MCPcopy Index your code
hub / github.com/NVIDIA/TensorRT-LLM / VilaVisionWrapper

Class VilaVisionWrapper

tensorrt_llm/tools/multimodal_builder.py:639–648  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

637 image = image.to(args.device, torch.float16)
638
639 class VilaVisionWrapper(torch.nn.Module):
640
641 def __init__(self, tower, projector):
642 super().__init__()
643 self.tower = tower
644 self.projector = projector
645
646 def forward(self, image):
647 features = self.tower(image)
648 return self.projector(features)
649
650 model = AutoModel.from_pretrained(
651 args.model_path,

Callers 1

build_vila_engineFunction · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected