TensorFlow pipeline RASA

Has anyone noticed tensorflow pipeline overfitting the data. How is it internally working?. Which api does it call it internally and can we tweak the parameters to get better result.

Yes, I’m facing a similar issue. I think you need to add more phrases in nlu.md file.

Hi, TensorFlow pipeline uses limited vocabulary that it could find in training data, so it has troubles to classify test sentences with a lot of unseen words. Did you check documentation for description: http://rasa.com/docs/nlu/pipeline/#intent-classifier-tensorflow-embedding?

Hi, I am trying to install tensorflow_embedding pipeline, using the recommended “pip install rasa_nlu[tensorflow]” command in my Ananconda command prompt. While all other requirements were satisfied, I ran into the following error.

May I know which distribution should I should to install this pipeline? I am using Python 3.7

Collecting tensorflow; extra == “tensorflow” (from rasa_nlu[tensorflow]) Could not find a version that satisfies the requirement tensorflow; extra == “tensorflow” (from rasa_nlu[tensorflow]) (from versions: ) No matching distribution found for tensorflow; extra == “tensorflow” (from rasa_nlu[tensorflow])

@ghostshiphost Has the same error, last I checked latest tensorflow supports for python 3.6. I suggest create a virtual environment for python 3.6 and then install

1 Like