KaneAI - World’s First E2E Software Testing Agent.Book a Demo Join TestMu Conference

How to use start_example method in hypothesis

Best Python code snippet using hypothesis

datasets.py

Source:datasets.py

1#TODO : use a Timeseries Generator of Keras : https://keras.io/preprocessing/sequence/2__author__ = 'AlexBioY'3import numpy as np4import importlib5import os6from collections import namedtuple7from src.data.data_sources import get_combined_cleaned_onecoin_df8from src.data.settings import  DATASET_TRANSFORM9import logging10logger = logging.getLogger(__name__)11logger.setLevel(logging.DEBUG)12# TODO: do it smarter (use keras function ot scipy) or use matrix multiplication13def _normalize_dataset(X):14    for example in range(X.shape[0]):15        X[example, :, 0] = (X[example, :, 0] - X[example, -1, 0]) / (np.max(X[example, :, 0]) - np.min(X[example, :, 0]))16        X[example, :, 1] = (X[example, :, 1] - X[example, -1, 1]) / (np.max(X[example, :, 1]) - np.min(X[example, :, 1]))17        X[example, :, 2] = (X[example, :, 2] - X[example, -1, 2]) / (np.max(X[example, :, 2]) - np.min(X[example, :, 2]))18        X[example, :, 3] = (X[example, :, 3] - X[example, -1, 3]) / (np.max(X[example, :, 3]) - np.min(X[example, :, 3]))19    return X20def df_to_X_onecoin(data_df, ds_transform):21    '''22    Transform an input ts into array [ examples, time points back fatures (LSTM modules), feature dimension ],23    Labels computing can be set to a different function24    '''25    win_size = DATASET_TRANSFORM[ds_transform].win_size26    stride = DATASET_TRANSFORM[ds_transform].stride27    label_func = DATASET_TRANSFORM[ds_transform].label_func28    num_classes = DATASET_TRANSFORM[ds_transform].num_classes29    future = DATASET_TRANSFORM[ds_transform].future30    res_period = DATASET_TRANSFORM[ds_transform].res_period31    n = len(data_df)32    if (n - win_size) < 0:33        logger.error("   DATASET is smaller then win_size! we need more data")34    num_examples = int((n - win_size) / stride) # how many times we can srtide via the timeseries (number of possible examples)35    # (4968, 96, 1)36    predictors = data_df.shape[1]  # make prediction based on multivatiate ts, price and volume37    label_dummy_classes = num_classes38    data_set = np.zeros([num_examples, win_size, predictors])39    labels = np.zeros([num_examples, label_dummy_classes])40    # form training examples by shifting triugh the dataset41    logger.info("   One coin: Converting dataframe to dataset array,  " + str(num_examples) + " examples")42    for start_example in range(0, num_examples):43        end_example = start_example + win_size44        # build X array45        data_set[start_example, :, 0] = data_df[start_example:end_example]['price'].values.reshape([-1, 1])[:, 0]46        data_set[start_example, :, 1] = data_df[start_example:end_example]['volume'].values.reshape([-1, 1])[:, 0]47        data_set[start_example, :, 2] = data_df[start_example:end_example]['price_var'].values.reshape([-1, 1])[:, 0]48        data_set[start_example, :, 3] = data_df[start_example:end_example]['volume_var'].values.reshape([-1, 1])[:, 0]49        data_set[start_example, :, 4] = data_df[start_example:end_example]['price_max'].values.reshape([-1, 1])[:, 0]50        data_set[start_example, :, 5] = data_df[start_example:end_example]['price_min'].values.reshape([-1, 1])[:, 0]51        #TODO: add blockchain info here52        # here we set the future values either to following proce values of price-max values (in case we predict max53        if label_func == 'label_3class_max_hit':54            future_values = data_df[end_example-1 : end_example + future]['price_max']55            threshold_1 = DATASET_TRANSFORM[ds_transform].threshold_156            threshold_2 = DATASET_TRANSFORM[ds_transform].threshold_257        elif label_func == 'label_3class_min_hit':58            future_values = data_df[end_example-1 : end_example + future]['price_min']59            threshold_1 = DATASET_TRANSFORM[ds_transform].threshold_160            threshold_2 = DATASET_TRANSFORM[ds_transform].threshold_261        elif label_func == 'label_2class_max_hit':62            future_values = data_df[end_example - 1: end_example + future]['price_min']63            threshold_1 = DATASET_TRANSFORM[ds_transform].threshold_164            threshold_2 = None65        else:66            future_values = data_df[end_example-1 : end_example + future]['price'] # we also need the last price from example67            threshold_1 = DATASET_TRANSFORM[ds_transform].return_target68            threshold_2 = None69        #build Y array (labels)70        module = importlib.import_module('src.data.datasets')71        func_obj = getattr(module, label_func)72        labels[start_example, :] = func_obj(future_values, threshold_1, threshold_2)73        # assert the array dimencions74        if start_example % 3000 == 0:75            logger.info("   ... df->array examples completed: " + str(start_example))76    logger.info("   One coin: finished.")77    return data_set, labels78def label_3class_return_target(future_prices, threshold_1, threshold_2):79    '''80    calculate a dummy class number out of 90 future prices as 0 - same / 1 - up / 2 - down81    '''82    return_target = threshold_183    # 0 -same, 1-up, 2 -down84    label_dummy_classes=385    open_price = future_prices[0]86    close_price = future_prices[-1]87    price_return = close_price - open_price88    percentage_return = 1 - (open_price - price_return) / open_price89    label = 0 if (abs(percentage_return) < return_target) else np.sign(percentage_return)90    dummy_labels = np.zeros([1,label_dummy_classes]).astype(int)91    # 0 - same / 1 - up / 2 - down92    if label == 0:93        dummy_labels[0, 0] = 194    elif label == 1:95        dummy_labels[0, 1] = 196    elif label == -1:97        dummy_labels[0, 2] = 198    return dummy_labels99def label_2class_return_target(future_prices, threshold_1, threshold_2):100    # NOTE: return tagret  is ignored here101    return_target = threshold_1102    # 1 - up, - 1 - down103    label_dummy_classes = 2104    open_price = future_prices[0]105    close_price = future_prices[-1]106    price_return = close_price - open_price107    percentage_return = 1 - (open_price - price_return) / open_price108    label = np.sign(percentage_return)109    dummy_labels = np.zeros([1, label_dummy_classes]).astype(int)110    # 0 - same / 1 - up / 2 - down111    if label == 1:112        dummy_labels[0, 0] = 1113    elif label == -1:114        dummy_labels[0, 1] = 1115    return dummy_labels116def label_3class_max_hit(future_prices, threshold_1, threshold_2):117    # 0 -same, 1-threshold_1, 2 -threshold_2118    label_dummy_classes=3119    open_price = future_prices[0]120    close_price = future_prices[-1]121    # min_price = np.min(future_prices)122    # percent_min = 1 - (open_price - min_price) / open_price123    max_price = np.max(future_prices)124    percent_max = (max_price-open_price) / open_price125    if  percent_max > threshold_2:126        label = 2127    elif percent_max > threshold_1:128        label = 1129    else:130        label = 0131    dummy_labels = np.zeros([1,label_dummy_classes]).astype(int)132    # 0 - same / 1 - up / 2 - down133    if label == 0:134        dummy_labels[0, 0] = int(1)135    elif label == 1:136        dummy_labels[0, 1] = int(1)137    elif label == 2:138        dummy_labels[0, 2] = int(1)139    return dummy_labels140def label_2class_max_hit(future_prices, threshold_1, threshold_2):141    # create a label set as follows:142    # label 1 (maxhit) if anytime within future_prices the price max value exceeded threshold_1 value143    # (i.e there was an opportunity to trade144    # labels: 0 -same, 1-threshold_1145    label_dummy_classes=2146    open_price = future_prices[0]147    close_price = future_prices[-1]148    # min_price = np.min(future_prices)149    # percent_min = 1 - (open_price - min_price) / open_price150    max_price = np.max(future_prices)151    percent_max = (max_price-open_price) / open_price152    if percent_max > threshold_1:153        label = 1154    else:155        label = 0156    dummy_labels = np.zeros([1,label_dummy_classes]).astype(int)157    # 0 - same / 1 - up / 2 - down158    if label == 0:159        dummy_labels[0, 0] = int(1)160    elif label == 1:161        dummy_labels[0, 1] = int(1)162    return dummy_labels163def combine_all_coins(COINS_LIST, db_name, ds_transform):164    '''165    Build the a full dataset X, Y by fusind all datasets of each coin from COIN_LIST166    - for each pair get ts of price and volume, calculate variance and build a df [time, price, vol, price_var, vol_var]167    - split this ts into pieces of win_size ad calculate a label for each168    - pile them up int one dataset169    '''170    res_period = DATASET_TRANSFORM[ds_transform].res_period171    X = []  # (147319, 200, 4) - 4 is price, volume, price_var, volume_var172    Y = []  # (147319, 3)  - 3 is number of classes173    logger.info(" > Form data set X array from a coin list:" + str(COINS_LIST))174    for transaction_coin, counter_coin in COINS_LIST:175        # retrieve a time series df from DB as [time,price,volume, price_var, volume_var]176        data_df = get_combined_cleaned_onecoin_df(db_name, transaction_coin, counter_coin, res_period)177        # TODO: cut all old data from period of high volatility178        data_df = data_df.loc['2018-03-01':]179        # convert this df into a array of shape of (147319, 200, 4) = (examples, time_back, features)180        # all parameters of data transformation are in data.settings181        X_train_one, Y_train_one = df_to_X_onecoin(data_df, ds_transform)182        del data_df183        # pile up into one array184        if X == []:185            X = X_train_one186            Y = Y_train_one187        else:188            X = np.concatenate((X, X_train_one), axis=0)189            Y = np.concatenate((Y, Y_train_one), axis=0)190    # delete all examples with NaN inside191    idx2delete = []192    for n in range(X.shape[0] - 1):193        if np.isnan(X[n, :, :]).any():194            idx2delete.append(n)195    X = np.delete(X, (idx2delete), axis=0)196    Y = np.delete(Y, (idx2delete), axis=0)197    if DATASET_TRANSFORM[ds_transform].num_classes == 3:198        logger.info("> X,Y Datasets have been built: same= " + str(sum(Y[:,0])) + ' | UP= ' + str(sum(Y[:,1])) + ' | DOWN= ' + str(sum(Y[:,2])))199    elif DATASET_TRANSFORM[ds_transform].num_classes == 2:200        logger.info("> X,Y Datasets have been built: UP= " + str(sum(Y[:, 0])) + ' | DOWN= ' + str( sum(Y[:, 1])) )201    else:202        logger.error("UNKNOWN NUMBER OF CLASSES!")203    # normalize204    # TODO: can I do it in-place?205    X = _normalize_dataset(X)206    # sanity check207    logger.info("   ... Sanity Checking for NaN in dataset: check for any nan")208    for n in range(X.shape[0]):209        if np.isnan(X[n, :, :]).any():210            logger.info(n)211    logger.info("=======> final X dataset shape: " + str(X.shape))212    logger.info("=======> final Y dataset shape: " + str(Y.shape))213    # # TODO: check if folder is exists214    # np.save("data/processed/"+fname_x, X)215    # np.save("data/processed/"+fname_y, Y)...

run_glrp_grid_mnist.py

Source:run_glrp_grid_mnist.py

1#!python2"""3Running the GLRP on GCNN model trained on MNIST data. Digits are graph signals on 8-nearest neighbor graph.4The model can be retrained uncommenting a part of the code.5"""6import numpy as np7from matplotlib import pyplot as plt8from sklearn.metrics import accuracy_score, f1_score9from tensorflow.examples.tutorials.mnist import input_data10from components import glrp_scipy, visualize_mnist11from lib import models, graph, coarsening12import time13COARSENING_LEVELS = 4  # to satisfy pooling of size 4 two times we need 4 level14DIR_DATA = "./data/mnist"15METRIC = 'euclidean'16NUMBER_EDGES = 817M = 28  # size of the digit's picture side18FEATURE_NUM = M * M19EPS = 1e-7  # for adjacency matrix20if __name__ == "__main__":21    # !!!22    # creating the adjacency matrix with all the non-zero weights equal to 123    z = graph.grid(M)24    dist, idx = graph.distance_sklearn_metrics(z, k=NUMBER_EDGES, metric=METRIC)25    A = graph.adjacency(dist, idx)26    A[A > EPS] = 127    graphs, perm = coarsening.coarsen(A, levels=COARSENING_LEVELS, self_connections=False)28    L = [graph.laplacian(A, normalized=True) for A in graphs]29    mnist = input_data.read_data_sets(DIR_DATA, one_hot=False)30    train_data = mnist.train.images.astype(np.float32)31    val_data = mnist.validation.images.astype(np.float32)32    test_data = mnist.test.images.astype(np.float32)33    train_labels = mnist.train.labels34    val_labels = mnist.validation.labels35    test_labels = mnist.test.labels36    train_data = coarsening.perm_data(train_data, perm)37    val_data = coarsening.perm_data(val_data, perm)38    test_data = coarsening.perm_data(test_data, perm)39    common = {}40    common['dir_name'] = 'mnist_grid_ones/'41    common['num_epochs'] = 3042    common['batch_size'] = 10043    common['decay_steps'] = mnist.train.num_examples / common['batch_size']44    common['eval_frequency'] = 30 * common['num_epochs']45    common['brelu'] = 'b1relu'46    common['pool'] = 'mpool1'47    C = max(mnist.train.labels) + 1  # number of classes48    common['regularization'] = 5e-449    common['dropout'] = 0.550    common['learning_rate'] = 0.0351    common['decay_rate'] = 0.9552    common['momentum'] = 0.953    common['F'] = [32, 64]54    common['K'] = [25, 25]55    common['p'] = [4, 4]56    common['M'] = [512, C]57    name = 'cgconv_cgconv_softmax_momentum'58    params = common.copy()59    params['dir_name'] += name60    params['filter'] = 'chebyshev5'61    model = models.cgcnn(L, **params)62    # !!!63    # To train again uncomment this part64    # start = time.time()65    # accuracy, loss, t_step, trained_losses = model.fit(train_data, train_labels, val_data, val_labels)66    # end = time.time()67    probas_ = model.get_probabilities(test_data)68    f1 = 100 * f1_score(test_labels, np.argmax(probas_, axis=1), average='weighted')69    acc = 100 * accuracy_score(test_labels, np.argmax(probas_, axis=1))70    print("\n\tTest F1 weighted: ", f1)71    print("\tTest Accuraccy:", acc, "\n")72    data_to_test = val_data[0:common["batch_size"], ]73    probas_ = model.get_probabilities(data_to_test)74    labels_by_network = np.argmax(probas_, axis=1)75    labels_data_to_test = val_labels[0:common["batch_size"], ]76    I = np.eye(10)77    labels_hot_encoded = I[labels_by_network]78    glrp = glrp_scipy.GraphLayerwiseRelevancePropagation(model, data_to_test, labels_hot_encoded)79    rel = glrp.get_relevances()[-1]  # getting the relevances corresponding to the input layer80    data_to_test = coarsening.perm_data_back(data_to_test, perm, FEATURE_NUM)81    rel = coarsening.perm_data_back(rel, perm, FEATURE_NUM)82    results_dir = './figures/'83    start_example = 984    end_example = 1785    visualize_mnist.plot_numbers(data_to_test[start_example:end_example, ], rel[start_example:end_example, ],86                                 labels_data_to_test[start_example:end_example, ],87                                 labels_by_network[start_example:end_example, ], results_dir)88    # start_example = 989    # end_example = 1790    for i in range(start_example, end_example):  # 9, 1791        heatmap = visualize_mnist.get_heatmap(rel[i,])92        fig = plt.figure()93        ax = fig.add_subplot(111)94        ax.axis('off')95        ax.imshow(heatmap, cmap='Reds', interpolation='bilinear')...

main.py

Source:main.py

...13  os.system('termux-info | grep Linux>f.txt || echo not>f.txt')14  versio = str(open('f.txt').read())15  if 'Linux' in versio:16    startup_termux()17    start_example()18  else:19    startup_linux()20    lol = start_example() 21    sleep(5)22    killprocess(lol)23def start_example():24  starting_command = subprocess.run(["python3", "lo.py"], capture_output=True)25  print(starting_command)26  return(starting_command)27def killprocess(starting):28  starting.kill29  ...

Automation Testing Tutorials

Learn to execute automation testing from scratch with LambdaTest Learning Hub. Right from setting up the prerequisites to run your first automation test, to following best practices and diving deeper into advanced test scenarios. LambdaTest Learning Hubs compile a list of step-by-step guides to help you be proficient with different test automation frameworks i.e. Selenium, Cypress, TestNG etc.