I have only recently realised this and want to know if my chatbot is training properly with all the NLU modules i have listed. Both DIETClassifier and ResponseSelector is at 0% epoch X_dense and throws deprecated message but then it continues to train on the line that states after final_data[k] so I am confused what this means and whether its training using those items at all. my Response selector doesnt seem to work so this might have something to do with it
Hi,
Can you please specify the rasa version you are using?
Hey I am using Rasa 1.10.10 and Python 3.7.6, Windows 10.
Can you create a smaller subset of your data and see if the problem persists. If it does, it will help if you can share that smaller subset of training data here so that I can reproduce the error on my side.
Does not seem to be based on training data, I have also tried using a dietclassifier for example in one of the small sample projects that was created halfway through the Rasa Masterclass. and the same issue persists. the image seems to indicate an issue with diet_classifier.py line 595 and model_data.py:386.for DIETClassifier, and model_data.py:386 for ResponseSelector as seen in the first screenshot above.
sample training data from one of my project nlu.md (1.4 KB)
Has this issue been looked at? I have noticed another user with same issue.
The warning originates from tensorflow under the hood and is indicating some possible issue with the data wrangling we do inside. The reported accuracies seem fine in both the screenshots, so unlikely this is affecting the training. I am still curious to see what brings up that warning. Can you share the link of the sample project from the Rasa Masterclass where you encountered this?
Also, please share the version of tensorflow installed in your virtual environment.
The tensorflow version I am using is 2.1.0
As for the sample project, what I mean is I created the project using Rasa init, and followed the same steps in the masterclass. I do not have my work uploaded to git so I am unable to provide a link. DIET works as extractor, but I am also having problems with responseselector so I am unsure if this is part of the problem.
What problem are you having with the response selector. Sorry, it’s a bit tough to debug this without having a reproducible example.