in main.py, the code is referring to the placeholder_tokens param (main is the only place it is referenced), while the model refers to placeholder_strings.
it seem that currently the training process ignores the --placeholder_tokens cmd-line argument.
That's a bug alright. The config file is using placeholder_strings as expected, so you can either fix the parameter name here or modify the placeholders in the config. I'll fix this either way.
in main.py, the code is referring to the placeholder_tokens param (main is the only place it is referenced), while the model refers to placeholder_strings.
it seem that currently the training process ignores the --placeholder_tokens cmd-line argument.
Am I missing something, or is it simply a typo?
Thanks and keep up the amazing work.