John 93e31f9f80 Sparse DiT blocks for the SLAT flows ... SparseDiTAttention / SparseProjectAttention / ModulatedSparseTransformerCrossBlock - the sparse counterparts of the dense DiT pieces, attending within each batch item. The modulation needs care: it is per batch item ([B, 6C]) while features are a flat [N, C] stack, so each row must pick up its own item's shift/scale/gate. Broadcasting would silently apply item 0's modulation to everything when B == 1. Verified end to end: slat_flow matches upstream at correlation 1.00000000.		2026-08-02 12:10:31 +10:00
..
__init__.py	Sparse DiT blocks for the SLAT flows	2026-08-02 12:10:31 +10:00
conv.py	Shared TRELLIS-lineage sparse core in MLX	2026-08-02 10:31:07 +10:00
convert.py	Share the 5-D layout disambiguation	2026-08-02 11:25:09 +10:00
dit.py	Sparse DiT blocks for the SLAT flows	2026-08-02 12:10:31 +10:00
ops.py	Shared TRELLIS-lineage sparse core in MLX	2026-08-02 10:31:07 +10:00
tensor.py	Add masked upsample and parameterise downsample reduction (16/16 tests)	2026-08-02 10:42:20 +10:00