RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:5 and cuda:0! (when checking argument for argument weight in method wrapper_CUDA__cudnn_convolution)

I get this error when I multi-GPU use llava's inference code.How should I fix the code to make it work on multiple gpu's? Thanks,a lot.@John-Ge
` embedding_output = self.vision_tower.embeddings(images)`

Start with llava's inference code and the reported `self.vision_tower.embeddings(images)` call. Trace where the images, vision tower, and model weights are placed across GPUs, then verify multi-GPU inference runs without the reported cross-device RuntimeError.

Written by the indexing model from the issue text.