An Awesome New Example of a Fast.AI Powered Product In Industry

Hello folks, I want to share some more exciting research we are doing at Github. This new example uses the language model, and generally builds off the idea @jeremy presented in class re: mapping pictures to word embeddings. This new (prototype) product is natural language semantic code search, where you can search code semantically. The results are very impressive, and we share and end-to-end example with data and code which we have written in this blog post.

Below is a live example in a Jupyter notebook (I recommend taking a look at the blog post for an example with a user-interface mockup) that you will be able to build yourself if you do the tutorial!. Thanks to @jeremy and the community for the support!


This work is wonderful. Congratulations on your result @hamelsmu. I was working through the same notebook trying to access hidden states for extracting sentence embeddings and the schematic in your blog post simplified it all. :wink:


This is incredible! How do I install it?

That is awesome!

Amazing tool!

great work @hamelsmu.