ImportError: cannot import name 'NiftiDataset' from 'monai.data' (/usr/local/lib/python3.7/dist-packages/monai/data/__init__.py)

[Sebagam]

Hi all,

When trying to run MONAI on google colab:

ImportError: cannot import name 'NiftiDataset' from 'monai.data' (/usr/local/lib/python3.7/dist-packages/monai/data/init.py)

This issue was not present a few months ago.

Best,
Sebastian

[Nic-Ma]

Hi @Sebagam ,

Thanks for your interest here.
We already changed it to ImageDataset in v0.5.

Thanks.

[Sebagam]

Ok, thanks.
It looks that I have other compatibility issues.
How can I install v0.4?

[Nic-Ma]

Maybe you can try pip install monai==0.4.0?

Thanks.

[Nic-Ma]

But I strongly suggest you to update with MONAI v0.6 as we integrated many enhancements.

Thanks.

[Nic-Ma]

Hi @Sebagam ,

If you check the error log, you can find that the data shape is wrong:
Shape: (1, 166, 256, 256, 1)
Actually your data already has channel dim at the last dim, you should not use AddChannel transform, you can use EnsureChannelFirst or AsChannelFirst transform.

Thanks.

[Sebagam]

Thank you, Nic, Do you know what the following issue mean? RuntimeError: DataLoader worker (pid(s) 1038) exited unexpectedly

…

On Sun, Sep 5, 2021 at 11:52 AM Nic Ma ***@***.***> wrote: Hi @Sebagam <https://github.com/Sebagam> , If you check the error log, you can find that the data shape is wrong: Shape: (1, 166, 256, 256, 1) Actually your data already has channel dim at the last dim, you should not use AddChannel transform, you can use EnsureChannelFirst or AsChannelFirst transform. Thanks. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#2886 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB7M2OO3XFR74B774XAQDTLUAN74TANCNFSM5DLSHURQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[Nic-Ma]

Hi @Sebagam ,

I think maybe it's some PyTorch multi-processing issue? Can you try again with num_workers=0?

Thanks.

[Nic-Ma]

Hi @Sebagam ,

Seems something wrong with your data shape?
Could you please help add EnsureChannelFirst transform at the beginning of your transform chain?

Thanks.

[Sebagam]

I got this error now: ValueError: meta_dict not available, EnsureChannelFirst is not in use.

…

On Wed, Sep 8, 2021 at 8:59 PM Nic Ma ***@***.***> wrote: Hi @Sebagam <https://github.com/Sebagam> , Seems something wrong with your data shape? Could you please help add EnsureChannelFirst transform at the beginning of your transform chain? Thanks. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#2886 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB7M2OKHVIK3RWSIQZGUKXLUA72E3ANCNFSM5DLSHURQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[Nic-Ma]

Hi @Sebagam ,

As you are using array-based transforms instead of dict-based transforms, it may not work well.
Could you please help change to AsChannelFirst transform?

Thanks.

[pip-alireza]

Tried running the Swin-unet-r on BTCV data, and it keeps giving the metadata not available error. I used the code provided for BTCV data, do you know what is causing this? (checked the metadata and it is available. Adding the EnsureChannelFirst/AsChannelFirst did not help but caused the caching dataset to fail.)

Here is the error:
File "/home/alireza/anaconda3/envs/pytorch2/lib/python3.11/site-packages/monai/transforms/utility/array.py", line 204, in call
raise ValueError(msg)
ValueError: Metadata not available and channel_dim=None, EnsureChannelFirst is not in use.

[KumoLiu]

Hi @pip-alireza, you could specify the channel_dim in EnsureChannelFirst.

MONAI/monai/transforms/utility/array.py

Lines 180 to 183 in 90d2acb

	channel_dim: This argument can be used to specify the original channel dimension (integer) of the input array.
	It overrides the `original_channel_dim` from provided MetaTensor input.
	If the input array doesn't have a channel dim, this value should be ``'no_channel'``.
	If this is set to `None`, this class relies on `img` or `meta_dict` to provide the channel dimension.

Something like:

EnsureChannelFirst(channel_dim="no_channel")

Hope it helps, thanks!

[preet012]

I was facing the same issue. Adding EnsureChannelFirstd(keys=["image", "label"],channel_dim='no_channel'), right after LoadImaged in transforms helped.

[mtoan65]

@preet012 Can you send the details of that revised code snippet here?

[Nic-Ma]

I think maybe you can print out the model predictions, see whether it's overfitting?
@wyli @ericspod Could you guys help please share some comments about the training status?

Thanks in advance.

[Sebagam]

I think that I managed to fix it. Do you know how to turn off this output? INFO:nibabel.global:pixdim[0] (qfac) should be 1 (default) or -1; setting qfac to 1 /usr/local/lib/python3.7/dist-packages/monai/data/utils.py:542: UserWarning: Modifying image pixdim from [1.1999999 1. 1. 0. ] to [ 1.19999993 1. 1. 218.40411641] warnings.warn(f"Modifying image pixdim from {pixdim} to {norm}") Thanks

…

On Thu, Sep 9, 2021 at 11:58 AM Nic Ma ***@***.***> wrote: I think maybe you can print out the model predictions, see whether it's overfitting? @wyli <https://github.com/wyli> @ericspod <https://github.com/ericspod> Could you guys help please share some comments about the training status? Thanks in advance. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#2886 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB7M2OKXRI2LFE3ITJEKVZLUBDDQVANCNFSM5DLSHURQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[Nic-Ma]

I think you can turn off all the warnings, referring to:
https://stackoverflow.com/questions/14463277/how-to-disable-python-warnings

Thanks.

[Sebagam]

Hi Nic, Yes, I know that command. The problem is the Nibabel information messages: INFO:nibabel.global:pixdim[0] (qfac) should be 1 (default) or -1; setting qfac to 1 pixdim[0] (qfac) should be 1 (default) or -1; setting qfac to 1 Do you know how to suppress them? Thanks

…

On Thu, Sep 9, 2021 at 8:05 PM Nic Ma ***@***.***> wrote: I think you can turn off all the warnings, referring to: https://stackoverflow.com/questions/14463277/how-to-disable-python-warnings Thanks. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#2886 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB7M2OOISBSPLX6MWAA5TE3UBE4VDANCNFSM5DLSHURQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[Nic-Ma]

Hi @wyli ,

Do you know something about this Nibabel warning?

Thanks in advance.