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 13, 07:00 UTC
[BugFix] DeepSeek-OCR: apply NoRepeatNGramLogitsProcessor to greedy path (#28617)
Signed-off-by: Yuanping Song <yuanping.song@outlook.com>
Yuanping Song committed
Nov 13, 2025 at 10:24 UTC
3035d1a166821272d4e7eb204e2c613bb02bacd7
1 file changed
+1
-1
vllm/model_executor/models/deepseek_ocr.py
+1
-1
@@ -161,7 +161,7 @@ class NGramPerReqLogitsProcessor(AdapterLogitsProcessor):
161
)
162
163
def is_argmax_invariant(self) -> bool:
164
- return True
164
+ return False
165
166
def new_req_logits_processor(
167
self,