[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
+8
View File
@@ -323,6 +323,14 @@ class LLMEngine:
engine_core_timestamp=outputs.timestamp,
iteration_stats=iteration_stats,
)
mm_processor_cache = self.renderer.mm_processor_cache
if mm_processor_cache is not None:
for engine_core_output in outputs.outputs:
if engine_core_output.mm_cache_miss_hashes:
for mm_hash in engine_core_output.mm_cache_miss_hashes:
mm_processor_cache.invalidate(mm_hash)
self.output_processor.update_scheduler_stats(outputs.scheduler_stats)
# 3) Abort any reqs that finished due to stop strings.