We recently announced our new 14-episode series focused on developing AI assistants with Rasa Open Source (version 2.x and later) from scratch. Hosted by our Developer Relation team @rctatman@Juste & @koaning, each episode is dedicated to a specific topic and combines theoretical information with code examples and demos.
The entire process of developing a contextual AI assistant will be covered, you don’t have to be a Rasa or machine learning expert to follow the series, but you will definitely become an experienced Rasa developer by the end of it!
Here is a full overview and schedule for upcoming episodes:
This is amazing and so awesome. I love how everything will be broken down into logical segments. For a newbie like me, this will be perfect. Thank you!
Looking forward as well! Maybe you guys can cover things like docker deployment, building a docker custom action image, using a custom action image in kubernetes, deploying to production, conditional logic in forms. The other aspects are pretty clear, but the above not really
Thanks!
While deployment was already covered in the Live Coding playlist, I agree with @georgebrianb, it would be nice to have a short video about it! Maybe that’s episode 13?
@georgebrianb thank you for the feedback and suggestions @koaning is currently working on a “Understanding Rasa Deployments” course. We will share more info soon.
Awesome initiative.
Can you point me to resources where I can run a DIETClassifier on huge data inside a GPU VM instance?
My setup is GCP VM instance with GPU (Cuda 11.2).
When running DIETClassifier, it is not recognizing the GPU and it uses only CPU which takes around 20 hours to train 100 Epochs. Any suggestions are welcome.
Hi there and thanks for this series which has been great so far. I just finished watching Custom Forms and I really think it could for the basis for a use-case I have. Is the pizza bot example that Vincent was using available on github? I look at one that is there but it’s not the same as the one Vincent had in his video. Thanks.
Could you start a new thread/question with information about your situation? Slow training times may have a lot of causes and a GPU may not always be the best solution.