KAIBER NN Editor for PyTorch

Powerful and intuitive neural network editor for PyTorch

KAIBER NN Editor for PyTorch is a browser-based development support tool that streamlines the design of AI models through interactive operations.

Deep learning neural networks are currently implemented with error-prone code descriptions. Deep Insight has developed a PyTorch-specific GUI-based neural network design tool based on the expertise gained from the edge AI development tool KAIBER.

With an intuitive visual design interface compatible with the PyTorch framework, which is growing in popularity for its flexible features such as Define by Run, the neural network layers and their interactions can be visually described to enable AI model behavior without sacrificing expressiveness or performance. It allows for easy understanding and rapid design and generation of NN code. It also facilitates debugging and automates PyTorch-specific development practices for framework and code description.

Free trial beta 5 (Chrome compatible)
Easy access!

~ Try sample AI models below ~

Sample AI models
Download

The version has been upgraded to beta 5 (0.5.0).
Please check the version using the i mark in the upper right corner of the NN Editor
and clear your browser cache.

Key points of KAIBER NN Editor for PyTorch

Native support
for PyTorch

Rather than supporting a variety of frameworks in a generic fashion, PyTorch specialization focuses on a single design methodology, simplifies integration with previously developed Python code, and reduces bugs by automatically checking for consistency with PyTorch's own NN design methodology.

Designing neural network with chart graphs

The focus is on providing a visual and clear interface in the neural network design process. Rather than replacing all code, model architecture and data flow visualization makes complex models easy to debug and maintain.

Visual and code-based hybrid approach

Supports "Python-based deep learning model" design methodology. It automatically generates code for the learning models available in regular PyTorch and can be combined with Python preprocessing programs and neural network implementations.

Visual and code-based hybrid approach

Supports "Python-based deep learning model" design methodology. It automatically generates code for the learning models available in regular PyTorch and can be combined with Python preprocessing programs and neural network implementations.

Main Features

Visual model design

✔︎ Designing Neural Networks with Mouse Operation
　　 Intuitive design experience with visual grasp of model structure

✔︎ Supports operator drag & drop, cut, copy and paste
Easy to create models with similar repeated structures

✔︎ Supports Undo and Redo
Easy to undo and redo even if you make a mistake

Operator Settings

✔︎ Display setting items in list format for each operator
View a list of all available settings.

Real-time error checking

✔︎ Real-time display of the shape of each operator output
Check output shapes and find bugs quickly

Output Python code

✔︎ Created NN models can be exported as source code for PyTorch　　　
Quickly incorporate into your AI learning code

Document links

✔︎ Link to PyTorch operator documentation on screen
Quick reference to the relevant document

Operator Settings

✔︎ Display setting items in list format for each operator
View a list of all available settings.