[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
+3
View File
@@ -993,6 +993,9 @@ class GPUModelRunner(
self.encoder_cache.clear()
self.late_interaction_runner.clear()
def release_late_interaction_query_cache(self, query_keys: list[str]) -> None:
self.late_interaction_runner.release_queries(query_keys)
def _get_positions(self, num_tokens: Any):
if isinstance(num_tokens, int):
if self.uses_mrope: