You can switch the default pre-trained model loading from GPUNet-0 to one of the following models listed below. Loads NVIDIA GPUNet-0 model by default pre-trained on IMAGENET dataset. device ( "cpu" ) print ( f 'Using for inference' ) Load Pretrained model device ( "cuda" ) ! nvidia - smi else : device = torch. filterwarnings ( 'ignore' ) % matplotlib inline if torch. Import torch from PIL import Image import ansforms as transforms import numpy as np import json import requests import matplotlib.pyplot as plt import warnings warnings. These are needed for preprocessing images and visualization. To run the example you need some extra python packages installed. You can switch the default pre-trained model loading from GPUNet-0 to one of these: GPUNet-1, GPUNet-2, GPUNet-P0, GPUNet-P1, GPUNet-D1 or GPUNet-D2. In the example below the pretrained GPUNet-0 model is loaded by default to perform inference on image and present the result. You can use this notebook to quickly load each one of listed models to perform inference runs. This notebook allows you to load and test all the the GPUNet model implementation listed in our CVPR-2022 paper. GPUNets are a new family of deployment and production ready Convolutional Neural Networks from NVIDIA auto-designed to max out the performance of NVIDIA GPU and TensorRT.Ĭrafted by NVIDIA AI using novel Neural Architecture Search(NAS) methods, GPUNet demonstrates state-of-the-art inference performance up to 2x faster than EfficientNet-X and FBNet-V3.
0 Comments
Leave a Reply. |