在 VirtualBox Ubuntu 18.04 上使用 Rasa/Tensorflow 时出错

Question

I have a laptop with Windows 10 Pro with Python 3.6.

This is my pip freeze on Windows:

numpy==1.19.5
absl-py==0.10.0
aiofiles==0.6.0
aiohttp==3.7.3
APScheduler==3.6.3
astor==0.8.1
async-generator==1.10
async-timeout==3.0.1
attrs==20.3.0
bidict==0.21.2
blis==0.7.4
boto3==1.16.50
botocore==1.19.50
bz2file==0.98
cached-property==1.5.2
cachetools==4.2.0
catalogue==1.0.0
certifi==2020.12.5
cffi==1.14.4
chardet==3.0.4
click==7.1.2
cloudpickle==1.6.0
colorama==0.4.4
colorclass==2.2.0
coloredlogs==10.0
colorhash==1.0.3
ConfigArgParse==1.2.3
contextvars==2.4
cryptography==3.3.1
cycler==0.10.0
cymem==2.0.5
dataclasses==0.8
decorator==4.4.2
dill==0.3.3
dnspython==1.16.0
docopt==0.6.2
dopamine-rl==3.0.1
fbmessenger==6.0.0
Flask==1.1.2
future==0.18.2
gast==0.2.2
gevent==1.5.0
gin-config==0.4.0
google-api-core==1.24.1
google-api-python-client==1.12.8
google-auth==1.24.0
google-auth-httplib2==0.0.4
google-pasta==0.2.0
googleapis-common-protos==1.52.0
greenlet==0.4.17
grpcio==1.34.0
gunicorn==20.0.4
gym==0.18.0
h11==0.8.1
h2==3.2.0
h5py==3.1.0
hpack==3.0.0
hstspreload==2020.12.22
httpcore==0.3.0
httplib2==0.18.1
httptools==0.1.1
httpx==0.9.3
humanfriendly==9.1
hyperframe==5.2.0
idna==2.10
idna-ssl==1.1.0
immutables==0.14
importlib-metadata==3.3.0
importlib-resources==4.1.1
itsdangerous==1.1.0
Jinja2==2.11.2
jmespath==0.10.0
jsonpickle==1.4.2
jsonschema==3.2.0
Keras-Applications==1.0.8
Keras-Preprocessing==1.1.2
kfac==0.2.0
kiwisolver==1.3.1
Markdown==3.3.3
MarkupSafe==1.1.1
matplotlib==3.3.3
mattermostwrapper==2.2
mesh-tensorflow==0.1.18
mpmath==1.1.0
multidict==4.6.1
murmurhash==1.0.5
networkx==2.3
oauth2client==4.1.3
opencv-python==4.5.1.48
opt-einsum==3.3.0
packaging==19.2
pika==1.0.1
Pillow==7.2.0
plac==1.1.3
preshed==3.0.5
promise==2.3
prompt-toolkit==3.0.9
protobuf==3.14.0
pyasn1==0.4.8
pyasn1-modules==0.2.8
pycparser==2.20
pydot==1.4.1
pyglet==1.5.0
PyJWT==1.7.1
pykwalify==1.7.0
pymongo==3.11.2
pyparsing==2.4.7
pypng==0.0.20
pyreadline==2.1
pyrsistent==0.17.3
python-crfsuite==0.9.7
python-dateutil==2.8.1
python-engineio==4.0.0
python-socketio==5.0.4
python-telegram-bot==11.1.0
pytz==2019.3
PyYAML==5.3.1
questionary==1.9.0
rasa==1.6.0
rasa-sdk==1.6.1
redis==3.3.11
requests==2.25.1
requests-async==0.5.0
requests-toolbelt==0.9.1
rfc3986==1.4.0
rocketchat-API==0.6.36
rsa==4.6
ruamel.yaml==0.15.100
s3transfer==0.3.3
sanic==19.9.0
Sanic-Cors==0.9.9.post1
sanic-jwt==1.5.0
Sanic-Plugins-Framework==0.9.4.post1
scikit-learn==0.20.4
scipy==1.5.4
six==1.15.0
sklearn-crfsuite==0.3.6
slackclient==1.3.2
sniffio==1.2.0
spacy==2.3.5
SQLAlchemy==1.3.22
srsly==1.0.5
sympy==1.7.1
tabulate==0.8.7
tensor2tensor==1.14.1
tensorboard==1.15.0
tensorflow==1.15.3
tensorflow-datasets==3.1.0
tensorflow-estimator==1.15.1
tensorflow-gan==2.0.0
tensorflow-hub==0.11.0
tensorflow-metadata==0.26.0
tensorflow-probability==0.7.0
termcolor==1.1.0
terminaltables==3.1.0
thinc==7.4.5
tqdm==4.55.1
twilio==6.50.1
typing-extensions==3.7.4.3
tzlocal==2.1
uritemplate==3.0.1
urllib3==1.26.2
wasabi==0.8.0
wcwidth==0.2.5
webexteamssdk==1.6
websocket-client==0.54.0
websockets==8.1
Werkzeug==1.0.1
wrapt==1.12.1
yarl==1.6.3
zipp==3.4.0

The project I'm working on uses Rasa 1.6.0 and with these requirements, everything runs as expected on Windows.

However, for a specific task, I need to be on Ubuntu and so I installed Ubuntu 18.04 on a VirtualBox 6.1 machine, installed python and pip and run pip install -r requirements.txt command to install the exact same requirements on that machine, inside a virtual environment.

After this, if I try to run rasa train on a shell or import tensorflow as tf on a python console, an error appears:

Illegal instruction (core dumped)

Any ideias on what the cause for this error could be? I saw some people saying it could be some incompatibility of tensorflow with the architecture of the processor being used, however in my case that should not be the problem since everything works on Windows (right?).

I also tried to uninstall tensorflow and followed some online tutorials to Build from source but now, when I run rasa --version , tensorflow is no longer found:

$ rasa --version
Traceback (most recent call last):
  File "/home/<USER>/Desktop/venv_rasa/bin/rasa", line 8, in <module>
    sys.exit(main())
  File "/home/<USER>/Desktop/venv_rasa/lib/python3.6/site-packages/rasa/__main__.py", line 69, in main
    set_log_level(log_level)
  File "/home/<USER>/Desktop/venv_rasa/lib/python3.6/site-packages/rasa/utils/common.py", line 68, in set_log_level
    update_tensorflow_log_level()
  File "/home/<USER>/Desktop/venv_rasa/lib/python3.6/site-packages/rasa/utils/common.py", line 108, in update_tensorflow_log_level
    import tensorflow as tf
ModuleNotFoundError: No module named 'tensorflow'

But, inside a Python console, I can import tensorflow but if I try so access anything, other errors are shown:

>>> import tensorflow as tf
>>> tf.__version__
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: module 'tensorflow' has no attribute '__version__'
>>> 
>>> mnist = tf.keras.datasets.mnist
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: module 'tensorflow' has no attribute 'keras'

Am I missing anything? Am I doing something wrong?

Thanks in advance!

Answer 1

I saw some people saying it could be some incompatibility of tensorflow with the architecture of the processor being used, however in my case that should not be the problem since everything works on Windows (right?).

I think this may in fact be your problem. When you're using VirtualBox, you're no longer in the same environment as on your windows machine, and you could be running into incompatibilities there that were not happening in your local windows environment.

The fact that the module you built from source is not found, is most likely because it isn't in your PYTHONPATH .

I would recommend using a virtual environment and following eg these instructions for installing tensorflow on Ubuntu, since that's what your VirtualBox is running: https://linuxize.com/post/how-to-install-tensorflow-on-ubuntu-18-04/