thuEcstasy1/InfuseLoop-4
Text Generation • 0.6B • Updated • 362
InfuseLoop-4 and InfuseLoop-8: cross-trajectory latent injection for scalable parallel looped Transformers. 0.6B base models pretrained on 6B tokens.
Note 4-loop model; final 6B-token checkpoint.
Note 8-loop model; final 6B-token checkpoint.