@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 4c0d93f4b2
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 15, 23:30 UTC

[V1][Bugfix] Copy encoder input ids to fix set iteration issue during VLM abort (#13173)

Signed-off-by: andoorve <37849411+andoorve@users.noreply.github.com>

Murali Andoorveedu committed Feb 12, 2025 at 12:58 UTC 4c0d93f4b2de241336f4732cb5799cee8fedcb52
1 file changed +1 -1
vllm/v1/core/encoder_cache_manager.py
+1 -1
@@ -54,7 +54,7 @@ class EncoderCacheManager:
54
55 def free(self, request: Request) -> None:
56 """Free all cached input ids for the request."""
57 - input_ids = self.get_cached_input_ids(request)
57 + input_ids = self.get_cached_input_ids(request).copy()
58 for input_id in input_ids:
59 self.free_encoder_input(request, input_id)
60