[Attention][MLA] Fuse Kimi-K3 chunked-context K/V packing (#51772)

Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Yongye Zhu
2026-08-12 22:42:57 -07:00
committed by GitHub
co-authored by Claude Opus 5 mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent 63344913a6
commit 903d2efe7e
15 changed files with 996 additions and 35 deletions
@@ -24,7 +24,7 @@ class CustomMLAPrefillBackend(MLAPrefillBackend):
def run_prefill_new_tokens(self, q, k, v, return_softmax_lse):
raise NotImplementedError
def run_prefill_context_chunk(self, chunk, q, k, v):
def run_prefill_context_chunk(self, chunk, q, k, v, out=None):
raise NotImplementedError
@@ -112,7 +112,7 @@ def test_register_custom_backend_as_decorator():
def run_prefill_new_tokens(self, q, k, v, return_softmax_lse):
raise NotImplementedError
def run_prefill_context_chunk(self, chunk, q, k, v):
def run_prefill_context_chunk(self, chunk, q, k, v, out=None):
raise NotImplementedError
assert MLAPrefillBackendEnum.CUSTOM.is_overridden()