biondizzle/nvfp4-megamoe-kernel
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

DEBUG: print TMA partition tensor shapes

Browse Source
This commit is contained in:
biondizzle
2026-05-22 22:25:36 +00:00
parent 16f60e2dd1
commit 1b35e0f967
1 changed files with 9 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
tests/unit/test_fmha_v3_stage_c.py
View File

@@ -179,7 +179,16 @@ class FmhaV3StageCMulti:
b_lay = cute.make_layout(cute.slice_(cl_vmnk,(0,None,0,0)).shape)
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))
# DEBUG: print TMA partition tensor shapes BEFORE pre-slice
print(f"tAgQ shape before: {cute.shape(tAgQ)}")
print(f"tBgK shape before: {cute.shape(tBgK)}")
print(f"tVgV shape before: {cute.shape(tVgV)}")
print(f"tBsK shape: {cute.shape(tBsK)}")
print(f"tVsV shape: {cute.shape(tVsV)}")
tAgQ = tAgQ[(None,0,None,0)]; tBgK = tBgK[(None,None,0,0)]; tVgV = tVgV[(None,0,None,0)]
# DEBUG: after pre-slice
print(f"tBgK shape after pre-slice: {cute.shape(tBgK)}")
print(f"tVgV shape after pre-slice: {cute.shape(tVgV)}")
tCrQ = qk_mma.make_fragment_A(sQ); tCrK = qk_mma.make_fragment_B(sK)
tCrV = pv_mma.make_fragment_B(sV)