Xircuits is a Jupyterlab-based extension that enables visual, low-code, training workflows. It allows anyone to easily create executable Python code in seconds.
It is created by data scientists for data scientists.
Rich Xircuits Canvas Interface
Unreal Engine-like Chain Component Interface
Custom Nodes and Ports
Smart Link and Type Check Logic
Xircuits generates executable Python scripts from the canvas. As they're very customizable, you can perform DevOps automation like actions. Consider this Xircuits workflow which trains an mnist classifier.
You can run the code generated Python script in Xircuits, but you can also take the same script to train 3 types of models in one go using bash script:
TrainModel.py --epoch 5 --model "resnet50"
TrainModel.py --epoch 5 --model "vgg16"
TrainModel.py --epoch 5 --model "mobilenet"
Famous Python Library Support
Didn't find what you're looking for? Creating Xircuits components is very easy! If it's in Python - it can be made into a component. Your creativity is the limit, create components that are easily extendable!
Created a cool component library? All your colleagues need to do is to drop your component library folder in theirs and they can immediately use your components.
And many more.
How it Works
- Create a new Xircuits from the Jupyterlab launcher or open an existing .xircuits file.
- Drag components from the Xircuits Component Tray to form your training workflow. Share data among your components using the inPorts and outPorts.
- Press Compile button! The Xircuits will compile the components into a working Python code.
- Run it!
Xircuits will execute the compiled Python code on the Output Viewer.