├── run.sh ├── Figs └── C3.jpg ├── utils ├── __pycache__ │ ├── network.cpython-37.pyc │ ├── network.cpython-38.pyc │ ├── resnet.cpython-37.pyc │ ├── resnet.cpython-38.pyc │ ├── transform.cpython-37.pyc │ ├── transform.cpython-38.pyc │ ├── contrastive_loss.cpython-37.pyc │ └── contrastive_loss.cpython-38.pyc ├── save_model.py └── yaml_config_hook.py ├── modules ├── __pycache__ │ ├── network.cpython-37.pyc │ ├── network.cpython-38.pyc │ ├── resnet.cpython-37.pyc │ ├── resnet.cpython-38.pyc │ ├── transform.cpython-37.pyc │ ├── transform.cpython-38.pyc │ ├── contrastive_loss.cpython-37.pyc │ └── contrastive_loss.cpython-38.pyc ├── contrastive_loss.py ├── network.py ├── transform.py └── resnet.py ├── evaluation ├── __pycache__ │ ├── evaluation.cpython-37.pyc │ └── evaluation.cpython-38.pyc └── evaluation.py ├── config └── config.yaml ├── train.py ├── CIFAR_10_log.out ├── README.md ├── main.py └── LICENSE /run.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/run.sh -------------------------------------------------------------------------------- /Figs/C3.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/Figs/C3.jpg -------------------------------------------------------------------------------- /utils/__pycache__/network.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/utils/__pycache__/network.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/network.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/utils/__pycache__/network.cpython-38.pyc -------------------------------------------------------------------------------- /utils/__pycache__/resnet.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/utils/__pycache__/resnet.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/resnet.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/utils/__pycache__/resnet.cpython-38.pyc -------------------------------------------------------------------------------- /modules/__pycache__/network.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/modules/__pycache__/network.cpython-37.pyc -------------------------------------------------------------------------------- /modules/__pycache__/network.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/modules/__pycache__/network.cpython-38.pyc -------------------------------------------------------------------------------- /modules/__pycache__/resnet.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/modules/__pycache__/resnet.cpython-37.pyc -------------------------------------------------------------------------------- /modules/__pycache__/resnet.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/modules/__pycache__/resnet.cpython-38.pyc -------------------------------------------------------------------------------- /utils/__pycache__/transform.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/utils/__pycache__/transform.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/transform.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/utils/__pycache__/transform.cpython-38.pyc -------------------------------------------------------------------------------- /modules/__pycache__/transform.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/modules/__pycache__/transform.cpython-37.pyc -------------------------------------------------------------------------------- /modules/__pycache__/transform.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/modules/__pycache__/transform.cpython-38.pyc -------------------------------------------------------------------------------- /evaluation/__pycache__/evaluation.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/evaluation/__pycache__/evaluation.cpython-37.pyc -------------------------------------------------------------------------------- /evaluation/__pycache__/evaluation.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/evaluation/__pycache__/evaluation.cpython-38.pyc -------------------------------------------------------------------------------- /utils/__pycache__/contrastive_loss.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/utils/__pycache__/contrastive_loss.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/contrastive_loss.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/utils/__pycache__/contrastive_loss.cpython-38.pyc -------------------------------------------------------------------------------- /modules/__pycache__/contrastive_loss.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/modules/__pycache__/contrastive_loss.cpython-37.pyc -------------------------------------------------------------------------------- /modules/__pycache__/contrastive_loss.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Armanfard-Lab/C3/HEAD/modules/__pycache__/contrastive_loss.cpython-38.pyc -------------------------------------------------------------------------------- /utils/save_model.py: -------------------------------------------------------------------------------- 1 | import os 2 | import torch 3 | 4 | 5 | def save_model(args, model, optimizer, current_epoch): 6 | out = os.path.join(args.model_path, "checkpoint_{}.tar".format(current_epoch)) 7 | state = {'net': model.state_dict(), 'optimizer': optimizer.state_dict(), 'epoch': current_epoch} 8 | torch.save(state, out) 9 | -------------------------------------------------------------------------------- /config/config.yaml: -------------------------------------------------------------------------------- 1 | # general 2 | seed: 42 3 | workers: 8 4 | dataset_dir: "./datasets" 5 | 6 | # train options 7 | dataset_size: 60000 8 | class_num: 10 9 | batch_size: 128 10 | test_batch_size: 500 11 | image_size: 224 12 | start_epoch: 0 13 | max_epochs: 20 14 | dataset: "CIFAR-10" 15 | 16 | 17 | # model options 18 | resnet: "ResNet34" # ResNet18 / ResNet34 / ResNet50 19 | feature_dim: 128 20 | model_path: "save/CIFAR-10" 21 | reload: False 22 | 23 | # loss options 24 | learning_rate: 0.00001 25 | weight_decay: 0. 26 | zeta: 0.6 27 | 28 | 29 | -------------------------------------------------------------------------------- /modules/contrastive_loss.py: -------------------------------------------------------------------------------- 1 | import torch 2 | 3 | 4 | def C3_loss(z_i, z_j, batch_size, zeta): 5 | 6 | z = torch.cat((z_i, z_j), dim=0) 7 | multiply = torch.matmul(z, z.T) 8 | 9 | a = torch.ones([batch_size]) 10 | mask = 2 * (torch.diag(a, -batch_size) + torch.diag(a, batch_size) + torch.eye(2 * batch_size)) 11 | mask = mask.cuda() 12 | 13 | exp_mul = torch.exp(multiply) 14 | numerator = torch.sum(torch.where((multiply + mask) > zeta, exp_mul, torch.zeros(multiply.shape).cuda()), dim=1) 15 | den = torch.sum(exp_mul, dim=1) 16 | 17 | return -torch.sum(torch.log(torch.div(numerator, den))) / batch_size 18 | -------------------------------------------------------------------------------- /utils/yaml_config_hook.py: -------------------------------------------------------------------------------- 1 | import os 2 | import yaml 3 | 4 | 5 | def yaml_config_hook(config_file): 6 | """ 7 | Custom YAML config loader, which can include other yaml files (I like using config files 8 | insteaad of using argparser) 9 | """ 10 | 11 | # load yaml files in the nested 'defaults' section, which include defaults for experiments 12 | with open(config_file) as f: 13 | cfg = yaml.safe_load(f) 14 | for d in cfg.get("defaults", []): 15 | config_dir, cf = d.popitem() 16 | cf = os.path.join(os.path.dirname(config_file), config_dir, cf + ".yaml") 17 | with open(cf) as f: 18 | l = yaml.safe_load(f) 19 | cfg.update(l) 20 | 21 | if "defaults" in cfg.keys(): 22 | del cfg["defaults"] 23 | 24 | return cfg 25 | -------------------------------------------------------------------------------- /train.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import torch 3 | from modules import transform, resnet, network, contrastive_loss 4 | from torch.nn.functional import normalize 5 | 6 | def train_net(net, data_loader, optimizer, batch_size, zeta): 7 | 8 | net.train() 9 | for param in net.parameters(): 10 | param.requires_grad = True 11 | 12 | for step, ((x_i, x_j), _) in enumerate(data_loader): 13 | optimizer.zero_grad() 14 | 15 | x_i = x_i.to('cuda') 16 | x_j = x_j.to('cuda') 17 | h_i = net.resnet(x_i) 18 | h_j = net.resnet(x_j) 19 | 20 | z_i = normalize(net.instance_projector(h_i), dim=1) 21 | z_j = normalize(net.instance_projector(h_j), dim=1) 22 | 23 | loss = contrastive_loss.C3_loss(z_i, z_j, batch_size, zeta) 24 | loss.backward() 25 | optimizer.step() 26 | 27 | return net , optimizer 28 | -------------------------------------------------------------------------------- /modules/network.py: -------------------------------------------------------------------------------- 1 | import torch.nn as nn 2 | import torch 3 | from torch.nn.functional import normalize 4 | 5 | 6 | class Network(nn.Module): 7 | def __init__(self, resnet, feature_dim, class_num): 8 | super(Network, self).__init__() 9 | self.resnet = resnet 10 | self.feature_dim = feature_dim 11 | self.cluster_num = class_num 12 | self.instance_projector = nn.Sequential( 13 | nn.Linear(self.resnet.rep_dim, self.resnet.rep_dim), 14 | nn.ReLU(), 15 | nn.Linear(self.resnet.rep_dim, self.feature_dim), 16 | ) 17 | self.cluster_projector = nn.Sequential( 18 | nn.Linear(self.resnet.rep_dim, self.resnet.rep_dim), 19 | nn.ReLU(), 20 | nn.Linear(self.resnet.rep_dim, self.cluster_num), 21 | nn.Softmax(dim=1) 22 | ) 23 | 24 | def forward(self, x_i, x_j): 25 | h_i = self.resnet(x_i) 26 | h_j = self.resnet(x_j) 27 | 28 | z_i = normalize(self.instance_projector(h_i), dim=1) 29 | z_j = normalize(self.instance_projector(h_j), dim=1) 30 | 31 | c_i = self.cluster_projector(h_i) 32 | c_j = self.cluster_projector(h_j) 33 | 34 | return z_i, z_j, c_i, c_j 35 | 36 | def forward_cluster(self, x): 37 | h = self.resnet(x) 38 | c = self.cluster_projector(h) 39 | c = torch.argmax(c, dim=1) 40 | return c 41 | -------------------------------------------------------------------------------- /CIFAR_10_log.out: -------------------------------------------------------------------------------- 1 | epoch: 0 2 | ACC: 0.7701166666666667 3 | NMI: 0.677791163171621 4 | ARI: 0.6069096232114414 5 | epoch: 1 6 | ACC: 0.7786 7 | NMI: 0.6878734537062069 8 | ARI: 0.620410202264679 9 | epoch: 2 10 | ACC: 0.8102 11 | NMI: 0.712980578724114 12 | ARI: 0.6608287202941374 13 | epoch: 3 14 | ACC: 0.8254166666666667 15 | NMI: 0.72907347209768 16 | ARI: 0.6857929928175632 17 | epoch: 4 18 | ACC: 0.8313 19 | NMI: 0.736224580530419 20 | ARI: 0.6956548604007671 21 | epoch: 5 22 | ACC: 0.8312166666666667 23 | NMI: 0.7367812041865238 24 | ARI: 0.6943580934273499 25 | epoch: 6 26 | ACC: 0.8334333333333334 27 | NMI: 0.7406088679019149 28 | ARI: 0.6981705807349273 29 | epoch: 7 30 | ACC: 0.8321666666666667 31 | NMI: 0.7409049056249303 32 | ARI: 0.6953900294612881 33 | epoch: 8 34 | ACC: 0.8354333333333334 35 | NMI: 0.7439649890104261 36 | ARI: 0.7017268059720909 37 | epoch: 9 38 | ACC: 0.83585 39 | NMI: 0.7446183697891985 40 | ARI: 0.7017045031570542 41 | epoch: 10 42 | ACC: 0.83675 43 | NMI: 0.7453026426639774 44 | ARI: 0.7034332748395484 45 | epoch: 11 46 | ACC: 0.83925 47 | NMI: 0.7488559796687856 48 | ARI: 0.7080651781065155 49 | epoch: 12 50 | ACC: 0.8382833333333334 51 | NMI: 0.7478013607075835 52 | ARI: 0.7056364975676672 53 | epoch: 13 54 | ACC: 0.8384833333333334 55 | NMI: 0.7483681698995862 56 | ARI: 0.706346528737891 57 | epoch: 14 58 | ACC: 0.8386333333333333 59 | NMI: 0.7497345846553385 60 | ARI: 0.7067066636097123 61 | epoch: 15 62 | ACC: 0.8392833333333334 63 | NMI: 0.7475620710498015 64 | ARI: 0.7084747302785441 65 | epoch: 16 66 | ACC: 0.838 67 | NMI: 0.7477611369036427 68 | ARI: 0.7057603770174042 69 | epoch: 17 70 | ACC: 0.8395333333333334 71 | NMI: 0.748683708617457 72 | ARI: 0.7084227476299111 73 | epoch: 18 74 | ACC: 0.8384 75 | NMI: 0.7460593651920591 76 | ARI: 0.70574805383224 77 | epoch: 19 78 | ACC: 0.8407166666666667 79 | NMI: 0.7496362702834038 80 | ARI: 0.7093040509830213 81 | -------------------------------------------------------------------------------- /modules/transform.py: -------------------------------------------------------------------------------- 1 | import torchvision 2 | import cv2 3 | import numpy as np 4 | 5 | 6 | class GaussianBlur: 7 | def __init__(self, kernel_size, min=0.1, max=2.0): 8 | self.min = min 9 | self.max = max 10 | self.kernel_size = kernel_size 11 | 12 | def __call__(self, sample): 13 | sample = np.array(sample) 14 | prob = np.random.random_sample() 15 | if prob < 0.5: 16 | sigma = (self.max - self.min) * np.random.random_sample() + self.min 17 | sample = cv2.GaussianBlur(sample, (self.kernel_size, self.kernel_size), sigma) 18 | return sample 19 | 20 | 21 | class Transforms: 22 | def __init__(self, size, s=1.0, mean=None, std=None, blur=False): 23 | self.train_transform = [ 24 | torchvision.transforms.RandomResizedCrop(size=size), 25 | torchvision.transforms.RandomHorizontalFlip(), 26 | torchvision.transforms.RandomApply([torchvision.transforms.ColorJitter(0.8 * s, 0.8 * s, 0.8 * s, 0.2 * s)], 27 | p=0.8), 28 | torchvision.transforms.RandomGrayscale(p=0.2), 29 | ] 30 | if blur: 31 | self.train_transform.append(GaussianBlur(kernel_size=23)) 32 | self.train_transform.append(torchvision.transforms.ToTensor()) 33 | self.test_transform = [ 34 | torchvision.transforms.Resize(size=(size, size)), 35 | torchvision.transforms.ToTensor(), 36 | ] 37 | if mean and std: 38 | self.train_transform.append(torchvision.transforms.Normalize(mean=mean, std=std)) 39 | self.test_transform.append(torchvision.transforms.Normalize(mean=mean, std=std)) 40 | self.train_transform = torchvision.transforms.Compose(self.train_transform) 41 | self.test_transform = torchvision.transforms.Compose(self.test_transform) 42 | 43 | def __call__(self, x): 44 | return self.train_transform(x), self.train_transform(x) 45 | -------------------------------------------------------------------------------- /evaluation/evaluation.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import torch 3 | from sklearn.metrics import normalized_mutual_info_score, adjusted_rand_score 4 | 5 | 6 | nmi = normalized_mutual_info_score 7 | ari = adjusted_rand_score 8 | 9 | def acc(y_true, y_pred): 10 | """ 11 | Calculate clustering accuracy. Require scikit-learn installed 12 | # Arguments 13 | y: true labels, numpy.array with shape `(n_samples,)` 14 | y_pred: predicted labels, numpy.array with shape `(n_samples,)` 15 | # Return 16 | accuracy, in [0,1] 17 | """ 18 | y_true = y_true.astype(np.int64) 19 | y_pred = y_pred.astype(np.int64) 20 | assert y_pred.size == y_true.size 21 | D = max(y_pred.max(), y_true.max()) + 1 22 | w = np.zeros((D, D), dtype=np.int64) 23 | for i in range(y_pred.size): 24 | w[y_pred[i], y_true[i]] += 1 25 | from scipy.optimize import linear_sum_assignment 26 | 27 | ind = linear_sum_assignment(w.max() - w) 28 | accuracy = 0.0 29 | for i in ind[0]: 30 | accuracy = accuracy + w[i, ind[1][i]] 31 | return accuracy / y_pred.size 32 | 33 | 34 | 35 | def net_evaluation(net,test_loader,dataset_size,test_batch_size = 500): 36 | 37 | pred_label_c = torch.zeros([dataset_size]).cuda() 38 | true_label_new = torch.zeros([dataset_size]) 39 | 40 | net.eval() 41 | for param in net.parameters(): 42 | param.requires_grad = False 43 | 44 | 45 | my_counter = 0 46 | for step, (x, y) in enumerate(test_loader): 47 | x = x.cuda() 48 | h = net.resnet(x) 49 | c = net.cluster_projector(h) 50 | c = torch.argmax(c, dim=1) 51 | 52 | pred_label_c[my_counter * test_batch_size:(my_counter + 1) * test_batch_size] = c 53 | true_label_new[my_counter * test_batch_size:(my_counter + 1) * test_batch_size] = y 54 | my_counter += 1 55 | 56 | my_acc = acc(true_label_new.cpu().numpy(), pred_label_c.cpu().numpy()) 57 | my_nmi = nmi(true_label_new.cpu().numpy(), pred_label_c.cpu().numpy()) 58 | my_ari = ari(true_label_new.cpu().numpy(), pred_label_c.cpu().numpy()) 59 | 60 | print("ACC:", my_acc) 61 | print("NMI:", my_nmi) 62 | print("ARI:", my_ari) -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # C3: Cross-instance guided Contrastive Clustering 2 | PyTorch implementation of the paper "C3: Cross-instance guided Contrastive Clustering" 3 | 4 |
Overview
5 | 6 | ## Implementation 7 | 8 | Please first downlaod the **`CIFAR_10_initial.zip`** from [this link](https://drive.google.com/file/d/1deqzG-eUztltgdQ0H2Y29N83_i9Tr_uN/view?usp=sharing) and put it in the same folder as **`main.py`** and then run the code. 9 | 10 | ## Citation 11 | 12 | You can find the preprint of our paper on [arXiv](https://arxiv.org/abs/2211.07136). 13 | 14 | Please cite our paper if you use the results or codes of our work. 15 | 16 | ``` 17 | @article{sadeghi2022c3, 18 | title={C3: Cross-instance guided Contrastive Clustering}, 19 | author={Sadeghi, Mohammadreza and Hojjati, Hadi and Armanfard, Narges}, 20 | journal={British Machine Vision Conference 2023}, 21 | year={2023} 22 | } 23 | ``` 24 | 25 | ## Abstract 26 | 27 | >Clustering is the task of gathering similar data samples into clusters without using any predefined labels. It has been widely studied in machine learning literature, and recent advancements in deep learning have revived interest in this field. Contrastive clustering (CC) models are a staple of deep clustering in which positive and negative pairs of each data instance are generated through data augmentation. CC models aim to learn a feature space where instance-level and cluster-level representations of positive pairs are grouped together. Despite improving the SOTA, these algorithms ignore the cross-instance patterns, which carry essential information for improving clustering performance. In this paper, we propose a novel contrastive clustering method, Cross-instance guided Contrastive Clustering (C3), that considers the cross-sample relationships to increase the number of positive pairs. In particular, we define a new loss function that identifies similar instances using the instance-level representation and encourages them to aggregate together. Extensive experimental evaluations show that our proposed method can outperform state-of-the-art algorithms on benchmark computer vision datasets: we improve the clustering accuracy by 6.8%, 2.8%, 4.9%, 1.3% and 0.4% on CIFAR-10, CIFAR-100, ImageNet-10, ImageNet-Dogs, and Tiny-ImageNet, respectively. 28 | -------------------------------------------------------------------------------- /main.py: -------------------------------------------------------------------------------- 1 | import os 2 | import numpy as np 3 | import torch 4 | import torchvision 5 | import argparse 6 | from collections import OrderedDict 7 | 8 | from modules import transform, resnet, network 9 | from utils import yaml_config_hook 10 | from torch.utils import data 11 | import torch.utils.data.distributed 12 | from evaluation import evaluation 13 | from train import train_net 14 | 15 | 16 | def main(): 17 | parser = argparse.ArgumentParser() 18 | config = yaml_config_hook.yaml_config_hook("config/config.yaml") 19 | 20 | for k, v in config.items(): 21 | parser.add_argument(f"--{k}", default=v, type=type(v)) 22 | args = parser.parse_args() 23 | if not os.path.exists(args.model_path): 24 | os.makedirs(args.model_path) 25 | 26 | torch.manual_seed(args.seed) 27 | torch.cuda.manual_seed_all(args.seed) 28 | torch.cuda.manual_seed(args.seed) 29 | np.random.seed(args.seed) 30 | 31 | # prepare data--------------------------------------------------------------------------------------------------------------------------------------------------- 32 | #train data 33 | train_dataset = torchvision.datasets.CIFAR10( 34 | root=args.dataset_dir, 35 | download=True, 36 | train=True, 37 | transform=transform.Transforms(size=args.image_size, s=0.5), 38 | ) 39 | test_dataset = torchvision.datasets.CIFAR10( 40 | root=args.dataset_dir, 41 | download=True, 42 | train=False, 43 | transform=transform.Transforms(size=args.image_size, s=0.5), 44 | ) 45 | dataset = data.ConcatDataset([train_dataset, test_dataset]) 46 | data_loader = torch.utils.data.DataLoader(dataset, batch_size=args.batch_size, shuffle=True, drop_last=True, 47 | pin_memory=True) 48 | 49 | 50 | # test data 51 | test_dataset_1 = torchvision.datasets.CIFAR10( 52 | root=args.dataset_dir, 53 | download=True, 54 | train=True, 55 | transform=transform.Transforms(size=args.image_size).test_transform, 56 | ) 57 | test_dataset_2 = torchvision.datasets.CIFAR10( 58 | root=args.dataset_dir, 59 | download=True, 60 | train=False, 61 | transform=transform.Transforms(size=args.image_size).test_transform, 62 | ) 63 | dataset_test = data.ConcatDataset([test_dataset_1, test_dataset_2]) 64 | test_loader = torch.utils.data.DataLoader( 65 | dataset=dataset_test, 66 | batch_size=args.test_batch_size, 67 | shuffle=False) 68 | 69 | 70 | # Initializing our network with a network trained with CC ------------------------------------------------------------------------------------------------------- 71 | res = resnet.get_resnet(args.resnet) 72 | net = network.Network(res, args.feature_dim, args.class_num) 73 | net = net.to('cuda') 74 | checkpoint = torch.load('CIFAR_10_initial', map_location=torch.device('cuda:0')) 75 | new_state_dict = OrderedDict() 76 | for k, v in checkpoint['net'].items(): 77 | name = k[7:] # remove `module.` 78 | new_state_dict[name] = v 79 | net.load_state_dict(new_state_dict) 80 | 81 | # optimizer --------------------------------------------------------------------------------------------------------------------------------------------- 82 | optimizer = torch.optim.Adam(net.parameters(), lr=args.learning_rate) 83 | 84 | # train loop --------------------------------------------------------------------------------------------------------------------------------------------------- 85 | for epoch in range(args.start_epoch, args.max_epochs): 86 | 87 | print("epoch:", epoch) 88 | evaluation.net_evaluation(net,test_loader,args.dataset_size, args.test_batch_size) 89 | net, optimizer = train_net(net, data_loader, optimizer, args.batch_size, args.zeta) 90 | 91 | state = {'net': net.state_dict(), 'optimizer': optimizer.state_dict(), 'epoch': epoch} 92 | with open('CIFAR_10_C3_loss_epoch_{}'.format(epoch), 'wb') as out: 93 | torch.save(state, out) 94 | 95 | 96 | if __name__ == "__main__": 97 | main() 98 | -------------------------------------------------------------------------------- /modules/resnet.py: -------------------------------------------------------------------------------- 1 | import torch.nn as nn 2 | import torch 3 | from torchvision.models.resnet import Bottleneck, BasicBlock, conv1x1 4 | 5 | 6 | class ResNet(nn.Module): 7 | 8 | def __init__(self, block, layers, num_classes=1000, zero_init_residual=False, 9 | groups=1, width_per_group=64, replace_stride_with_dilation=None, 10 | norm_layer=None): 11 | super(ResNet, self).__init__() 12 | if norm_layer is None: 13 | norm_layer = nn.BatchNorm2d 14 | self._norm_layer = norm_layer 15 | 16 | self.inplanes = 64 17 | self.dilation = 1 18 | if replace_stride_with_dilation is None: 19 | # each element in the tuple indicates if we should replace 20 | # the 2x2 stride with a dilated convolution instead 21 | replace_stride_with_dilation = [False, False, False] 22 | if len(replace_stride_with_dilation) != 3: 23 | raise ValueError("replace_stride_with_dilation should be None " 24 | "or a 3-element tuple, got {}".format(replace_stride_with_dilation)) 25 | self.groups = groups 26 | self.base_width = width_per_group 27 | self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=7, stride=2, padding=3, 28 | bias=False) 29 | self.bn1 = norm_layer(self.inplanes) 30 | self.relu = nn.ReLU(inplace=True) 31 | self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) 32 | self.layer1 = self._make_layer(block, 64, layers[0]) 33 | self.layer2 = self._make_layer(block, 128, layers[1], stride=2, 34 | dilate=replace_stride_with_dilation[0]) 35 | self.layer3 = self._make_layer(block, 256, layers[2], stride=2, 36 | dilate=replace_stride_with_dilation[1]) 37 | self.layer4 = self._make_layer(block, 512, layers[3], stride=2, 38 | dilate=replace_stride_with_dilation[2]) 39 | self.avgpool = nn.AdaptiveAvgPool2d((1, 1)) 40 | self.rep_dim = 512 * block.expansion 41 | 42 | for m in self.modules(): 43 | if isinstance(m, nn.Conv2d): 44 | nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') 45 | elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)): 46 | nn.init.constant_(m.weight, 1) 47 | nn.init.constant_(m.bias, 0) 48 | 49 | # Zero-initialize the last BN in each residual branch, 50 | # so that the residual branch starts with zeros, and each residual block behaves like an identity. 51 | # This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677 52 | if zero_init_residual: 53 | for m in self.modules(): 54 | if isinstance(m, Bottleneck): 55 | nn.init.constant_(m.bn3.weight, 0) 56 | elif isinstance(m, BasicBlock): 57 | nn.init.constant_(m.bn2.weight, 0) 58 | 59 | def _make_layer(self, block, planes, blocks, stride=1, dilate=False): 60 | norm_layer = self._norm_layer 61 | downsample = None 62 | previous_dilation = self.dilation 63 | if dilate: 64 | self.dilation *= stride 65 | stride = 1 66 | if stride != 1 or self.inplanes != planes * block.expansion: 67 | downsample = nn.Sequential( 68 | conv1x1(self.inplanes, planes * block.expansion, stride), 69 | norm_layer(planes * block.expansion), 70 | ) 71 | 72 | layers = [] 73 | layers.append(block(self.inplanes, planes, stride, downsample, self.groups, 74 | self.base_width, previous_dilation, norm_layer)) 75 | self.inplanes = planes * block.expansion 76 | for _ in range(1, blocks): 77 | layers.append(block(self.inplanes, planes, groups=self.groups, 78 | base_width=self.base_width, dilation=self.dilation, 79 | norm_layer=norm_layer)) 80 | 81 | return nn.Sequential(*layers) 82 | 83 | def _forward_impl(self, x): 84 | # See note [TorchScript super()] 85 | x = self.conv1(x) 86 | x = self.bn1(x) 87 | x = self.relu(x) 88 | x = self.maxpool(x) 89 | 90 | x = self.layer1(x) 91 | x = self.layer2(x) 92 | x = self.layer3(x) 93 | x = self.layer4(x) 94 | 95 | x = self.avgpool(x) 96 | x = torch.flatten(x, 1) 97 | 98 | return x 99 | 100 | def forward(self, x): 101 | return self._forward_impl(x) 102 | 103 | 104 | def get_resnet(name): 105 | resnet18 = ResNet(block=BasicBlock, layers=[2, 2, 2, 2]) 106 | resnet34 = ResNet(block=BasicBlock, layers=[3, 4, 6, 3]) 107 | resnet50 = ResNet(block=Bottleneck, layers=[3, 4, 6, 3]) 108 | 109 | resnets = { 110 | "ResNet18": resnet18, 111 | "ResNet34": resnet34, 112 | "ResNet50": resnet50, 113 | } 114 | if name not in resnets.keys(): 115 | raise KeyError(f"{name} is not a valid ResNet version") 116 | return resnets[name] 117 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Creative Commons Legal Code 2 | 3 | Attribution-NonCommercial-ShareAlike 3.0 Unported 4 | 5 | CREATIVE COMMONS CORPORATION IS NOT A LAW FIRM AND DOES NOT PROVIDE 6 | LEGAL SERVICES. DISTRIBUTION OF THIS LICENSE DOES NOT CREATE AN 7 | ATTORNEY-CLIENT RELATIONSHIP. CREATIVE COMMONS PROVIDES THIS 8 | INFORMATION ON AN "AS-IS" BASIS. CREATIVE COMMONS MAKES NO WARRANTIES 9 | REGARDING THE INFORMATION PROVIDED, AND DISCLAIMS LIABILITY FOR 10 | DAMAGES RESULTING FROM ITS USE. 11 | 12 | License 13 | 14 | THE WORK (AS DEFINED BELOW) IS PROVIDED UNDER THE TERMS OF THIS CREATIVE 15 | COMMONS PUBLIC LICENSE ("CCPL" OR "LICENSE"). THE WORK IS PROTECTED BY 16 | COPYRIGHT AND/OR OTHER APPLICABLE LAW. ANY USE OF THE WORK OTHER THAN AS 17 | AUTHORIZED UNDER THIS LICENSE OR COPYRIGHT LAW IS PROHIBITED. 18 | 19 | BY EXERCISING ANY RIGHTS TO THE WORK PROVIDED HERE, YOU ACCEPT AND AGREE 20 | TO BE BOUND BY THE TERMS OF THIS LICENSE. TO THE EXTENT THIS LICENSE MAY 21 | BE CONSIDERED TO BE A CONTRACT, THE LICENSOR GRANTS YOU THE RIGHTS 22 | CONTAINED HERE IN CONSIDERATION OF YOUR ACCEPTANCE OF SUCH TERMS AND 23 | CONDITIONS. 24 | 25 | 1. Definitions 26 | 27 | a. "Adaptation" means a work based upon the Work, or upon the Work and 28 | other pre-existing works, such as a translation, adaptation, 29 | derivative work, arrangement of music or other alterations of a 30 | literary or artistic work, or phonogram or performance and includes 31 | cinematographic adaptations or any other form in which the Work may be 32 | recast, transformed, or adapted including in any form recognizably 33 | derived from the original, except that a work that constitutes a 34 | Collection will not be considered an Adaptation for the purpose of 35 | this License. For the avoidance of doubt, where the Work is a musical 36 | work, performance or phonogram, the synchronization of the Work in 37 | timed-relation with a moving image ("synching") will be considered an 38 | Adaptation for the purpose of this License. 39 | b. "Collection" means a collection of literary or artistic works, such as 40 | encyclopedias and anthologies, or performances, phonograms or 41 | broadcasts, or other works or subject matter other than works listed 42 | in Section 1(g) below, which, by reason of the selection and 43 | arrangement of their contents, constitute intellectual creations, in 44 | which the Work is included in its entirety in unmodified form along 45 | with one or more other contributions, each constituting separate and 46 | independent works in themselves, which together are assembled into a 47 | collective whole. A work that constitutes a Collection will not be 48 | considered an Adaptation (as defined above) for the purposes of this 49 | License. 50 | c. "Distribute" means to make available to the public the original and 51 | copies of the Work or Adaptation, as appropriate, through sale or 52 | other transfer of ownership. 53 | d. "License Elements" means the following high-level license attributes 54 | as selected by Licensor and indicated in the title of this License: 55 | Attribution, Noncommercial, ShareAlike. 56 | e. "Licensor" means the individual, individuals, entity or entities that 57 | offer(s) the Work under the terms of this License. 58 | f. "Original Author" means, in the case of a literary or artistic work, 59 | the individual, individuals, entity or entities who created the Work 60 | or if no individual or entity can be identified, the publisher; and in 61 | addition (i) in the case of a performance the actors, singers, 62 | musicians, dancers, and other persons who act, sing, deliver, declaim, 63 | play in, interpret or otherwise perform literary or artistic works or 64 | expressions of folklore; (ii) in the case of a phonogram the producer 65 | being the person or legal entity who first fixes the sounds of a 66 | performance or other sounds; and, (iii) in the case of broadcasts, the 67 | organization that transmits the broadcast. 68 | g. "Work" means the literary and/or artistic work offered under the terms 69 | of this License including without limitation any production in the 70 | literary, scientific and artistic domain, whatever may be the mode or 71 | form of its expression including digital form, such as a book, 72 | pamphlet and other writing; a lecture, address, sermon or other work 73 | of the same nature; a dramatic or dramatico-musical work; a 74 | choreographic work or entertainment in dumb show; a musical 75 | composition with or without words; a cinematographic work to which are 76 | assimilated works expressed by a process analogous to cinematography; 77 | a work of drawing, painting, architecture, sculpture, engraving or 78 | lithography; a photographic work to which are assimilated works 79 | expressed by a process analogous to photography; a work of applied 80 | art; an illustration, map, plan, sketch or three-dimensional work 81 | relative to geography, topography, architecture or science; a 82 | performance; a broadcast; a phonogram; a compilation of data to the 83 | extent it is protected as a copyrightable work; or a work performed by 84 | a variety or circus performer to the extent it is not otherwise 85 | considered a literary or artistic work. 86 | h. "You" means an individual or entity exercising rights under this 87 | License who has not previously violated the terms of this License with 88 | respect to the Work, or who has received express permission from the 89 | Licensor to exercise rights under this License despite a previous 90 | violation. 91 | i. "Publicly Perform" means to perform public recitations of the Work and 92 | to communicate to the public those public recitations, by any means or 93 | process, including by wire or wireless means or public digital 94 | performances; to make available to the public Works in such a way that 95 | members of the public may access these Works from a place and at a 96 | place individually chosen by them; to perform the Work to the public 97 | by any means or process and the communication to the public of the 98 | performances of the Work, including by public digital performance; to 99 | broadcast and rebroadcast the Work by any means including signs, 100 | sounds or images. 101 | j. "Reproduce" means to make copies of the Work by any means including 102 | without limitation by sound or visual recordings and the right of 103 | fixation and reproducing fixations of the Work, including storage of a 104 | protected performance or phonogram in digital form or other electronic 105 | medium. 106 | 107 | 2. Fair Dealing Rights. Nothing in this License is intended to reduce, 108 | limit, or restrict any uses free from copyright or rights arising from 109 | limitations or exceptions that are provided for in connection with the 110 | copyright protection under copyright law or other applicable laws. 111 | 112 | 3. License Grant. Subject to the terms and conditions of this License, 113 | Licensor hereby grants You a worldwide, royalty-free, non-exclusive, 114 | perpetual (for the duration of the applicable copyright) license to 115 | exercise the rights in the Work as stated below: 116 | 117 | a. to Reproduce the Work, to incorporate the Work into one or more 118 | Collections, and to Reproduce the Work as incorporated in the 119 | Collections; 120 | b. to create and Reproduce Adaptations provided that any such Adaptation, 121 | including any translation in any medium, takes reasonable steps to 122 | clearly label, demarcate or otherwise identify that changes were made 123 | to the original Work. For example, a translation could be marked "The 124 | original work was translated from English to Spanish," or a 125 | modification could indicate "The original work has been modified."; 126 | c. to Distribute and Publicly Perform the Work including as incorporated 127 | in Collections; and, 128 | d. to Distribute and Publicly Perform Adaptations. 129 | 130 | The above rights may be exercised in all media and formats whether now 131 | known or hereafter devised. The above rights include the right to make 132 | such modifications as are technically necessary to exercise the rights in 133 | other media and formats. Subject to Section 8(f), all rights not expressly 134 | granted by Licensor are hereby reserved, including but not limited to the 135 | rights described in Section 4(e). 136 | 137 | 4. Restrictions. The license granted in Section 3 above is expressly made 138 | subject to and limited by the following restrictions: 139 | 140 | a. You may Distribute or Publicly Perform the Work only under the terms 141 | of this License. You must include a copy of, or the Uniform Resource 142 | Identifier (URI) for, this License with every copy of the Work You 143 | Distribute or Publicly Perform. You may not offer or impose any terms 144 | on the Work that restrict the terms of this License or the ability of 145 | the recipient of the Work to exercise the rights granted to that 146 | recipient under the terms of the License. You may not sublicense the 147 | Work. You must keep intact all notices that refer to this License and 148 | to the disclaimer of warranties with every copy of the Work You 149 | Distribute or Publicly Perform. When You Distribute or Publicly 150 | Perform the Work, You may not impose any effective technological 151 | measures on the Work that restrict the ability of a recipient of the 152 | Work from You to exercise the rights granted to that recipient under 153 | the terms of the License. This Section 4(a) applies to the Work as 154 | incorporated in a Collection, but this does not require the Collection 155 | apart from the Work itself to be made subject to the terms of this 156 | License. If You create a Collection, upon notice from any Licensor You 157 | must, to the extent practicable, remove from the Collection any credit 158 | as required by Section 4(d), as requested. If You create an 159 | Adaptation, upon notice from any Licensor You must, to the extent 160 | practicable, remove from the Adaptation any credit as required by 161 | Section 4(d), as requested. 162 | b. You may Distribute or Publicly Perform an Adaptation only under: (i) 163 | the terms of this License; (ii) a later version of this License with 164 | the same License Elements as this License; (iii) a Creative Commons 165 | jurisdiction license (either this or a later license version) that 166 | contains the same License Elements as this License (e.g., 167 | Attribution-NonCommercial-ShareAlike 3.0 US) ("Applicable License"). 168 | You must include a copy of, or the URI, for Applicable License with 169 | every copy of each Adaptation You Distribute or Publicly Perform. You 170 | may not offer or impose any terms on the Adaptation that restrict the 171 | terms of the Applicable License or the ability of the recipient of the 172 | Adaptation to exercise the rights granted to that recipient under the 173 | terms of the Applicable License. You must keep intact all notices that 174 | refer to the Applicable License and to the disclaimer of warranties 175 | with every copy of the Work as included in the Adaptation You 176 | Distribute or Publicly Perform. When You Distribute or Publicly 177 | Perform the Adaptation, You may not impose any effective technological 178 | measures on the Adaptation that restrict the ability of a recipient of 179 | the Adaptation from You to exercise the rights granted to that 180 | recipient under the terms of the Applicable License. This Section 4(b) 181 | applies to the Adaptation as incorporated in a Collection, but this 182 | does not require the Collection apart from the Adaptation itself to be 183 | made subject to the terms of the Applicable License. 184 | c. You may not exercise any of the rights granted to You in Section 3 185 | above in any manner that is primarily intended for or directed toward 186 | commercial advantage or private monetary compensation. The exchange of 187 | the Work for other copyrighted works by means of digital file-sharing 188 | or otherwise shall not be considered to be intended for or directed 189 | toward commercial advantage or private monetary compensation, provided 190 | there is no payment of any monetary compensation in con-nection with 191 | the exchange of copyrighted works. 192 | d. If You Distribute, or Publicly Perform the Work or any Adaptations or 193 | Collections, You must, unless a request has been made pursuant to 194 | Section 4(a), keep intact all copyright notices for the Work and 195 | provide, reasonable to the medium or means You are utilizing: (i) the 196 | name of the Original Author (or pseudonym, if applicable) if supplied, 197 | and/or if the Original Author and/or Licensor designate another party 198 | or parties (e.g., a sponsor institute, publishing entity, journal) for 199 | attribution ("Attribution Parties") in Licensor's copyright notice, 200 | terms of service or by other reasonable means, the name of such party 201 | or parties; (ii) the title of the Work if supplied; (iii) to the 202 | extent reasonably practicable, the URI, if any, that Licensor 203 | specifies to be associated with the Work, unless such URI does not 204 | refer to the copyright notice or licensing information for the Work; 205 | and, (iv) consistent with Section 3(b), in the case of an Adaptation, 206 | a credit identifying the use of the Work in the Adaptation (e.g., 207 | "French translation of the Work by Original Author," or "Screenplay 208 | based on original Work by Original Author"). The credit required by 209 | this Section 4(d) may be implemented in any reasonable manner; 210 | provided, however, that in the case of a Adaptation or Collection, at 211 | a minimum such credit will appear, if a credit for all contributing 212 | authors of the Adaptation or Collection appears, then as part of these 213 | credits and in a manner at least as prominent as the credits for the 214 | other contributing authors. For the avoidance of doubt, You may only 215 | use the credit required by this Section for the purpose of attribution 216 | in the manner set out above and, by exercising Your rights under this 217 | License, You may not implicitly or explicitly assert or imply any 218 | connection with, sponsorship or endorsement by the Original Author, 219 | Licensor and/or Attribution Parties, as appropriate, of You or Your 220 | use of the Work, without the separate, express prior written 221 | permission of the Original Author, Licensor and/or Attribution 222 | Parties. 223 | e. For the avoidance of doubt: 224 | 225 | i. Non-waivable Compulsory License Schemes. In those jurisdictions in 226 | which the right to collect royalties through any statutory or 227 | compulsory licensing scheme cannot be waived, the Licensor 228 | reserves the exclusive right to collect such royalties for any 229 | exercise by You of the rights granted under this License; 230 | ii. Waivable Compulsory License Schemes. In those jurisdictions in 231 | which the right to collect royalties through any statutory or 232 | compulsory licensing scheme can be waived, the Licensor reserves 233 | the exclusive right to collect such royalties for any exercise by 234 | You of the rights granted under this License if Your exercise of 235 | such rights is for a purpose or use which is otherwise than 236 | noncommercial as permitted under Section 4(c) and otherwise waives 237 | the right to collect royalties through any statutory or compulsory 238 | licensing scheme; and, 239 | iii. Voluntary License Schemes. The Licensor reserves the right to 240 | collect royalties, whether individually or, in the event that the 241 | Licensor is a member of a collecting society that administers 242 | voluntary licensing schemes, via that society, from any exercise 243 | by You of the rights granted under this License that is for a 244 | purpose or use which is otherwise than noncommercial as permitted 245 | under Section 4(c). 246 | f. Except as otherwise agreed in writing by the Licensor or as may be 247 | otherwise permitted by applicable law, if You Reproduce, Distribute or 248 | Publicly Perform the Work either by itself or as part of any 249 | Adaptations or Collections, You must not distort, mutilate, modify or 250 | take other derogatory action in relation to the Work which would be 251 | prejudicial to the Original Author's honor or reputation. Licensor 252 | agrees that in those jurisdictions (e.g. Japan), in which any exercise 253 | of the right granted in Section 3(b) of this License (the right to 254 | make Adaptations) would be deemed to be a distortion, mutilation, 255 | modification or other derogatory action prejudicial to the Original 256 | Author's honor and reputation, the Licensor will waive or not assert, 257 | as appropriate, this Section, to the fullest extent permitted by the 258 | applicable national law, to enable You to reasonably exercise Your 259 | right under Section 3(b) of this License (right to make Adaptations) 260 | but not otherwise. 261 | 262 | 5. Representations, Warranties and Disclaimer 263 | 264 | UNLESS OTHERWISE MUTUALLY AGREED TO BY THE PARTIES IN WRITING AND TO THE 265 | FULLEST EXTENT PERMITTED BY APPLICABLE LAW, LICENSOR OFFERS THE WORK AS-IS 266 | AND MAKES NO REPRESENTATIONS OR WARRANTIES OF ANY KIND CONCERNING THE 267 | WORK, EXPRESS, IMPLIED, STATUTORY OR OTHERWISE, INCLUDING, WITHOUT 268 | LIMITATION, WARRANTIES OF TITLE, MERCHANTABILITY, FITNESS FOR A PARTICULAR 269 | PURPOSE, NONINFRINGEMENT, OR THE ABSENCE OF LATENT OR OTHER DEFECTS, 270 | ACCURACY, OR THE PRESENCE OF ABSENCE OF ERRORS, WHETHER OR NOT 271 | DISCOVERABLE. SOME JURISDICTIONS DO NOT ALLOW THE EXCLUSION OF IMPLIED 272 | WARRANTIES, SO THIS EXCLUSION MAY NOT APPLY TO YOU. 273 | 274 | 6. Limitation on Liability. EXCEPT TO THE EXTENT REQUIRED BY APPLICABLE 275 | LAW, IN NO EVENT WILL LICENSOR BE LIABLE TO YOU ON ANY LEGAL THEORY FOR 276 | ANY SPECIAL, INCIDENTAL, CONSEQUENTIAL, PUNITIVE OR EXEMPLARY DAMAGES 277 | ARISING OUT OF THIS LICENSE OR THE USE OF THE WORK, EVEN IF LICENSOR HAS 278 | BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. 279 | 280 | 7. Termination 281 | 282 | a. This License and the rights granted hereunder will terminate 283 | automatically upon any breach by You of the terms of this License. 284 | Individuals or entities who have received Adaptations or Collections 285 | from You under this License, however, will not have their licenses 286 | terminated provided such individuals or entities remain in full 287 | compliance with those licenses. Sections 1, 2, 5, 6, 7, and 8 will 288 | survive any termination of this License. 289 | b. Subject to the above terms and conditions, the license granted here is 290 | perpetual (for the duration of the applicable copyright in the Work). 291 | Notwithstanding the above, Licensor reserves the right to release the 292 | Work under different license terms or to stop distributing the Work at 293 | any time; provided, however that any such election will not serve to 294 | withdraw this License (or any other license that has been, or is 295 | required to be, granted under the terms of this License), and this 296 | License will continue in full force and effect unless terminated as 297 | stated above. 298 | 299 | 8. Miscellaneous 300 | 301 | a. Each time You Distribute or Publicly Perform the Work or a Collection, 302 | the Licensor offers to the recipient a license to the Work on the same 303 | terms and conditions as the license granted to You under this License. 304 | b. Each time You Distribute or Publicly Perform an Adaptation, Licensor 305 | offers to the recipient a license to the original Work on the same 306 | terms and conditions as the license granted to You under this License. 307 | c. If any provision of this License is invalid or unenforceable under 308 | applicable law, it shall not affect the validity or enforceability of 309 | the remainder of the terms of this License, and without further action 310 | by the parties to this agreement, such provision shall be reformed to 311 | the minimum extent necessary to make such provision valid and 312 | enforceable. 313 | d. No term or provision of this License shall be deemed waived and no 314 | breach consented to unless such waiver or consent shall be in writing 315 | and signed by the party to be charged with such waiver or consent. 316 | e. This License constitutes the entire agreement between the parties with 317 | respect to the Work licensed here. There are no understandings, 318 | agreements or representations with respect to the Work not specified 319 | here. Licensor shall not be bound by any additional provisions that 320 | may appear in any communication from You. This License may not be 321 | modified without the mutual written agreement of the Licensor and You. 322 | f. The rights granted under, and the subject matter referenced, in this 323 | License were drafted utilizing the terminology of the Berne Convention 324 | for the Protection of Literary and Artistic Works (as amended on 325 | September 28, 1979), the Rome Convention of 1961, the WIPO Copyright 326 | Treaty of 1996, the WIPO Performances and Phonograms Treaty of 1996 327 | and the Universal Copyright Convention (as revised on July 24, 1971). 328 | These rights and subject matter take effect in the relevant 329 | jurisdiction in which the License terms are sought to be enforced 330 | according to the corresponding provisions of the implementation of 331 | those treaty provisions in the applicable national law. If the 332 | standard suite of rights granted under applicable copyright law 333 | includes additional rights not granted under this License, such 334 | additional rights are deemed to be included in the License; this 335 | License is not intended to restrict the license of any rights under 336 | applicable law. 337 | 338 | 339 | Creative Commons Notice 340 | 341 | Creative Commons is not a party to this License, and makes no warranty 342 | whatsoever in connection with the Work. Creative Commons will not be 343 | liable to You or any party on any legal theory for any damages 344 | whatsoever, including without limitation any general, special, 345 | incidental or consequential damages arising in connection to this 346 | license. Notwithstanding the foregoing two (2) sentences, if Creative 347 | Commons has expressly identified itself as the Licensor hereunder, it 348 | shall have all rights and obligations of Licensor. 349 | 350 | Except for the limited purpose of indicating to the public that the 351 | Work is licensed under the CCPL, Creative Commons does not authorize 352 | the use by either party of the trademark "Creative Commons" or any 353 | related trademark or logo of Creative Commons without the prior 354 | written consent of Creative Commons. Any permitted use will be in 355 | compliance with Creative Commons' then-current trademark usage 356 | guidelines, as may be published on its website or otherwise made 357 | available upon request from time to time. For the avoidance of doubt, 358 | this trademark restriction does not form part of this License. 359 | 360 | Creative Commons may be contacted at https://creativecommons.org/. 361 | --------------------------------------------------------------------------------