Question while training

[xiaozhou]

when I do this step:
python continue_train.py --data-dir dataset_location --save-dir save_destination --train-size train_set_size --use-count
There is an error that I have no file "final_categories_frequency"

How can I get this file?

[xiaozhou]

Neither do I find this file in github or downloaded files

[Kai Wang]

Please double check your --data-dir argument, this should point to one of the datasets (bedroom/living/office) generated by create_data.py, instead of the original suncg_data.
final_categories_frequency should be automatically created in the process

[xiaozhou]

Thanks a lot!
I succuessfully deal with that.

[xiaozhou]

When I do " python create_data.py "
There is an error:



I wonder if there are some bugs in code or it's my mistake

This is my environment:
win7
Anaconda
python3.6.6
pytorch0.4.1
with all python libraryin requirements.txt
cuda9.0

[Kai Wang]

Should be a problem with Windows only. It's a problem with some code I copied elsewhere in order to suppress the annoying C++ warnings, just comment out the with stdout_redirected() line in the respective file and you should be all set.
An actual fix might come later...