[Core] Add MRV2 virtual-batch PCP for MLA (#46570)

Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
Co-authored-by: Matthew Bonanni <mbonanni@redhat.com>
This commit is contained in:
Lucas Wilkinson
2026-07-19 02:53:15 +00:00
committed by GitHub
co-authored by Codex OpenAI Codex Matthew Bonanni
parent 9243e0124e
commit b6ff8a2f50
39 changed files with 1204 additions and 196 deletions
@@ -1114,6 +1114,13 @@ def test_sparse_backend_prefill_correctness(
@pytest.mark.parametrize(
"seq_lens,query_lens,workspace_size,max_logits_bytes,expected",
[
(
torch.tensor([0]),
torch.tensor([0]),
100,
1000,
[],
),
# Logits constraint triggers split (M*N exceeds budget)
# req0: M=10, N=100 -> 1000 elems (4000 bytes) - fits in 5000
# req1: adding M=10, N=100 -> new_M=20, new_N=200 -> 4000 elems > 1250