Transformers Neural Network — Interactive Visualizer
Interactive Visualization • Live Training • GPT vs T5
GPT (Decoder-only)
T5 (Encoder-Decoder)
3D Visualization
I love AI
Hello world
Machine learning
Neural networks
Train
Pause
Forward
Backward
Reset
Loss:
0.000
Epoch:
0
Mode:
GPT
Heads:
4
Input
Encoder
Decoder
Output
Click on a token to inspect multi-head attention and gradients.