Machine-Learning-Tokyo/Interactive_Tools

Curated collection of browser-based interactive visualizations for understanding ML/DL model internals, spanning transformer attention mechanisms, CNN feature maps, GAN training dynamics, and embedding spaces using techniques like activation atlases and attention visualization. Integrates with HuggingFace transformers, TensorFlow, and embedding projection algorithms (PCA, t-SNE, UMAP). Tools run inference client-side in the browser, enabling real-time experimentation without server infrastructure.

2,803 stars. No commits in the last 6 months.