[Perf][Qwen4Exp] Merge QSA QKVG and indexer QK projections (#59533)

Signed-off-by: Thien Tran <gau.nernst@yahoo.com.sg>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: namgyu-youn <namgyu.dev@gmail.com>
This commit is contained in:
Thien Tran
2026-10-05 07:03:29 +00:00
committed by GitHub
co-authored by Codex namgyu-youn
parent 92044241a0
commit 042ab0305c
6 changed files with 181 additions and 119 deletions
+8 -4
View File
@@ -74,13 +74,14 @@ def _make_main_inputs(num_tokens: int, fp8_cache: bool) -> dict:
slots = torch.randperm(num_blocks * MAIN_PAGE, device="cuda")[:num_tokens]
slots[-1] = -1
norm_weights = torch.randn(2, MAIN_D, dtype=torch.bfloat16, device="cuda") * 0.2
qkv_width = 2 * (MAIN_HQ + MAIN_HK) * MAIN_D
return dict(
main_qkv=torch.randn(
num_tokens,
2 * (MAIN_HQ + MAIN_HK) * MAIN_D,
qkv_width + (HQ + 1) * D,
dtype=torch.bfloat16,
device="cuda",
),
)[:, :qkv_width],
main_q_norm_weight=norm_weights[0],
main_k_norm_weight=norm_weights[1],
main_eps=EPS,
@@ -316,9 +317,12 @@ def test_qsa_fused_prepare_matches_unfused(
)
unfused_compressed = fused_compressed.clone()
# TP=4 QKVG width makes the indexer slice row-strided, as in the merged
# projection. The exact width is unimportant; the larger stride is tested.
main_qkvg_width = 3584
projected_qk = torch.randn(
num_tokens, (HQ + 1) * D, dtype=torch.bfloat16, device=device
)
num_tokens, main_qkvg_width + (HQ + 1) * D, dtype=torch.bfloat16, device=device
)[:, main_qkvg_width:]
q_weight = torch.randn(D, dtype=torch.bfloat16, device=device) * 0.2
k_weight = torch.randn(D, dtype=torch.bfloat16, device=device) * 0.2