#.o2z4bcj7i Inception-v3. This model was trained for several weeks on multiple GPUs on ImageNet. Transfer learning does not require GPUs to trainTraining across the full training set (40,000 images) took less than a minute on my Macbook Pro without GPU support. 1080 gpu 2개로 cifar10 학습 시간은 백만 steps에 20시간 정도 걸렸다.Image recognition 실행해보면, 1번 prediction에 2초쯤 걸리는 것 같다.