A tool that automatically writes TensorFlow code from input-output examples

What is it about?

Deep learning frameworks like TensorFlow and PyTorch have steep learning curves. We present a tool, called TF-Coder, that automatically writes TensorFlow code that is consistent with a user-provided example of the desired behavior. TF-Coder solves 63 of 70 real-world tasks within 5 minutes, sometimes finding simpler solutions in less time compared to experienced human programmers. The tool is publicly available, free to use, and has solved hundreds of unique problems from real users.

Featured Image

Photo by Markus Spiske on Unsplash

Photo by Markus Spiske on Unsplash

Why is it important?

Deep learning is becoming increasingly popular and successful, partially due to the development of powerful deep learning frameworks. However, these libraries have steep learning curves. We introduce a tool called TF-Coder that automatically writes TensorFlow code, in order to help programmers develop their deep learning models more quickly and with more confidence. TF-Coder can also identify alternative code implementations that are simpler or more efficient.