D1: Debug TMA partition shapes at hd=512

This commit is contained in:
2026-05-24 04:43:12 +00:00
parent 9201a844dd
commit 24b9310682

View File

@@ -181,6 +181,7 @@ class FmhaKernel:
tBsK,tBgK = cpasync.tma_partition(tma_k,0,b_lay,cute.group_modes(sK,0,3),cute.group_modes(tCgK,0,3))
tVsV,tVgV = cpasync.tma_partition(tma_v,0,b_lay,cute.group_modes(sV,0,3),cute.group_modes(tCgV,0,3))
tAgQ = tAgQ[(None,0,None,0)]; tBgK = tBgK[(None,0,None,0)]; tVgV = tVgV[(None,0,None,0)]
print(f"TMA: tAgQ shape={cute.shape(tAgQ)}, tBgK shape={cute.shape(tBgK)}, tVgV shape={cute.shape(tVgV)}")
tCrQ = qk_mma.make_fragment_A(sQ); tCrK = qk_mma.make_fragment_B(sK)
tCrV = pv_mma.make_fragment_B(sV)