[CPU] optimize flash_attn_varlen_func (#15708)

This commit is contained in:
Ma Mingfei
2026-01-29 22:07:05 -08:00
committed by GitHub
parent 336dc4579e
commit 88f7759402
10 changed files with 1189 additions and 339 deletions
+1
View File
@@ -183,6 +183,7 @@ class TestExtendAttention(CustomTestCase):
self._test_extend_attention_once(1, 123, 1, 1, 128, 96, is_mla)
self._test_extend_attention_once(1, 123, 16, 1, 128, 96, is_mla)
self._test_extend_attention_once(4, 1230, 16, 4, 128, 96, is_mla)
self._test_extend_attention_once(1, 9000, 16, 1, 32, 32, is_mla)
if __name__ == "__main__":