KeyError: 'Accessing unknown key in a struct: wd'

[Taemra]

Hi @felixkreuk .

Thanks for your codes. I have been successfully used predict.py to get the phoneme boundaries, but I would like to train a new model on my own data so that the result may get better. I have configured the timit_path in the conf.yaml to my own datasets which has three sub directory (train/val/test), and did not change anything else in documents.

Then using "python main.py" or "python main.py data=timit", I got the following error reports:

`Traceback (most recent call last): File "main.py", line 77, in main()

File "D:\anaconda\lib\site-packages\hydra\main.py", line 20, in decorated_main run_hydra(

File "D:\anaconda\lib\site-packages\hydra_internal\utils.py", line 171, in run_hydra hydra.run(

File "D:\anaconda\lib\site-packages\hydra_internal\hydra.py", line 82, in run return run_job(

File "D:\anaconda\lib\site-packages\hydra\plugins\common\utils.py", line 109, in run_job ret.return_value = task_function(task_cfg)

File "main.py", line 28, in main cfg.wd = os.getcwd()

File "D:\anaconda\lib\site-packages\omegaconf\dictconfig.py", line 73, in setattr self.setitem(key, value)

File "D:\anaconda\lib\site-packages\omegaconf\dictconfig.py", line 40, in setitem raise KeyError(

KeyError: 'Accessing unknown key in a struct : wd'`

The problem may be very simple since I am not familiar with code-writing. But really hope to know how to use the main.py to train a model based on my own data.

[felixkreuk]

Hi @Taemra, could you please make sure you are using hydra version hydra-core==0.11.3? You can do that by running pip intall hydra-core==0.11.3. Let me know if this solves your error. Thanks.

[Taemra]

Hi Felix. It helps, the error is solved. Thank you very much. :D