AttributeError : module tensorflow has no attribute logging

(Bui Duc Huy) #1

Hello, i’m new to rasa. Recently i tried to train rasa with the data in this link:

I typed this in the command line: python -m rasa train -c config.yml --data data/

After i hit enter, an error occured: tf_log_level = tf.logging.ERROR AttributeError: module ‘tensorflow’ has no attribute ‘logging’

In my knowledge, tensorflow 2.0 doesn’t have tf.logging anymore and maybe it’s the cause of my problem. I have installed both tensorflow 1 and tensorflow 2 on my computer. I don’t know how to specify for the program to use a specific tensorflow version and i would like to receive some help on this matter. Thanks for your time.

(Tanja Bergmann) #2

Rasa requires tensorflow version 1.13.1. I would recommend using a virtual environment, such as Virtualenv — virtualenv 16.6.1 documentation. Check out Install TensorFlow with pip  |  TensorFlow on how to install tensorflow via pip in a virtualenv.

1 Like
(Bui Duc Huy) #3

Thanks for your support. I installed virtualenv and then installed tensorflow 1.13 with it. I just want to be sure that after i activate the virtual enviroment in command line: myenvpath activate

I can use the command below and it will use the tensorflow version in my virtual enviroment or i have to do some config first: rasa train -c config.yml --data/ nlp

(Tanja Bergmann) #4

To activate your virtualenv you need to execute source <path-to-env-folder>/bin/activate. You also need to install rasa in that virtualenv otherwise you will not be able to execute your command. But in general only the packages inside the virtualenv are used when you execute something inside the virtualenv. Btw. the command should look like: rasa train nlp -c config.yml --data/