Hello everyone,
I’m currently working on developing a chatbot using Rasa, and I’ve been experimenting with different configurations in my pipeline. I’m using TensorBoard to track my model’s performance. Specifically, I’m trying to determine the optimal value for evaluate_on_number_of_examples in the DIETClassifier component.
However, I’m uncertain about the best approach to choose this value of evaluate_on_number_of_examples.
Could anyone provide insights or advice on how to choose the right value for evaluate_on_number_of_examples? What factors should I consider when making this decision? Are there any best practices or guidelines that could help me optimize this parameter for my specific use case?
Thank you in advance for your help!
this is my pipline pipeline:
- name: SpacyNLP model: “fr_core_news_md”
- name: SpacyTokenizer
- name: SpacyFeaturizer
- name: RegexFeaturizer
- name: LexicalSyntacticFeaturizer
- name: CountVectorsFeaturizer stop_words: [“le”, “la”, “et”, “de”, “un”, “une”, “les”, “vous”, “ces”, “au”, “nous”, “vous”, “je”, “il”]
- name: CountVectorsFeaturizer analyzer: “char_wb” min_ngram: 1 max_ngram: 3
- name: DIETClassifier constrain_similarities: True epochs: 200 ranking_length: 5 evaluate_on_number_of_examples: 25 evaluate_every_number_of_epochs: 5 tensorboard_log_directory: “.tensorboard” tensorboard_log_level: “epoch”
- name: RegexEntityExtractor
- name: ResponseSelector
epochs: 200