trainer can't work

[Sakura4036]

when I try to run this trainer , .next() can't run

-> "AttributeError: 'dict_keyiterator' object has no attribute 'next'"

I use:
python=3.7
torch=11.3

This method has been deleted in python 3?
Why not use next(iter)?

or what should I do to fix it?

[Baijiong-Lin]

Please change torch from version 1.13 to 1.8. It can run.

[Sakura4036]

OK, thanks. I can try it later.

[Sakura4036]

I change env to torch 1.8.0, still same error!

[Baijiong-Lin]

What is your run command? Can you run the example like office-31 successfully?

[Sakura4036]

No, I can't , same error

[Sakura4036]

[Baijiong-Lin]

Office-31 is a multi-input MTL problem where each task has its own input data, thus multi_input should be set as True as follows:
python train_office.py --multi_input

Please refer to Docs and README for more detail.

[Sakura4036]

it can run!

I checked my Dataset and Dataloader, it's similar to the dataset used in LibMTL , it seems no problem

Then what other problems may cause this error？

[Baijiong-Lin]

Do you mean there is still an error when applying LibMTL to your own dataset? Have you defined your own Dataset and Dataloader correctly? Please refer to Docs for the detail about applying LibMTL to a new dataset.

[Sakura4036]

OK，thanks， I will try it later