experiments/sgd/detector/train_detector.py [96:106]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    try:
        genotype = eval('genotypes.%s' % args.arch)
        net_args = {'C': args.init_channels,  # 48 if genotype == DARTS else 128
                    'genotype': genotype,
                    'n_cells': args.layers,   # 14 if genotype == DARTS else 12
                    'C_mult': int(genotype != ViT) + 1,  # assume either ViT or DARTS-style architecture
                    'preproc': genotype != ViT,
                    'stem_type': 1}  # assume that the ImageNet-style stem is used by default
    except:
        deepnets = DeepNets1M(split=args.split,
                              nets_dir=args.data_dir,
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



experiments/sgd/train_net.py [62:72]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    try:
        genotype = eval('genotypes.%s' % args.arch)
        net_args = {'C': args.init_channels,
                    'genotype': genotype,
                    'n_cells': args.layers,
                    'C_mult': int(genotype != ViT) + 1,  # assume either ViT or DARTS-style architecture
                    'preproc': genotype != ViT,
                    'stem_type': 1}  # assume that the ImageNet-style stem is used by default
    except:
        deepnets = DeepNets1M(split=args.split,
                              nets_dir=args.data_dir,
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -