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, 15:45 UTC
[fix] Revert "fixing mm placeholder replacement issue with gemma3" (#28285)
Signed-off-by: Kevin H. Luu <khluu000@gmail.com>
Kevin H. Luu committed
Nov 7, 2025 at 04:09 UTC
8e19d470af0a39a25dec799f81c0fb922777c240
1 file changed
+1
-1
vllm/model_executor/models/gemma3_mm.py
+1
-1
@@ -401,7 +401,7 @@ class Gemma3MultiModalProcessor(BaseMultiModalProcessor[Gemma3ProcessingInfo]):
401
402
def get_repl_toks(tok: int) -> list[int]:
403
if tok == newline_3:
404
- return [newline_2, newline_1]
404
+ return [newline_1, newline_2]
405
if tok == newline_4:
406
return [newline_2, newline_2]
407