Training Accuracy

Hi,! I am new to the rasa framework. I trained the bot with 100 epochs and 200 epochs, but both times I got the same accuracy. Is it possible to happen? Pls, help me with this. Thanks in advance

Hey @Vidush_05, welcome to the forum :slight_smile:

It is indeed possible. It may be that the pattern you’re trying to model is very hard and not even 200 epochs is enough. Or it may be that the model has already learnt everything that was learnable…

If you want more specific help, I’m afraid you’ll have to provide more info :wink: Can you post an example of the data you’re trying to learn? Also, can you specify whether this is about NLU or Core training, whether you mean training or test-set accuracy, and possibly share your config?

Sorry for the late response. When I train Core I experienced thisone. Yeah,I will share my config file. nlu_config.yml (247 Bytes)
policies.yml (189 Bytes)

@Vidush_05 unfortunately I can’t say anything more after seeing your configs – they aren’t surprising in any way though :slightly_smiling_face: Can you go back to the other questions I asked?

Could you pls tell me what are the details you need?

Sorry @Vidush_05, I realised that maybe it’d help to take a step back and clarify what it is that you want to know. I said it was perfectly possible to get the same accuracy even if you increase the number of epochs. Based on this, do you have a further question?

Thanks for your response. If I face any trouble I’ll post. Thanks

1 Like