[Bugfix][Pooling] Handle multimodal cache misses without crashing (#58975)

Signed-off-by: linchaoying6 <linchaoying6@jd.com>
Co-authored-by: linchaoying6 <linchaoying6@jd.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
This commit is contained in:
Chaoying
2026-10-02 10:14:33 -04:00
committed by GitHub
co-authored by linchaoying6 OpenAI Codex
parent 443dccd3c9
commit 592c6f3fbc
21 changed files with 525 additions and 17 deletions
+4
View File
@@ -1012,6 +1012,10 @@ class GPUModelRunner(LoRAModelRunnerMixin):
if self.pooling_runner is not None:
self.pooling_runner.clear()
def release_late_interaction_query_cache(self, query_keys: list[str]) -> None:
if self.pooling_runner is not None:
self.pooling_runner.release_late_interaction_query_cache(query_keys)
@torch.inference_mode()
def profile_cudagraph_memory(self) -> int:
"""Estimate the GPU memory required to capture CUDA graphs."""