Differential + parallel attn#2625

Merged

rwightman merged 3 commits intomainfrom

Dec 8, 2025

Commits on Dec 7, 2025

Trying out differential + parallel block, and fused out project option for ParallelScalingBlock to see if it's worthwhile now...
rwightman
committed

Commits on Dec 8, 2025

Add dwee model def, and weights for dwee and dpwee
rwightman
committed
Add weights_only=False to avg_checkpoints, needs to load training state dict
rwightman
committed