Relative Content

Tag Archive for pytorchresnet

The same backbone, one convergent and one divergent

I am training resnet, one implemented by mmdet and one by torchvision.As a result, mmdet cannot converge, but torchvision converges.I checked both architectures and they are identical. They use the same hyperparameters.

The same backbone, one convergent and one divergent

I am training resnet, one implemented by mmdet and one by torchvision.As a result, mmdet cannot converge, but torchvision converges.I checked both architectures and they are identical. They use the same hyperparameters.