 4.51.0

Thanks,
Sekhar H.

[Nic-Ma]

Hi @sekh77 ,

This is not exact 0.3.0, please try to use pip install monai==0.3.0.
Thanks.

[sekh77]

Ok, sure! Thank you.

Sekhar H.

[sekh77]

Hi @Nic-Ma -

I reinstalled MONAI to 0.3.0. But I still see exactly the same error after epoch 2. Here's my new config report:

MONAI version: 0.3.0
Python version: 3.7.7 (default, May 7 2020, 21:25:33) [GCC 7.3.0]
OS version: Linux (5.4.0-1031-azure)
Numpy version: 1.18.1
Pytorch version: 1.6.0
MONAI flags: HAS_EXT = False, USE_COMPILED = False

Optional dependencies:
Pytorch Ignite version: 0.4.2
Nibabel version: 3.2.0
scikit-image version: 0.16.2
Pillow version: 7.2.0
Tensorboard version: NOT INSTALLED or UNKNOWN VERSION.
gdown version: 3.11.0
TorchVision version: 0.8.1
ITK version: 5.1.1
tqdm version: 4.51.0

Thanks,
Sekhar H.

[sekh77]

Here's the error after reinstalling to 0.3.0:

epoch 2 average loss: 0.9216

ValueError Traceback (most recent call last)
in
49 val_outputs = post_trans(val_outputs)
50 # compute overall mean dice
---> 51 value, not_nans = dice_metric(y_pred=val_outputs, y=val_labels)
52 not_nans = not_nans.item()
53 metric_count += not_nans

ValueError: not enough values to unpack (expected 2, got 1)

Thanks,
Sekhar H.

[Nic-Ma]

Hi @sekh77 ,

Sorry I didn't make it clear, you should also update this tutorial repo to use tag 0.3.0.

Thanks.

[sekh77]

How do I do that? I downloaded the file - brats_segmentation_3d.ipynb - locally into my VM, and running it using my Jupyter Notebook.

Thanks,
Sekhar H.

[Meddebma]

Hi,

I'm using google colab and I tried to make the same training and evaluation with my own spleen dataset, but I've got the same problem after the 2nd epoch.
I reinstalled monai with the Version 0.3.0 but I still get this error after the 2nd epoch:
"ValueError: not enough values to unpack (expected 2, got 1)"
Is it already debugged or what should I wait?

Thanks
Aymen

[sekh77]

Hi @Nic-Ma -

Is this issue already fixed? Can I use the latest version?

Thanks,
Sekhar H.

[Nic-Ma]

Hi @sekh77 ,

Yes, I think so, you can use either the latest MONAI and latest tutorial, or MONAI 0.3 and tutorial 0.3.

Thanks.

[Meddebma]

Hi @sekh77,

did it work for you? I still have the same problem after installing MONAI 0.3 :/

Thanks,
Aymen

[yiheng-wang-nv]

Hi @sekh77,

did it work for you? I still have the same problem after installing MONAI 0.3 :/

Thanks,
Aymen

Hi, which tutorial version you used under MONAI 0.3? Could you please attach the error here? The MONAI version as well as the tutorial version should be consistent.

[Meddebma]

Hi @yiheng-wang-nv

here is the link:
https://github.com/Meddebma/pyradiomics/blob/master/Spleen_dict.ipynb

thanks!

[yiheng-wang-nv]

Hi @yiheng-wang-nv

here is the link:
https://github.com/Meddebma/pyradiomics/blob/master/Spleen_dict.ipynb

thanks!

I see, the code is the updated version that the dice metric will return two values. Please install the latest MONAI and run your notebook.

[Meddebma]

Hi @yiheng-wang-nv
here is the link:
https://github.com/Meddebma/pyradiomics/blob/master/Spleen_dict.ipynb
thanks!

I see, the code is the updated version that the dice metric will return two values. Please install the latest MONAI and run your notebook.

I did it in the beginning of code with: %pip install monai==0.3.0
should l do it differently?

Thanks @yiheng-wang-nv

[yiheng-wang-nv]

Hi @yiheng-wang-nv

here is the link:
https://github.com/Meddebma/pyradiomics/blob/master/Spleen_dict.ipynb

thanks!

Please use the commented line instead:
pip install git+https://github.com/Project-MONAI/MONAI#egg=MONAI

[Meddebma]

Hi @yiheng-wang-nv
here is the link:
https://github.com/Meddebma/pyradiomics/blob/master/Spleen_dict.ipynb
thanks!

Please use the commented line instead:
pip install git+https://github.com/Project-MONAI/MONAI#egg=MONAI

I just did it, I still have the same issue :/

https://github.com/Meddebma/pyradiomics/blob/master/Spleen_dict.ipynb

[yiheng-wang-nv]

Hi @yiheng-wang-nv
here is the link:
https://github.com/Meddebma/pyradiomics/blob/master/Spleen_dict.ipynb
thanks!

Please use the commented line instead:
pip install git+https://github.com/Project-MONAI/MONAI#egg=MONAI

I just did it, I still have the same issue :/

https://github.com/Meddebma/pyradiomics/blob/master/Spleen_dict.ipynb

According to the log in your link:
Requirement already satisfied: MONAI from git+https://github.com/Project-MONAI/MONAI#egg=MONAI in /usr/local/lib/python3.6/dist-packages (0.3.0)
Since your device already has MONAI 0.3.0, the latest version would not be installed.
Could you please uninstall MONAI 0.3.0 first, and then install the latest one?

[Meddebma]

Hi @yiheng-wang-nv
It worked very well thank you :)
Only the plotting of the GIF did not work, should I change something to see the visual output?
Thanks!

[pfdamasceno]

Hi, this seems to still be an issue for MONAI>=monai-0.6.0. Is that correct?