Rasa: 2.0.0rc4 clean install: Conflicting dependencies and VisibleDeprecationWarnings

[N-Olbert]

Rasa version: 2.0.0rc4 on Windows 10 1909

Python version: 3.8

After doing a fresh install in a clean venv, numpy shows some VisibleDeprecationWarnings during training a model:

lib\site-packages\rasa\nlu\classifiers\diet_classifier.py:648: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
model_data.add_features(key, sub_key, [np.array(_features)])

lib\site-packages\rasa\utils\tensorflow\model_data.py:587: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
final_data[key][sub_key].append(np.concatenate(np.array(f)))

lib\site-packages\rasa\utils\tensorflow\model_data_utils.py:197: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
attribute_features = {MASK: [np.array(attribute_masks)]}

Also note that a clean install is producing dependency conflicts:

pip check
tensorflow 2.3.1 has requirement numpy<1.19.0,>=1.16.0, but you have numpy 1.19.2.

[sara-tagger]

Thanks for raising this issue, @JustinaPetr will get back to you about it soon✨

Please also check out the docs and the forum in case your issue was raised there too 🤗

[dakshvar22]

Tested and the warnings are visible in Rasa==2.0.0 as well because of the numpy version being installed is 1.19.2 while the maximum version supported by tensorflow is 1.18.5. Will create a PR to pin the numpy version.

[alwx]

tensorflow==2.3.1 explicitly specifies that the support version of numpy should be <1.19.0 thus running poetry update updates the dependencies and includes the following line to the lockfile:

numpy = ">=1.16.0,<1.19.0"

I will prepare a PR that updates the dependencies.