Hi everyone, I’m trying to use DPTForDepthEstimation on GPU to accelerate inference, but seems like it has not been implemented for some reason. Here’s my code snippet and errors.
from transformers import DPTForDepthEstimation
depth_estimator = DPTForDepthEstimation.from_pretrained("Intel/dpt-hybrid-midas", device_map='cuda')
ValueError: DPTForDepthEstimation does not support `device_map='cuda'` yet.
Thanks for your help!