when i use regex in training data, it raise the error “fulldialoguetrackerfeaturizer is not json serializable”
Can you show an example of your training data? Have you tried the RegexFeaturizer
{“rasa_nlu_data”: {“regex_features”:[{‘name’: ‘number_arab’,‘pattern’: ‘[0-9]{1,}’}], “entity_synonyms”:[], “common_examples”:[]}
pipeline:
- name: “JiebaTokenizer”
- name: “RegexFeaturizer”
- name: “CRFEntityExtractor”
- name: “CountVectorsFeaturizer” OOV_token: oov token_pattern: “(?u)\b\w+\b”
- name: “EmbeddingIntentClassifier” epochs: 10
policies:
-
name: “EmbeddingPolicy” epochs: 10 featurizer:
- name: MaxHistoryTrackerFeaturizer
max_history: 3
state_featurizer:
- name: LabelTokenizerSingleStateFeaturizer
- name: MaxHistoryTrackerFeaturizer
max_history: 3
state_featurizer:
-
name: FallbackPolicy fallback_action_name: ‘action_default_fallback’ nlu_threshold: 0.5 core_threshold: 0.3
-
name: MemoizationPolicy max_history: 3