5 Commits

Author	SHA1	Message	Date
biondizzle	8f35b75164	D2: comprehensive head-packed test (n_h=1, 64, 128, hd=64, 128)	2026-05-25 17:16:05 +00:00
biondizzle	efdedab399	fix tests: use 3D tensors (M, hd, 1) matching kernel local_tile expectations	2026-05-25 16:54:56 +00:00
biondizzle	a4499f5aa8	fix tests: pad Q to 128 rows (M tile size) for all configs	2026-05-25 16:53:17 +00:00
biondizzle	af136eee27	fix: use CUstream instead of cuStream(0)	2026-05-25 16:51:52 +00:00
biondizzle	4826fa6afb	D2: add num_query_heads/batch_size params + head-packed test ... - FmhaKernel.__init__: add num_query_heads=1, batch_size=1 - Grid: (ceil_div(n_h*T, 128), 1, batch) for multi-CTA - Test: head-packed multi-head (Q reshaped to (n_h*T, hd)) - n_h=1 regression, n_h=128 Pro decode, n_h=64 Flash, hd=128	2026-05-25 16:50:49 +00:00