Research Checkpoint Archive

This repository stores checkpoints from flow-matching image-generation experiments. Each a/rXXX/ directory is one training run.

c.json: sanitized training configuration
v.jsonl: gradient-variance probes, when available
t.jsonl: training metrics, when available
pXXXXXXXX/m.pt: checkpoint at the encoded training step
pXXXXXXXX/q/*.json: FID records corresponding to that checkpoint

The checkpoint files are original PyTorch training checkpoints and include model, EMA, optimizer, scheduler, step, and argument state where available.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support