Pre-Trained vectors for tensorflow pipeline

We recently switched over to the tensorflow_embedding pipeline (mostly due to multiple intent support and less memory consumption).

However (due to the architecture of the pipeline), the model doesn’t generalize over new examples at all.

In many cases the models returns an incorrect intent with very high confidence and the correct intent (along with many others) is returned as zero.

Is it possible to have a structured way of solving this problem?

I tired the pipeline specified in this thread

but it didn’t work and I got the same results.

what type of SpaCy’s model are you using?? small, medium or large and in which language?

hey man, could you take a look at this for me?

https://forum.rasa.com/t/agent-handle-message-no-longer-works/815

I’m using English. How does one find out whether small,medium or large is being used?

Are you using the spacy model installed using rasa ? how did you install the spacy backend?

I used the command python -m spacy download en to download it. Running it right now seems to download the small model.

Probably why you don’t a see a difference in performance. The small model of Spacy does not contain a lot of word vectors. Tensorflow is non linear so my assumption is that it should fit better compared to SVM(sklearn pipeline)

But if you want to use pre-trained vectors, maybe try the larger models and see the difference

Okay. I will try that and let you know. But since I have the small models not installed, after installing the large model, how do I ensure that rasa uses the large model and not the small one?

nevermind, got it!