This command trains a ResNet50 with GN on ImageNet:
./imagenet-resnet-gn.py --data /path/to/imagenet
This script follows the exact setting in the original paper. Trained on 8 GPUs with a total batch size of 256, it achieves 24.0, 24.0, 24.1 top-1 validation accuracy in 3 independent runs, evaluated by the median of last 5 epochs. This matches the performance in the paper.
Training takes ~27 hours on 8 V100s. A training log is included.
The code is modified slightly from Tensorpack ResNet examples.
A trained model is available at Tensorpack model zoo.