Training model failed?

Hi everyone

I was updating my NLU data on Rasa x, I went to train my model and I keep on getting this error that my training has failed but there’s no explanation as to why or where the error is.

This is the first time this has happened as it has not been working since last night. I am at a complete loss on how to resolve it, any suggestions would be much appreciated.

Please find a screenshot of the error below.

Many thanks!

What’s your deployment? Local, Quick-Install, Helm Chart, or Docker-Compose?

Hello, it’s Docker-compose

@yasmincodes If you have Docker Desktop, check the logs, as on current status Rasa-X latest having issues. Try rasa-x 0.39.3 with Rasa 2.8.0 / 2.8.1

Hi yes I checked the logs and I had missing data in my training data. Docker told me what was missing and I was able to add the missing data back into the training data and now the model trains without errors. Thank you both :slight_smile:

@yasmincodes Great,Congrats and keep an eye on logs :eye: please close this topic with the solution for others. Good Luck with the project.

1 Like

I am having the same issue but I did the quick install ?Like training failed

@Rishabhh17 Not get you on this? can you elaborate.

I have done the server installations of Rasa X on azure instance. When I open the url and training the model it gets trained but when I go for (Talk to your bot),its taking the intents but not giving any response.

Did you check the Pod logs?

You can get the Pod names with

kubectl get pods --namespace rasa

Then get the logs with

kubectl logs <pod_name> --namespace rasa

Didnt get that.

You did a Quick-Install, this is equivalent to Helm Install.

Helm install uses Kubernetes. Kubernetes is made of Pods. Pods are made of Docker Containers.

You have Pods for Rasa X, the Action Server, NLU, Training, etc.

To see your pods, write the command

kubectl get pods --namespace rasa

Once you get the name of the pods, you can individually check their logs to see what went wrong with the command:

kubectl logs <pod_name> --namespace rasa

For example:

kubectl logs rasa-rasa-worker-1a2b3c --namespace rasa

I would encourage you to read more about Kubernetes. You can check the official Rasa Advanced Deployment Workshop.

1 Like

Means I have to go for Docker compose for server installation.

Not necessarily. I use Quick-Install (Helm) as well, never used Docker-Compose.

But try both and choose your favorite method if you want to!

I used Server Quick-Install and the bot gets deployed and its not responding to any intents.

Yes I understand that, did you check the logs and see any error?

Where i get those?

Already told you twice. Please see this reply.

1 Like