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, 17:15 UTC
[BugFix][V1] Quick fix for min_tokens with multiple EOS (#15407)
Signed-off-by: Nick Hill <nhill@redhat.com>
Nick Hill committed
Mar 24, 2025 at 15:58 UTC
623e2ed29f70dec5af892871f0307655e348c8a9
1 file changed
+2
-1
vllm/sampling_params.py
+2
-1
@@ -369,8 +369,9 @@ class SamplingParams(
369
self.top_k = -1
370
self.min_p = 0.0
371
self._verify_greedy_sampling()
372
+
373
# eos_token_id is added to this by the engine
373
- self._all_stop_token_ids = set(self.stop_token_ids)
374
+ self._all_stop_token_ids.update(self.stop_token_ids)
375
376
def _verify_args(self) -> None:
377
if not isinstance(self.n, int):