def __init__(self, hp: HyperParams):
set_random_state(hp.random_seed)
self.hp = hp
self.device = get_device()
self.net, self.loaders, self.opt_inner = self.configure()
self.criterion = torch.nn.CrossEntropyLoss()
lr_schedule = LinearDecayLR(hp.lr_outer, hp.steps_outer)
self.opt_outer = ReptileSGD(self.net, lr_schedule)
self.samples = {
s: next(iter(loader))
for s, loader in self.loaders.items()
}
self.support = (torch.empty(0), torch.empty(0))
self.save_path = f"{self.hp.root}/{self.hp.to_string()}.pt"
print(dict(save_path=self.save_path))
def __init__(
self,
hp: HyperParams,
loaders: Dict[str, MetaLoader],
net: torch.nn.Module,
use_gpu=True,
):
self.hp = hp
self.loaders = loaders
self.device = get_device(use_gpu)
self.rng = set_random_state()
self.net = net.to(self.device)
self.criterion = torch.nn.CrossEntropyLoss()
self.opt_inner = torch.optim.SGD(self.net.parameters(), lr=hp.lr_inner)
lr_schedule = LinearDecayLR(hp.lr_outer, hp.steps_outer)
self.opt_outer = ReptileSGD(self.net, lr_schedule)
self.batch_val = next(iter(self.loaders[Splits.val]))
def __init__(
self,
hparams: HyperParams,
loaders: Dict[str, torchmeta.utils.data.BatchMetaDataLoader],
net: torch.nn.Module,
):
self.hparams = hparams
self.loaders = loaders
self.device = get_device()
self.rng = set_random_state()
self.net = net.to(self.device)
self.criterion = torch.nn.CrossEntropyLoss()
self.opt_inner = torch.optim.SGD(self.net.parameters(),
lr=hparams.lr_inner)
lr_schedule = LinearDecayLR(hparams.lr_outer, hparams.steps_outer)
self.opt_outer = ReptileSGD(self.net,
lr_schedule,
num_accum=hparams.bs_outer)
self.batch_val = next(iter(self.loaders[Splits.val]))
parser.add_argument("--hid-dim",
type=int,
default=64,
help='Hidden layer dimensionalities.')
parser.add_argument('--batchnorm',
action='store_true',
default=False,
help="batchnorm")
parser.add_argument('--dropout', type=float, default=0.8, help='dropout')
parser.add_argument('--drop_bef',
action='store_true',
default=True,
help='Location of dropout')
args = parser.parse_args()
print(args)
acc_lists = []
random_seeds = generate_random_seeds(seed=1222, nums=args.runs)
for run in range(args.runs):
set_random_state(random_seeds[run])
acc_lists.append(main(args))
acc_lists = np.array(acc_lists)
mean = np.around(np.mean(acc_lists, axis=0), decimals=4)
std = np.around(np.std(acc_lists, axis=0), decimals=4)
print('Total acc: ', acc_lists)
print('mean', mean)
print('std', std)