Open Fly-hub opened 11 months ago
Why can only CPU be used for video inference code in reademe? Is there any code that uses GPU for inference?
@Fly-hub, you can do this by using the code below: model = MplugOwlForConditionalGeneration.from_pretrained( pretrained_ckpt, device_map={'': 1}, )
model = MplugOwlForConditionalGeneration.from_pretrained( pretrained_ckpt, device_map={'': 1}, )
Why can only CPU be used for video inference code in reademe? Is there any code that uses GPU for inference?