cmoyates cc2d31b435 feat(phase4d): shape + parity tests, fix attention transpose ... Fix MaskUnitAttention output transpose (0,2,3,1,4 → 0,3,2,1,4) to match PyTorch transpose(1,3) token ordering for windowed attention. Parity results (4 stages): Stage 0: max_abs 2.9e-4, mean 1.6e-5 Stage 1: max_abs 1.3e-4, mean 1.3e-5 Stage 2: max_abs 1.1e-2, mean 5.2e-5 (16 blocks, expected drift) Stage 3: max_abs 5.8e-4, mean 2.6e-5 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>		2026-03-01 05:51:36 -03:30
..
convert	feat(phase3): PyTorch→MLX weight converter (#1)	2026-03-01 05:16:56 -03:30
inference	init: repo scaffolding, deps, prompts, plan	2026-03-01 04:48:51 -03:30
io	init: repo scaffolding, deps, prompts, plan	2026-03-01 04:48:51 -03:30
model	feat(phase4d): shape + parity tests, fix attention transpose	2026-03-01 05:51:36 -03:30
utils	feat(phase2): MLX decoder + refiner with parity tests	2026-03-01 05:00:00 -03:30
__init__.py	init: repo scaffolding, deps, prompts, plan	2026-03-01 04:48:51 -03:30