Warning after RASA upgrade: Sanic tried to use loop.add_signal_handler

Rasa Open Source
1

Hi

I have upgraded my RASA:

Previous the following packages were installed:

rasa==1.4.0
spacy==2.1.8
rasa-sdk==1.4.0
numpy==1.16.0
python-socketio==4.3.1
python-engineio==3.9.3
livereload
pytest==5.2.2
cx_Oracle==7.2.2
python-dateutil==2.8.0
multidict==4.6.1

To start the chatbot I did this:

Terminal 1:

rasa train

Terminal 2:

rasa run --endpoints endpoints.yml --credentials credentials.yml

Terminal 3:

rasa run actions --actions actions

Terminal 4:

livereload

Now I have upgraded to the following version:

rasa==1.5.0
spacy==2.1.8
rasa-sdk==1.5.0
numpy==1.16.0
python-socketio==4.3.1
python-engineio==3.9.3
livereload
pytest==5.2.2
cx_Oracle==7.2.2
python-dateutil==2.8.0
multidict==4.6.1

In terminal 3 I now get the following warning:

2020-01-02 16:43:42 WARNING  sanic.root  - Sanic tried to use loop.add_signal_handler but it is not implemented on this platform.

What is this? How can I solve this?

Note: I get the same warning for the following upgrade:

rasa==1.5.2
spacy==2.1.8
rasa-sdk==1.5.0
numpy==1.16.0
python-socketio==4.3.1
python-engineio==3.9.3
livereload
pytest==5.2.2
cx_Oracle==7.2.2
python-dateutil==2.8.0
multidict==4.6.1

Upgrading even hire will raise more problems. Therefore I want to solve each problem step by step!

2

Am I really the only person with this problem?

3

Try installing sanic==19.9.0.

1 Like
4

Unfortunately, same problem even with installing sanic==19.9.0 :frowning:

5

I’m also having the same warning. I’ve searched the issue but have not found a solution to this. Since it is a warning, I’m letting it be.

1 Like
6

I upgraded to the latest RASA version:

rasa==1.6.0
spacy==2.1.8
rasa-sdk==1.6.0
numpy==1.16.0
python-socketio==4.3.1
python-engineio==3.9.3
livereload
pytest==5.2.2
cx_Oracle==7.2.2
python-dateutil==2.8.0
multidict==4.6.1
sanic==19.9.0

Now, I do not get the Sanic warning (see my first post) anymore. That’s good!

However, if I do rasa train it says:

UserWarning: Number of features (1) for attribute 'text_dense_features' does not match number of
tokens (12). Set 'return_sequence' to true in the corresponding featurizer in order to make use of the features in 'CRFEntityExtractor'.
  f"Number of features ({len(features)}) for attribute "

I am not sure if this is sth. I have to worry about. The problem is, the after training the model and running terminal 2, terminal 3 and terminal 4 commands (see first post) the interface of my chatbot remains empty. I do not get any critical error message, but it looks like my model is empty.

That is strange. Maybe it is related to the warning above?

7

It seems to be an error with the pipeline CRFEntityExtractor. Can you share your config.yml file??

There are later versions both for rasa and rasa_sdk : 1.6.1. Try installing that.

8

The old configurations are still working:

rasa==1.2.3
spacy==2.1.8
rasa-sdk==1.2.0
numpy==1.16.0
python-socketio==4.2.1
python-engineio==3.8.2
livereload
pytest==5.2.2
cx_Oracle==7.2.2

After rasa train and runing terminal 1-4 I have my bot running and talking to me. But I do not want to use these configurations for production, since they are too old.

The new configuration (see grey box in my previous post) show no critical error warning. Only the one I mentioned in my previous post above. You say that it seems like an error with the pipeline CRFEntityExtractor.

Taking the old (working) model and run it with the new configurations doesn’t start the bot properly --> Therefore I think, the problem is not the model itself, but maybe the pipeline that runs the model.

Here is my config.yml file:

language: "de_core_news_sm"             
pipeline: pretrained_embeddings_spacy   #  We started with this one! Used to be called as spacy_sklearn.


policies:
  - name: "KerasPolicy"
    epochs: 300
    max_history: 0
  - name: "MemoizationPolicy"
    max_history: 0
  - name: "FormPolicy"   # To use forms, you also need to include the FormPolicy in your policy configuration file.
  - name: "MappingPolicy"
  - name: "FallbackPolicy"
    # min confidence needed to accept an NLU prediction
    nlu_threshold: 0.3
    # min confidence needed to accept an action prediction from Rasa Core
    core_threshold: 0.3
    # name of the action to be called if the confidence of intent / action
    # is below the threshold
    fallback_action_name: 'action_default_fallback'
9

Can you try the “supervised_embeddings” pipeline? Also update to the latest versions of all packages and then retrain your model

10

It doesen’t work :frowning: I installed all packages as described in the grey box. I did rasa train . No error messages when I run terminal 1 -terminal 4. Everythink looks fine. However, I have no output. The chatbot does not realy start.

This is what Chrome browser says after starting the bot and pressing ctrl+shift+i:

image
image1913Ă—273 38.2 KB

Looks like the problem is related to the browser or livereload or sth. like this…?

11

Yes, there seems to be a problem with the cookie/access settings in the browser which does not allow rasa to run properly. I’m sorry to say that I cannot help you in solving the browser error. You can search for it and after it is solved, you can get back to me, if ther is any problem other than this :grimacing:

Also, I have not worked with livereload and therefore have no knowledge of it.

1 Like