Hi. :) I’m Adrien, co-founder of Streamlit. This project started with my personal frustration trying to build high-quality machine learning tools to visualize datasets and models. As a researcher and engineer, I’ve seen how poor tooling slows down machine learning projects. I wondered if we could make building beautiful, interactive tools as easy as writing simple Python scripts.

Over the past year, my teammates and I have worked with a select beta community of over a hundred machine learning engineers to make a free, open-source Python library called Streamlit. We wanted the process of coding beautiful, interactive ML tools to feel like training a neural net or performing an ad-hoc analysis in Jupyter. We think we've achieved this, and I'm really excited to share it with all of you!

We’d love to hear your thoughts and ideas. I’ll be here to answer any questions you might have!

This is pretty cool, it's kind of between a script and a notebook. How well does this work with TensorFlow?

You can see an example of using Streamlit to do realtime inference on a Tensorflow model in this blog post:

https://towardsdatascience.com/coding-ml-tools-like-you-code...

(About 90% through the article there's an animation.) We'll post the code on how we did this soon.

The secret is to use @st.cache(ignore_hash=True) to store Tensorflow state. If more questions / issues come up, please post them here or in our forums discuss.streamlit.io.

Streamlit and Tensorflow work perfectly together. Actually, Streamlit is a great way to explore Tensorflow models in realtime.

..and thanks Srynot!! :)

Also, here's our launch blog post! :)

https://towardsdatascience.com/coding-ml-tools-like-you-code...

Cool stuff! Architecture makes sense. I may try this next model I build.