Back to search
Top Contributors
14
commits
A minimal implementation of Drifting Models for 2D toy data. Unlike diffusion/flow models that iterate at inference, drifting models evolve the pushforward distribution during training and generate in a single forward pass (1-NFE). The drifting field V governs sample movement: V -> 0 as generated matches data.
Stars
70
Forks
5
Watchers
70
Open Issues
0
Overall repository health assessment
No package.json found
This might not be a Node.js project
14
commits