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 9, 21:48 UTC
[Bugfix] Fix Nemotron-Nano-v2-vlm static resolution (#32682)
Signed-off-by: Netanel Haber <58652339+netanel-haber@users.noreply.github.com>
Netanel Haber committed
Jan 21, 2026 at 08:28 UTC
27ca95b3c9e6e32ac02508e729c01865800fd036
1 file changed
+3
-1
vllm/model_executor/models/nano_nemotron_vl.py
+3
-1
@@ -1678,7 +1678,9 @@ class NemotronH_Nano_VL_V2(
1678
pixel_values_flat=pixel_values_flat, **kwargs
1679
)
1680
else:
1681
- return NanoNemotronVLImagePixelInputs(**kwargs)
1681
+ return NanoNemotronVLImagePixelInputs(
1682
+ num_patches=kwargs.pop("image_num_patches"), **kwargs
1683
+ )
1684
1685
def _process_image_input_dynamic(
1686
self, image_input: NanoNemotronVLImagePixelInputsDynamic