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 3, 10:30 UTC
[BugFix] support PP for Cohere vision model (#42819)
Signed-off-by: <conway.zhu@cohere.com> Signed-off-by: root <conway.zhu@cohere.com>
czhu-cohere committed
May 18, 2026 at 11:12 UTC
9758a6e5c5a556275c030db456d5d434ee999d58
1 file changed
+4
vllm/model_executor/models/cohere2_vision.py
+4
@@ -358,6 +358,10 @@ class Cohere2VisionForConditionalGeneration(
358
architectures=config.text_config.architectures,
359
)
360
361
+ self.make_empty_intermediate_tensors = (
362
+ self.language_model.make_empty_intermediate_tensors
363
+ )
364
+
365
@property
366
def dtype(self):
367
return next(self.parameters()).dtype