Does Rasa generate the exact same model every time with the same input? Or there are little variations? Maybe it depends on the choosen pipeline?
With the default settings, there will be some variation. If you would like the same model to be produced every time you train with the same data, you can set the random seed for each of the components you’re using. In Rasa 1.x you can do that in the config file using the random_seed argument, like so:
policies:
- name: TEDPolicy
max_history: 5
epochs: 100
random_seed: 42
Hi @rctatman, thank you! Does this apply to Rasa 2.x?
Yep, it should! (Didn’t have a 2.0 install handy when I answered yesterday to check.)