Back to search
Top Contributors
3
commits
3
commits
Conditional Flow Matching (CFM) for image generation on CIFAR-10, benchmarked against DDPM and DDIM. Comparing generation quality (FID) vs. sampling speed (NFE) using shared U-Net backbone.
Stars
0
Forks
0
Watchers
0
Open Issues
0
Overall repository health assessment
No package.json found
This might not be a Node.js project
3
commits
3
commits
Update README with ODE inversion results and project structure
b46cd90View on GitHubAdd NeurIPS report, Beamer presentation, ODE inversion notebook and results
b6c3ceaView on GitHubMove notebook to notebooks/cfm/04_cfm_training.ipynb
ae7c741View on GitHubUpdate README with results and visualizations, add all figures
8962b56View on GitHubFix 2D toy animation to show vector field arrows alongside particles
922b3c3View on GitHubAdd visualization notebook with all 4 proposed visualizations
5550abfView on GitHubAdd results analysis and Elora's conditional CFM evaluation results
d605577View on GitHubReorganize results into ddpm/ and cfm/ subfolders, add Elora evaluation notebook
04be5beView on GitHubFix cfm_sample NameError: move solver definition before training loop
0cab12fView on GitHub