Paths when running training

[bbieniek]

Hi,

While running training via run_bilstmcrf.py I ran it the following way:
Installed deidentify via pip
Checked out the code from github to have run_bilstmcrf.py

When running the run_bilstmcrf.py script, this resulted in issues with the paths for the training set as the scripts use base path in most cases.

Is this the correct way of running the training?

[jantrienes]

Hi @bbieniek,

thanks for reaching out. The recommended way to train a model with the run_bilstmcrf.py script is to 1) clone the repo, 2) create the virtual environment (see here), and 3) train the model according to these steps. No pip install is needed in this case, because you are working from the sources itself.

When running the run_bilstmcrf.py script, this resulted in issues with the paths for the training set as the scripts use base path in most cases.

All scripts assume that the training data resides in data/corpus/. Could you perhaps share the specific error you received? That would make it easier to analyze the problem. Thank you.

Best,
Jan

[jantrienes]

I'm closing this for now. Please let me know if you have additional questions.

[bbieniek]

Thanks Jan for your response. That helped!