Dependency errors ... unable to start rasa

Rasa Open Source
1

Dear Rasa-Team,

I’m trying now since several days to get your example bot running… I’ve made a German version out of it, include German spacy. Basically just changed the intents.

For whatever dependency reasons, its not working. Whether rasa core xy is not compatible with spacy, or Tensorflow version xy is not compatible with Python, or Jupiter not with rasa core and so on. Cost me so far 40 hours. All your examples are also not working.

At the moment my notebook displays:TypeError: fit() got multiple values for keyword argument ‘epochs’, my console shows a tensorflow error.

Versions: rasa_nlu: 0.13.8 rasa_core: 0.13.0a1 Tensorflow 1.10.0 iPython 6.5.0 Python 3.6.6 (w anaconda) Mac OS 10.14.1 (MacBook Pro, 2.3 Ghz, 2017)

The funny thing: Every example you provide on your website uses a different core, impossible without hours of installation (because the dependencies are of course never working) to run anything smooth.

Please let me know on which exact versions the rasa stack will run on Mac hardware.

thanks.

best, Patrick bot1complete.ipynb (56.6 KB)

2

Hi Patrick,

Thanks for your patience, I know it’s super frustrating when the software doesn’t work.

The solution to your problem is to move this epochs=200 argument into the class initialization of KerasPolicy and remove it from the agent.train command, i.e.

agent = Agent('domain.yml', policies=[MemoizationPolicy(), KerasPolicy(epochs=200), FallbackPolicy()])

# loading our neatly defined training dialogues
training_data = agent.load_data('stories.md')

agent.train(
    training_data,
    validation_split=0.0
)

Thanks for bringing this to our attention though. To prevent it from happening again, where did you find this notebook? I will update it.

3

Hi Tom,

I’m still unable to bring the bot up and working. I’ve now downgraded to spacy 1.9.0, rasa core 0.11.3 and rasa_nlu 0.13.0 and replaced TensorFlow with sklearn. Now I’m facing a new dependency issue, rasa don’t accept the Cymem in spacy - when I’m executing the import of spacy it’s ending up with an error:

ModuleNotFoundError Traceback (most recent call last) in () 3 import rasa_nlu 4 import rasa_core ----> 5 import spacy 6 import sklearn 7 import tensorflow

~/.local/lib/python3.6/site-packages/spacy/init.py in () 6 7 # These are imported as part of the API ----> 8 from thinc.neural.util import prefer_gpu, require_gpu 9 10 from .cli.info import info as cli_info

~/.local/lib/python3.6/site-packages/thinc/neural/init.py in () ----> 1 from ._classes.model import Model

~/.local/lib/python3.6/site-packages/thinc/neural/_classes/model.py in () 9 msgpack_numpy.patch() 10 —> 11 from … import util 12 from …train import Trainer 13 from …ops import NumpyOps, CupyOps

~/.local/lib/python3.6/site-packages/thinc/neural/util.py in () 1 from future import print_function, unicode_literals 2 import numpy ----> 3 from preshed.maps import PreshMap 4 from pathlib import Path 5

cymem.pxd in init preshed.maps()

ModuleNotFoundError: No module named ‘cymem’

The installation of spacy is without any issues, if I go and install Cymem again via pip-install it shows that it’s already installed - so it’s a pure rasa related problem.

at the moment I’m really not sure what it takes to bring rasa up to work, somehow the whole software seems not so mature O_o