code
stringlengths
17
6.64M
class LeaveOneOutSelectionMethod(SelectionMethod): 'Picks (hparams, step) by leave-one-out cross validation.' name = 'leave-one-domain-out cross-validation' @classmethod def _step_acc(self, records): 'Return the {val_acc, test_acc} for a group of records corresponding\n to a single ste...
def format_mean(data, latex): 'Given a list of datapoints, return a string describing their mean and\n standard error' if (len(data) == 0): return (None, None, 'X') mean = (100 * np.mean(list(data))) err = (100 * np.std((list(data) / np.sqrt(len(data))))) if latex: return (mean,...
def print_table(table, header_text, row_labels, col_labels, colwidth=10, latex=True): 'Pretty-print a 2D array of data, optionally with row/col labels' print('') if latex: num_cols = len(table[0]) print('\\begin{center}') print('\\adjustbox{max width=\\textwidth}{%') print(...
def print_results_tables(records, task, selection_method, latex): 'Given all records, print a results table for each dataset.' grouped_records = reporting.get_grouped_records(records, group_test_envs=(task != 'unsupervised_domain_generalization')).map((lambda group: {**group, 'sweep_acc': selection_method.swe...
def stage_path(data_dir, name): full_path = os.path.join(data_dir, name) if (not os.path.exists(full_path)): os.makedirs(full_path) return full_path
def download_and_extract(url, dst, remove=True): gdown.download(url, dst, quiet=False) if dst.endswith('.tar.gz'): tar = tarfile.open(dst, 'r:gz') tar.extractall(os.path.dirname(dst)) tar.close() if dst.endswith('.tar'): tar = tarfile.open(dst, 'r:') tar.extractall(...
def download_vlcs(data_dir): full_path = stage_path(data_dir, 'VLCS') download_and_extract('https://drive.google.com/uc?id=1skwblH1_okBwxWxmRsp9_qi15hyPpxg8', os.path.join(data_dir, 'VLCS.tar.gz'))
def download_mnist(data_dir): full_path = stage_path(data_dir, 'MNIST') MNIST(full_path, download=True)
def download_pacs(data_dir): full_path = stage_path(data_dir, 'PACS') download_and_extract('https://drive.google.com/uc?id=0B6x7gtvErXgfbF9CSk53UkRxVzg', os.path.join(data_dir, 'PACS.zip')) os.rename(os.path.join(data_dir, 'kfold'), full_path)
def download_office_home(data_dir): full_path = stage_path(data_dir, 'office_home') download_and_extract('https://drive.google.com/uc?id=0B81rNlvomiwed0V1YUxQdC1uOTg', os.path.join(data_dir, 'office_home.zip')) os.rename(os.path.join(data_dir, 'OfficeHomeDataset_10072016'), full_path)
def download_domain_net(data_dir): full_path = stage_path(data_dir, 'domain_net') urls = ['http://csr.bu.edu/ftp/visda/2019/multi-source/groundtruth/clipart.zip', 'http://csr.bu.edu/ftp/visda/2019/multi-source/infograph.zip', 'http://csr.bu.edu/ftp/visda/2019/multi-source/groundtruth/painting.zip', 'http://cs...
def download_terra_incognita(data_dir): full_path = stage_path(data_dir, 'terra_incognita') download_and_extract('https://lilablobssc.blob.core.windows.net/caltechcameratraps/eccv_18_all_images_sm.tar.gz', os.path.join(full_path, 'terra_incognita_images.tar.gz')) download_and_extract('https://lilablobssc....
def download_sviro(data_dir): full_path = stage_path(data_dir, 'sviro') download_and_extract('https://sviro.kl.dfki.de/?wpdmdl=1731', os.path.join(data_dir, 'sviro_grayscale_rectangle_classification.zip')) os.rename(os.path.join(data_dir, 'SVIRO_DOMAINBED'), full_path)
def todo_rename(records, selection_method, latex): grouped_records = reporting.get_grouped_records(records).map((lambda group: {**group, 'sweep_acc': selection_method.sweep_acc(group['records'])})).filter((lambda g: (g['sweep_acc'] is not None))) alg_names = Q(records).select('args.algorithm').unique() al...
class Job(): NOT_LAUNCHED = 'Not launched' INCOMPLETE = 'Incomplete' DONE = 'Done' def __init__(self, train_args, sweep_output_dir): args_str = json.dumps(train_args, sort_keys=True) args_hash = hashlib.md5(args_str.encode('utf-8')).hexdigest() self.output_dir = os.path.join(s...
def all_test_env_combinations(n): '\n For a dataset with n >= 3 envs, return all combinations of 1 and 2 test\n envs.\n ' assert (n >= 3) for i in range(n): (yield [i]) for j in range((i + 1), n): (yield [i, j])
def make_args_list(n_trials, dataset_names, algorithms, n_hparams_from, n_hparams, steps, data_dir, task, holdout_fraction, single_test_envs, hparams): args_list = [] for trial_seed in range(n_trials): for dataset in dataset_names: for algorithm in algorithms: if single_tes...
def ask_for_confirmation(): response = input('Are you sure? (y/n) ') if (not (response.lower().strip()[:1] == 'y')): print('Nevermind!') exit(0)
class Job(): NOT_LAUNCHED = 'Not launched' INCOMPLETE = 'Incomplete' DONE = 'Done' def __init__(self, train_args, sweep_output_dir, train_script='domainbed.scripts.train'): args_str = json.dumps(train_args, sort_keys=True) args_hash = hashlib.md5(args_str.encode('utf-8')).hexdigest() ...
def all_test_env_combinations(n): '\n For a dataset with n >= 3 envs, return all combinations of 1 and 2 test\n envs.\n ' assert (n >= 3) for i in range(n): (yield [i]) for j in range((i + 1), n): (yield [i, j])
def make_args_list(n_trials, dataset_names, algorithms, n_hparams_from, n_hparams, steps, data_dir, task, holdout_fraction, single_test_envs, wandb_proj, wandb_group, only_eval, always_rerun, warmstart_model_ckpt, hparams): args_list = [] for trial_seed in range(n_trials): for dataset in dataset_names...
def ask_for_confirmation(): response = input('Are you sure? (y/n) ') if (not (response.lower().strip()[:1] == 'y')): print('Nevermind!') exit(0)
class CLIPConLoss(nn.Module): 'CLIP text-image contrastive loss' def __init__(self, feature_dim, temperature=0.07, learnable_temperature=True, is_project=False, is_symmetric=True): super(CLIPConLoss, self).__init__() self.temperature = temperature if learnable_temperature: ...
def finite_mean(x): num_finite = torch.isfinite(x).float().sum() mean = torch.where(torch.isfinite(x), x, torch.tensor(0.0).to(x)).sum() if (num_finite != 0): mean = (mean / num_finite) else: return torch.tensor(0.0).to(x) return mean
class PLLogisticRegression(pl.LightningModule): '\n Logistic regression model\n ' def __init__(self, input_dim: int, num_classes: int, bias: bool=True, learning_rate: float=0.0001, optimizer: Optimizer=Adam, l1_strength: float=0.0, l2_strength: float=0.0, is_nonlinear: bool=False, **kwargs): "\...
def pretty(d, indent=0): for (key, value) in d.items(): print((('\t' * indent) + str(key))) if isinstance(value, dict): pretty(value, (indent + 1)) else: print((('\t' * (indent + 1)) + str(value)))
def get_record(output_dir): print('Loading records from:', output_dir) records = reporting.load_records(output_dir) print('Total records:', len(records)) return records
def get_results(out_dir, selection_method): 'Given all records, print a results table for each dataset.' records = get_record(out_dir) grouped_records = reporting.get_grouped_records(records, group_test_envs=True).map((lambda group: {**group, 'sweep_acc': selection_method.sweep_acc(group['records'], retur...
def plot_result(result_dict, plot_dataset, plot_y='acc_tgt', include=None, exclude=None, plot_std=False): plt.figure() sub_result_dict = result_dict[plot_dataset] plt_xs = [] plt_ys = [] plt_errs = [] for (k, v) in sub_result_dict.items(): subsub_result_dict = sub_result_dict[k] ...
def is_image_file(filename): return filename.lower().endswith(IMG_EXTENSIONS)
def make_dataset(dir, class_to_idx): images = [] dir = os.path.expanduser(dir) for target in sorted(class_to_idx.keys()): d = os.path.join(dir, target) if (not os.path.isdir(d)): continue for (root, _, fnames) in sorted(os.walk(d)): for fname in sorted(fname...
def pil_loader(path): with open(path, 'rb') as f: img = Image.open(f) return img.convert('RGB')
class CustomImageFolder(VisionDataset): def __init__(self, root, transform, perturbation_fn=None, idx_subsample_list=None): super().__init__(root, transform=transform, target_transform=None) (classes, class_to_idx) = self._find_classes(self.root) samples = make_dataset(self.root, class_to...
class DistributedSampler(Sampler): 'Sampler that restricts data loading to a subset of the dataset.\n It is especially useful in conjunction with\n :class:`torch.nn.parallel.DistributedDataParallel`. In such case, each\n process can pass a DistributedSampler instance as a DataLoader sampler,\n and loa...
class EvalSetting(): def __init__(self, name, dataset, size, perturbation_fn_cpu=None, perturbation_fn_gpu=None, metrics_fn=None, adversarial_attack=None, class_sublist=None, idx_subsample_list=None, parent_eval_setting=None, transform=None): super().__init__() self.name = name self.datas...
def accuracy_topk(logits, targets, topk=(1, 5)): maxk = max(topk) batch_size = targets.size(0) (_, pred) = logits.topk(maxk, 1, True, True) pred = pred.t() correct = pred.eq(targets.view(1, (- 1)).expand_as(pred)) res = {} for k in topk: correct_k = correct[:k].contiguous().view((-...
class StandardDataset(): def __init__(self, name=None, path=None): super().__init__() assert (bool(name) ^ bool(path)), 'Please specify one (and exactly one) of name or path' if (name is not None): assert (name in DATASET_NAMES), f'Dataset {name} is not recognized as an existi...
def corrupt_greyscale(image): return greyscale(image)
def accuracy_topk_subselected(logits, targets): targets = torch.tensor([class_sublist.index(x) for x in targets]) return accuracy_topk(logits, targets)
def corr_brightness_sev_1(image): return corruption_dict['brightness'](image, 0)
def corr_brightness_sev_2(image): return corruption_dict['brightness'](image, 1)
def corr_brightness_sev_3(image): return corruption_dict['brightness'](image, 2)
def corr_brightness_sev_4(image): return corruption_dict['brightness'](image, 3)
def corr_brightness_sev_5(image): return corruption_dict['brightness'](image, 4)
def corr_contrast_sev_1(image): return corruption_dict['contrast'](image, 0)
def corr_contrast_sev_2(image): return corruption_dict['contrast'](image, 1)
def corr_contrast_sev_3(image): return corruption_dict['contrast'](image, 2)
def corr_contrast_sev_4(image): return corruption_dict['contrast'](image, 3)
def corr_contrast_sev_5(image): return corruption_dict['contrast'](image, 4)
def corr_fog_sev_1(image): return corruption_dict['fog'](image, 0)
def corr_fog_sev_2(image): return corruption_dict['fog'](image, 1)
def corr_fog_sev_3(image): return corruption_dict['fog'](image, 2)
def corr_fog_sev_4(image): return corruption_dict['fog'](image, 3)
def corr_fog_sev_5(image): return corruption_dict['fog'](image, 4)
def corr_frost_sev_1(image): return corruption_dict['frost'](image, 0)
def corr_frost_sev_2(image): return corruption_dict['frost'](image, 1)
def corr_frost_sev_3(image): return corruption_dict['frost'](image, 2)
def corr_frost_sev_4(image): return corruption_dict['frost'](image, 3)
def corr_frost_sev_5(image): return corruption_dict['frost'](image, 4)
def corr_gaussian_blur_sev_1(image): return corruption_dict['gaussian_blur'](image, 0)
def corr_gaussian_blur_sev_2(image): return corruption_dict['gaussian_blur'](image, 1)
def corr_gaussian_blur_sev_3(image): return corruption_dict['gaussian_blur'](image, 2)
def corr_gaussian_blur_sev_4(image): return corruption_dict['gaussian_blur'](image, 3)
def corr_gaussian_blur_sev_5(image): return corruption_dict['gaussian_blur'](image, 4)
def corr_gaussian_noise_sev_1(image): return corruption_dict['gaussian_noise'](image, 0)
def corr_gaussian_noise_sev_2(image): return corruption_dict['gaussian_noise'](image, 1)
def corr_gaussian_noise_sev_3(image): return corruption_dict['gaussian_noise'](image, 2)
def corr_gaussian_noise_sev_4(image): return corruption_dict['gaussian_noise'](image, 3)
def corr_gaussian_noise_sev_5(image): return corruption_dict['gaussian_noise'](image, 4)
def corr_impulse_noise_sev_1(image): return corruption_dict['impulse_noise'](image, 0)
def corr_impulse_noise_sev_2(image): return corruption_dict['impulse_noise'](image, 1)
def corr_impulse_noise_sev_3(image): return corruption_dict['impulse_noise'](image, 2)
def corr_impulse_noise_sev_4(image): return corruption_dict['impulse_noise'](image, 3)
def corr_impulse_noise_sev_5(image): return corruption_dict['impulse_noise'](image, 4)
def corr_jpeg_compression_sev_1(image): return corruption_dict['jpeg_compression'](image, 0)
def corr_jpeg_compression_sev_2(image): return corruption_dict['jpeg_compression'](image, 1)
def corr_jpeg_compression_sev_3(image): return corruption_dict['jpeg_compression'](image, 2)
def corr_jpeg_compression_sev_4(image): return corruption_dict['jpeg_compression'](image, 3)
def corr_jpeg_compression_sev_5(image): return corruption_dict['jpeg_compression'](image, 4)
def corr_pixelate_sev_1(image): return corruption_dict['pixelate'](image, 0)
def corr_pixelate_sev_2(image): return corruption_dict['pixelate'](image, 1)
def corr_pixelate_sev_3(image): return corruption_dict['pixelate'](image, 2)
def corr_pixelate_sev_4(image): return corruption_dict['pixelate'](image, 3)
def corr_pixelate_sev_5(image): return corruption_dict['pixelate'](image, 4)
def corr_saturate_sev_1(image): return corruption_dict['saturate'](image, 0)
def corr_saturate_sev_2(image): return corruption_dict['saturate'](image, 1)
def corr_saturate_sev_3(image): return corruption_dict['saturate'](image, 2)
def corr_saturate_sev_4(image): return corruption_dict['saturate'](image, 3)
def corr_saturate_sev_5(image): return corruption_dict['saturate'](image, 4)
def corr_shot_noise_sev_1(image): return corruption_dict['shot_noise'](image, 0)
def corr_shot_noise_sev_2(image): return corruption_dict['shot_noise'](image, 1)
def corr_shot_noise_sev_3(image): return corruption_dict['shot_noise'](image, 2)
def corr_shot_noise_sev_4(image): return corruption_dict['shot_noise'](image, 3)
def corr_shot_noise_sev_5(image): return corruption_dict['shot_noise'](image, 4)
def corr_spatter_sev_1(image): return corruption_dict['spatter'](image, 0)
def corr_spatter_sev_2(image): return corruption_dict['spatter'](image, 1)
def corr_spatter_sev_3(image): return corruption_dict['spatter'](image, 2)
def corr_spatter_sev_4(image): return corruption_dict['spatter'](image, 3)
def corr_spatter_sev_5(image): return corruption_dict['spatter'](image, 4)
def corr_speckle_noise_sev_1(image): return corruption_dict['speckle_noise'](image, 0)