def leaderboard_performance(submission_file_path): real = utils_lung.read_test_labels(pathfinder.TEST_LABELS_PATH) pred = parse_predictions(submission_file_path) real = collections.OrderedDict(sorted(real.iteritems())) pred = collections.OrderedDict(sorted(pred.iteritems())) check_validity(real, pred) return log_loss(real.values(), pred.values())
if n_selection: all_candidates = all_candidates[:n_selection] return all_candidates # data iterators batch_size = 1 train_valid_ids = utils.load_pkl(pathfinder.VALIDATION_SPLIT_PATH) train_pids, valid_pids, test_pids = train_valid_ids[ 'training'], train_valid_ids['validation'], train_valid_ids['test'] print('n train', len(train_pids)) print('n valid', len(valid_pids)) id2label = utils_lung.read_labels(pathfinder.LABELS_PATH) id2label_test = utils_lung.read_test_labels(pathfinder.TEST_LABELS_PATH) id2label_all = id2label.copy() id2label_all.update(id2label_test) train_data_iterator = data_iterators.DSBPatientsDataGenerator( data_path=pathfinder.DATA_PATH, batch_size=batch_size, transform_params=p_transform, n_candidates_per_patient=n_candidates_per_patient, data_prep_fun=data_prep_function_train, candidates_prep_fun=candidates_prep_function, id2candidates_path=id2candidates_path, id2label=id2label_all, rng=rng, patient_ids=train_pids, random=True,
if n_selection: all_candidates = all_candidates[:n_selection] return all_candidates # data iterators batch_size = 1 train_valid_ids = utils.load_pkl(pathfinder.VALIDATION_SPLIT_PATH) train_pids, valid_pids, test_pids, stage2_pids = train_valid_ids['training'], train_valid_ids['validation'], train_valid_ids['test'], train_valid_ids['test_stage2'] print 'n train', len(train_pids) print 'n valid', len(valid_pids) print 'n test', len(test_pids) all_pids = train_pids + valid_pids + test_pids id2label = utils_lung.read_labels(pathfinder.LABELS_PATH) id2label_test = utils_lung.read_test_labels(pathfinder.TEST_LABELS_PATH) id2label_all = id2label.copy() id2label_all.update(id2label_test) train_data_iterator = data_iterators.DSBPatientsDataGenerator(data_path=pathfinder.DATA_PATH, batch_size=batch_size, transform_params=p_transform, n_candidates_per_patient=n_candidates_per_patient, data_prep_fun=data_prep_function_train, candidates_prep_fun = candidates_prep_function, id2candidates_path=id2candidates_path, id2label = id2label_all, rng=rng,