RASA X - Quick Install Method (Fresh installation) model not training

[Deprecated] Rasa X Community Edition
1

Hi Team, To do a fresh installation of RASA-X (via Quick Install method) with version 0.29.1, I follow the steps suggested in quick install. But when the docker images are brought up, all the containers are Up and running, but when i train a model, it is not training. On further check, i found that the rasa-production container, has the below logs:

2020-06-15 14:10:45 ERROR pika.adapters.utils.io_services_utils - Socket failed to connect: <socket.socket fd=21, family=AddressFamily.AF_INET, type=2049, proto=6, laddr=(‘172.18.0.7’, 55204)>; error=111 (Connection refused) 2020-06-15 14:10:45 ERROR pika.adapters.utils.connection_workflow - TCP Connection attempt failed: ConnectionRefusedError(111, ‘Connection refused’); dest=(<AddressFamily.AF_INET: 2>, <SocketKind.SOCK_STREAM: 1>, 6, ‘’, (‘172.18.0.2’, 5672)) 2020-06-15 14:10:45 ERROR pika.adapters.utils.connection_workflow - AMQPConnector - reporting failure: AMQPConnectorSocketConnectError: ConnectionRefusedError(111, ‘Connection refused’)

I have tried reinstalling the server, on a fresh VM too. I also tried installing an older version (0.25.2), but i am facing the same challenge there.

Please advise.

Best Regards, Ankit

2

Hey @ankit, you can safely ignore those logs, so long as they aren’t persisting over time (they happen during server startup when the Rasa X server is not yet running). If you are having trouble with training, can you check the rasa-worker logs for any errors? This is the container that does the model training.

3

Hi @erohmensing, Good to hear this from you. Actually, the model training on the Production Server is not happening and as soon as i go ahead an check the logs of the Production Container, I read these logs. Issue: Training the Model with the Custom Action Server integrated, but as soon as model is trained, the model does not reflect. How to debug this issue? Best Regards, Ankit

4

Are you sure you checked the right container? it should be rasa-worker not rasa-production.

as soon as i go ahead an check the logs of the Production Container, I read these logs. Issue: Training the Model with the Custom Action Server integrated, but as soon as model is trained, the model does not reflect.

This is a log message? That doesn’t seem right – can you post a traceback of what you’re seeing?

5

Hi @erohmensing,

Thank you for helping me with this issue. I may have shared multiple issues, but I am unable to do RCA given my limited knowledge of RASAX. I will share the entire scenario in detail below, covering all the observations. Here are the steps:

  1. When starting up the containers using “docker-compose up -d”, i get no issues and all docker containers are up. (image: docker

    container_status
    container_status1680×1050 120 KB

  2. Issue:When I login into RASAX and upload nlu, stories and domain files and click on Train, model starts to train.

    train_start
    train_start1680×1050 137 KB

Post training, I get a successful message that the model is existing and no changes were made.! train_complete|690x431

Whereas , in my models directory, there are no models:

no_model
no_model1680×1050 161 KB

  1. Issue: I noticed one thing that when I click on train_model, I get Status 500 in the network tab.! Error_captured_via_network_while_inspect_element|690x431

  2. Issue: I have done a fresh installation of RASAX and i am able to login; still I am get an “Authentication Token invalid” icon when I go on the login screen of the RASAX.

    rasax_login_screen
    rasax_login_screen1680×1050 103 KB

Appreciate your quick response :slight_smile:

Best Regards, Ankit

6

Thanks for the description! It’s as i thought, you’re having a training error (the 500 error). The “nothing has changed” modal is coming up incorrectly. To debug why it isn’t training (and therefore no model shows), you’ll want to check the training container’s logs with sudo docker-compose logs rasa-worker.

7

Hi @erohmensing,

Here are the logs from the rasa-worker:

rasa-worker_1 | /opt/venv/lib/python3.7/site-packages/rasa/core/brokers/pika.py:294: FutureWarning: Your Pika event broker config contains the deprecated queue key. Please use the queues key instead. rasa-worker_1 | self.queues = self._get_queues_from_args(queues, kwargs) rasa-worker_1 | 2020-06-17 03:51:36 ERROR pika.adapters.utils.io_services_utils - Socket failed to connect: <socket.socket fd=22, family=AddressFamily.AF_INET, type=SocketKind.SOCK_STREAM, proto=6, laddr=(‘172.19.0.7’, 51366)>; error=111 (Connection refused) rasa-worker_1 | 2020-06-17 03:51:36 ERROR pika.adapters.utils.connection_workflow - TCP Connection attempt failed: ConnectionRefusedError(111, ‘Connection refused’); dest=(<AddressFamily.AF_INET: 2>, <SocketKind.SOCK_STREAM: 1>, 6, ‘’, (‘172.19.0.3’, 5672)) rasa-worker_1 | 2020-06-17 03:51:36 ERROR pika.adapters.utils.connection_workflow - AMQPConnector - reporting failure: AMQPConnectorSocketConnectError: ConnectionRefusedError(111, ‘Connection refused’)

There are not other logs apart from the above mentioned.

Best Regards, Ankit Pachauri

8

Hi @erohmensing,

I found the issue. The Domain.yml file was having some inconsistencies. Once I corrected them, the model got trained and is now reflecting in the system Thank you for the proactive help and support :blush: We can close this ticket.

Best Regards, Ankit

1 Like
9

No problem @ankit! Glad you figured it out. Weird that we couldn’t find information about that in the logs!

10

Yes. That was weird

Actually, there were no logs generated. I compared it with another installation and that is when i figured out that the Logs were not generating, which in turn can be considered as Model not being trained. Nonetheless, issue was resolved.:v: :blush: