TedPolicy doesn't learn at all


I am training on a Windows with Rasa Open Source my bot. However, my Ted Policy does not learn at all. It always predicts no enity for the test stories. My NLU however, extract the enitities in a good way. I also changend the epochs training rate and the max history (both increased) which did not help

Hello, If your Ted Policy is not learning and always predicts no entity for the test stories, there are a few things you could try to debug the issue:

  1. Check your training data: Make sure that your training data contains enough examples with entities, and that the entities are properly labeled.
  2. Check your configuration: Ensure that the entity extraction configuration is correct, and that the Ted Policy is included in your pipeline.
  3. Try changing hyperparameters: Experiment with different hyperparameters for the Ted Policy, such as the learning rate and batch size.
  4. Use a different policy: If the Ted Policy is not working for your use case, try using a different policy, such as the Regex or CRF Entity Extractor.
  5. Check your version compatibility: Make sure that your Rasa Open Source version is compatible with the Ted Policy you are trying to use. If none of these solutions work, you may need to further investigate the issue or seek assistance from the Rasa community.