def get_model(model):
from models import resnet, base
if model == 'R32_C10':
rnet_checkpoint = 'cv/pretrained/R32_C10/pk_E_164_A_0.923.t7'
layer_config = [5, 5, 5]
rnet = resnet.FlatResNet32(base.BasicBlock,
layer_config,
num_classes=10)
agent = resnet.Policy32([1, 1, 1], num_blocks=15)
elif model == 'R110_C10':
rnet_checkpoint = 'cv/pretrained/R110_C10/pk_E_130_A_0.932.t7'
layer_config = [18, 18, 18]
rnet = resnet.FlatResNet32(base.BasicBlock,
layer_config,
num_classes=10)
agent = resnet.Policy32([1, 1, 1], num_blocks=54)
elif model == 'R32_C100':
rnet_checkpoint = 'cv/pretrained/R32_C100/pk_E_164_A_0.693.t7'
layer_config = [5, 5, 5]
rnet = resnet.FlatResNet32(base.BasicBlock,
layer_config,
num_classes=100)
agent = resnet.Policy32([1, 1, 1], num_blocks=15)
elif model == 'R110_C100':
rnet_checkpoint = 'cv/pretrained/R110_C100/pk_E_160_A_0.723.t7'
layer_config = [18, 18, 18]
rnet = resnet.FlatResNet32(base.BasicBlock,
layer_config,
num_classes=100)
agent = resnet.Policy32([1, 1, 1], num_blocks=54)
elif model == 'R101_ImgNet':
rnet_checkpoint = 'cv/pretrained/R101_ImgNet/ImageNet_R101_224_76.464'
layer_config = [3, 4, 23, 3]
rnet = resnet.FlatResNet224(base.Bottleneck,
layer_config,
num_classes=1000)
agent = resnet.Policy224([1, 1, 1, 1], num_blocks=33)
# elif model=='ResNext_C100':
# agent = resnet.Policy32([1,1,1], num_blocks=4*18)
# elif model=='ResNext_C10':
# agent = resnet.Policy32([1,1,1], num_blocks=4*18)
# load pretrained weights into flat ResNet
# rnet_checkpoint = torch.load(rnet_checkpoint)
# load_weights_to_flatresnet(rnet_checkpoint, rnet)
return agent
def get_resnet_model(dataset_type, blocks_list=[18, 18, 18]):
if dataset_type == Data.cifar10:
return resnet.FlatResNet32(base.BasicBlock,
blocks_list,
num_classes=10)
elif dataset_type == Data.cifar100:
return resnet.FlatResNet32(base.BasicBlock,
blocks_list,
num_classes=100)
elif dataset_type == Data.mini_imagenet:
return resnet.FlatResNet224(base.Bottleneck,
blocks_list,
num_classes=80)
else:
return resnet.FlatResNet224(base.Bottleneck,
blocks_list,
num_classes=1000)
print('E:%d Test Acc: %.3f' % (epoch, correct / total))
acc = acc + (correct/total)
# dataset and dataloader
trainset, testset = get_policy_datasets(Data.cifar10, Mode.with_policy, iter=args.iter)
trainloader = D.DataLoader(trainset, batch_size=args.batch_size, shuffle=True, num_workers=4)
testloader = D.DataLoader(testset, batch_size=256, shuffle=False, num_workers=4)
# global metric
best_test_acc = 0.0
epoch = 0
# configuration of the ResNet
if args.resume:
rnet = resnet.FlatResNet32(base.BasicBlock, [18, 18, 18], num_classes=10)
state = torch.load(os.path.join(savepath, 'rnet.t7'))
rnet.load_state_dict(state['net'])
best_test_acc = state['acc']
epoch = state['epoch']
else:
rnet = resnet.FlatResNet32(base.BasicBlock, [18, 18, 18], num_classes=10)
rnet_dict = torch.load(os.path.join(rootpath, 'rnet.t7'))
rnet.load_state_dict(rnet_dict['net'])
# Policy Network: ResNet-10
agent = Networks.PolicyNet10()
ckpt_dict = torch.load(os.path.join(rootpath, 'agent_best.t7'))
agent.load_state_dict(ckpt_dict['net'])
# Loss Fn and Optimizer
def get_model(model):
from models import resnet, base
if model == 'policy_satellite':
# agent = resnet.Policy224([1,1,1,1], num_blocks=289, num_feat=20)
# agent = resnet.Policy224GRU([1,1,1,1], num_blocks=289, num_feat=128)
# agent = resnet.PolicySeq()
agent = resnet.Policy2x2([1, 1, 1, 1], num_blocks=4)
rnet_hr = None
rnet_lr = None
elif model == 'R32_C10':
layer_config = [5, 5, 5]
rnet_hr = resnet.FlatResNet32(base.BasicBlock,
layer_config,
num_classes=10)
rnet_lr = resnet.FlatResNet32(base.BasicBlock,
layer_config,
num_classes=10)
agent = resnet.Policy32([1, 1, 1], num_blocks=16)
elif model == 'R32_C100':
layer_config = [5, 5, 5]
rnet_hr = resnet.FlatResNet32(base.BasicBlock,
layer_config,
num_classes=100)
rnet_lr = resnet.FlatResNet32(base.BasicBlock,
layer_config,
num_classes=100)
agent = resnet.Policy32([1, 1, 1], num_blocks=16)
elif model == 'R50_ImgNet':
agent = resnet.Policy224([1, 1, 1, 1], num_blocks=16)
""" High Res. Classifier """
rnet_hr = torchmodels.resnet50(pretrained=False)
set_parameter_requires_grad(rnet_hr, False)
num_ftrs = rnet_hr.fc.in_features
rnet_hr.fc = torch.nn.Linear(num_ftrs, 1000)
""" Low Res. Classifier """
rnet_lr = torchmodels.resnet50(pretrained=False)
set_parameter_requires_grad(rnet_lr, False)
num_ftrs = rnet_lr.fc.in_features
rnet_lr.fc = torch.nn.Linear(num_ftrs, 1000)
elif model == 'R34_fMoW':
agent = resnet.Policy224([1, 1, 1, 1], num_blocks=16)
""" High Res. Classifier """
rnet_hr = torchmodels.resnet34(pretrained=True)
set_parameter_requires_grad(rnet_hr, False)
num_ftrs = rnet_hr.fc.in_features
rnet_hr.fc = torch.nn.Linear(num_ftrs, 62)
""" Low Res. Classifier """
rnet_lr = torchmodels.resnet34(pretrained=True)
set_parameter_requires_grad(rnet_lr, False)
num_ftrs = rnet_lr.fc.in_features
rnet_lr.fc = torch.nn.Linear(num_ftrs, 62)
elif model == 'R34_CARS':
agent = resnet.Policy224([1, 1, 1, 1], num_blocks=16)
""" High Res. Classifier """
rnet_hr = torchmodels.resnet34(pretrained=True)
set_parameter_requires_grad(rnet_hr, False)
num_ftrs = rnet_hr.fc.in_features
rnet_hr.fc = torch.nn.Linear(num_ftrs, 196)
""" Low Res. Classifier """
rnet_lr = torchmodels.resnet34(pretrained=True)
set_parameter_requires_grad(rnet_lr, False)
num_ftrs = rnet_lr.fc.in_features
rnet_lr.fc = torch.nn.Linear(num_ftrs, 196)
return rnet_hr, rnet_lr, agent