Rasa falls into fallback zone everytime while running into docker container

Rasa Open Source
1

When I trained the model on my local PC, it ran perfectly. However, when I tried running it in a Docker container after installing all dependencies, I encountered a fallback error everytime.

Here is the version details: Rasa Version : 3.6.20 Minimum Compatible Version: 3.5.0 Rasa SDK Version : 3.6.2 Python Version : 3.8.18

Here is my config file:

recipe: default.v1

# The assistant project unique identifier
# This default value must be replaced with a unique assistant name within your deployment
assistant_id: 20240610-175227-regular-block

# Configuration for Rasa NLU.
language: en

policies:
  - name: MemoizationPolicy
    priority: 3
  - name: RulePolicy
    priority: 1
    core_fallback_threshold: 0.3
    core_fallback_action_name: "action_default_fallback"
    enable_fallback_prediction: True

pipeline:
  - name: SpacyNLP
    model: "en_core_web_md"
  - name: SpacyEntityExtractor
    dimension: ["NAME"]
  - name: WhitespaceTokenizer
    intent_split_symbol: "+"
  - name: RegexFeaturizer
  - name: LexicalSyntacticFeaturizer
  - name: CountVectorsFeaturizer
    analyzer: word
  - name: CountVectorsFeaturizer
    analyzer: char_wb
    min_ngram: 1
    max_ngram: 4
  - name: LanguageModelFeaturizer
    model_name: "bert"
    model_weights: "bert-base-uncased"
  - name: DucklingEntityExtractor
    url: "http://localhost:8000"
    dimensions:
      - time
      - phone_number
      - email
      - name
      - number
  - name: DIETClassifier
    epochs: 150
    constrain_similarities: true
    entity_recognition: true
    entity_roles: true
    entity_groups: true
  - name: components.address_city_state_detector.AddressCityStateDetector
  - name: RegexEntityExtractor
    use_lookup_tables: true
    use_regexes: true
    use_word_boundaries: true
  - name: EntitySynonymMapper
  - name: FallbackClassifier
    threshold: 0.6
    ambiguity_threshold: 0.1
2

maybe you should run duckling in a container too, so, change in you config , localhost to the name of container. Same thing to rasa action, as rasa nlu runs separete from rasa actions, as you run rasa shell and rasa run actions, you have to run rasa action in a separete container too, and in your endpoints set the name of the container that its running rasa actions. Then probably its gonna work. Run in a compose docker, its wil works easy