config.json

{
    "name": "ImageNet_LT_ResNeXt50_SADE",
    "n_gpu": 2,

    "arch": {
        "type": "ResNeXt50Model",
        "args": {
            "num_classes": 1000,
            "reduce_dimension": true,
            "use_norm": true,
            "returns_feat": true,
            "num_experts": 3
        }
    },
    "data_loader": {
        "type": "ImageNetLTDataLoader",
        "args":{
            "data_dir": "./data/imagenet",
            "batch_size": 64,
            "shuffle": true,
            "num_workers": 12
        }
    },
    "optimizer": {
        "type": "SGD",
        "args":{
            "lr": 0.025,
            "weight_decay": 5e-4,
            "momentum": 0.9,
            "nesterov": true
        }
    },
    "add_extra_info": true,
    "loss": {
        "type": "DiverseExpertLoss",
        "args":{
             "tau": 2
        }
    },
    "metrics": [
        "accuracy"
    ],
    "lr_scheduler": {
        "type": "CosineAnnealingLR",
        "args": {
            "T_max": 100,
            "eta_min": 0
        }
    },
    "trainer": {
        "epochs": 180,
        "save_dir": "saved/imagenet/sade_e180_inv2_bs64_lr0.025",
        "save_period": 180,
        "verbosity": 2,
        "monitor": "max val_accuracy",
        "tensorboard": false
    }
}