
Rasa version:1.4.1
Python version: 2.7.4
Operating system (windows, osx, ...): Ubuntu
This issue appears to stem from the requirement that the argument persist_nlu_data exists when calling the training function, but it is not being set to None when calling training via interactive mode.
Can be corrected by adding args.persist_nlu_data = None to the top of the interactive method.
I would think that interactive_core would require the same amendment.
Hi @nicolasfarina, thanks for bringing this up. This issue has been fixed in https://github.com/RasaHQ/rasa/commit/525a2443ec6863081ec7774a1a707272f687c944, but the fix hasn't been released yet. It will be included in 1.4.2. I'll leave this issue open until that's released.
Most helpful comment
This issue appears to stem from the requirement that the argument
persist_nlu_dataexists when calling the training function, but it is not being set to None when calling training via interactive mode.https://github.com/RasaHQ/rasa/blob/33daa89648222fec73c276eda8f1d4b61e985619/rasa/cli/interactive.py#L54-L64
Can be corrected by adding
args.persist_nlu_data = Noneto the top of the interactive method.I would think that
interactive_corewould require the same amendment.