Hello, As the title says, when i train a model with a new intent and upload it to my Rasa X server, the assistant cannot understand it (with the newest model activated). If I however change the language in the config file, and upload this new model, and change it back it can. How could this be? I’ve made a post earlier regarding the same problem, when trying to train the model via the Rasa X UI. Thanks in advance!

do you maybe have too few training data? How does your NLU pipeline look like?

so my pipeline looks like this: # Configuration for Rasa NLU. # https://rasa.com/docs/rasa/nlu/components/ language: da # # No configuration for the NLU pipeline was provided. The following default pipeline was used to train your model. # # If you'd like to customize it, uncomment and adjust the pipeline. # # See https://rasa.com/docs/rasa/tuning-your-model for more information. pipeline: - name: SpacyNLP model: en_core_web_md - name: SpacyTokenizer - name: SpacyEntityExtractor - name: SpacyFeaturizer pooling: mean - name: CountVectorsFeaturizer analyzer: char_wb min_ngram: 1 max_ngram: 2 - name: DIETClassifier - name: FallbackClassifier threshold: 0.8 ambiguity_threshold: 0.1 epochs: 1 # Configuration for Rasa Core. # https://rasa.com/docs/rasa/core/policies/ policies: # # No configuration for policies was provided. The following default policies were used to train your model. # # If you'd like to customize them, uncomment and adjust the policies. # # See https://rasa.com/docs/rasa/policies for more information. - name: "AugmentedMemoizationPolicy" max_history: 10 - name: TEDPolicy # max_history: 5 # epochs: 100 # constrain_similarities: true - name: RulePolicy core_fallback_threshold: 0.4 core_fallback_action_name: "action_default_fallback" enable_fallback_prediction: True and my nlu file: version: "2.0" nlu: - intent: goodbye examples: | - good afternoon - cu - good by - cee you later - good night - bye - goodbye - have a nice day - see you around - bye bye - see you later - intent: affirm1 examples: | - Ja det har jeg - Jeg har et spørgsmål mere - Jeg vil gerne vide noget mere - Det lyder godt - Okay - Super - Jeg har et spørgsmål - modtaget - Fantastisk - Spændende - intent: affirm2 examples: | - yes - ja - Y - Korrekt - Præcis - Nøjagtig - Nemlig - ja tak - ok - jep - intent: deny examples: | - no - n - Nej - Det har jeg ikke - Jeg har ikke flere spørgsmål - Niks - Ellers tak - Nej tak - Niksen biksen - Nej det har jeg ikke - intent: bot_challenge examples: | - are you a bot? - are you a human? - am I talking to a bot? - am I talking to a human? - Hvem er du? - intent: Hilsen examples: | - Dav - Goddag - Hej - Hej med dig - Halløjsa - Hejsa - Goddag med dig bot - hyggehejsa - Halløjsovs - Davs med dig - Hej bot - pænt Goddag - Halløj - Hej min ven - Hej dig - Godmorgen - Godaften - hej - hejsa - hello - hrj - Chatbot? - Hej DreamPlan - intent: Datasikkerhed examples: | - Er min data sikker? - Hvordan opbevarer Dreamplan mine data? - Hvad gør Dreamplan for at holde data sikkert? - Hvor godt er min data beskyttet? - Er Dreamplan en sikker service? - Hvor fortrolig er Dreamplan? …

Can you please format your code blocks using to make it easier to read? Thank you :raised_hands:

Sorry about that, thought the yaml output was fine, but can see it is much easier to read now :slight_smile: Thank you

It’s still not formatted :see_no_evil:

that ought to do it :sweat_smile:

Hi Blaabjerg96, If you want to share code you can use the markdown syntax and if you’re new there’s a syntax guide here . In your case, you seem to have copied code with extra whitespaces that are getting rendered. I wanted to double-check your config.yml file, since you seem to be training a multi-lingual system. This is independant of your Rasa X issue, but why are you using an English spaCy tokeniser/featurizer when you seem to also have Danish in there? If you’re interested in building a multi-lingual system you might prefer to use language agnostic sentence embeddings (LaBSE). It’s a bit heavyweight, but it’s a system trained to understand multiple languages. We’ve got an algorithm whiteboard video on the topic and you can find the relevant Rasa component described on our docs .

Thats the core of the issue. If I use this configuration I can train the assistant in Rasa X server, and if I add new intents the danish SpaCy tokeniser needs to be changed to english and back again to recognize the new intent. I am not looking to make a bi-lingual bot, but I kept the intent around, If I needed the name, so I just needed to change the data. Thanks for the guide by the way - should be correctly set up now - third times the charm right :sweat_smile:

Rasa X will only update model if I change the config file

[Deprecated] Rasa X Community Edition

Blaabjerg96 (Blaabjerg96) May 28, 2021, 3:41pm 7

that ought to do it

Topic		Replies	Views
Change language of Rasa X models [Deprecated] Rasa X Community Edition	1	907	October 21, 2019
Rasa X on AWS can only train a english chatbot [Deprecated] Rasa X Community Edition	0	325	May 11, 2021
[How To] migrate pre-trained Rasa model to Rasa X [Deprecated] Rasa X Community Edition	14	3211	March 16, 2021
How rasa x train model, How its work? [Deprecated] Rasa X Community Edition	1	317	June 12, 2021
Rasa X: Worker cannot train the model after modifying the docker file for third language [Deprecated] Rasa X Community Edition	0	1073	March 14, 2021

Rasa X will only update model if I change the config file

Related topics