How to use test_hi method in hypothesis

Best Python code snippet using hypothesis

bmd_final.py

Source:bmd_final.py

1import click2import mlflow3import logging4import os5import os.path6import shutil7import numpy as np8import xarray as xr9import tensorflow as tf10import matplotlib.pyplot as plt11import climdex.temperature as tdex12import climdex.precipitation as pdex13import experiments.maxt_experiment_base as maxt14import experiments.prcp_experiment_base as prcp15from utils.plot import image_map_factory16from utils.preprocessing import remove_monthly_means17from utils.distributions import normal, bernoulli_gamma18from utils.data import create_time_series_train_test_generator_v219from baselines.dscnn import create_bmd_cnn1020from normalizing_flows.models import VariationalModel21from normalizing_flows.utils import get_metrics22from experiments.common import load_data23def bmd_plot(x, y, mean, samples, latlon_lo, latlon_hi):24    x_max, x_min = np.quantile(x, 0.95), np.quantile(x, 0.05)25    y_max, y_min = np.quantile(y, 0.95), np.quantile(y, 0.05)26    vmin, vmax = np.minimum(x_min, y_min), np.maximum(x_max, y_max)27    fig, axs, plot_fn = image_map_factory(2, 3, figsize=(6,4), cmap='viridis', min_max=(vmin, vmax))28    t = 0 # use first sample; the ordering should be random each epoch29    plot_fn(axs[0,0], x[t].numpy(), latlon_lo[0], latlon_lo[1], title='lo-res true')30    plot_fn(axs[0,1], y[t].numpy(), latlon_hi[0], latlon_hi[1], title='hi-res true')31    plot_fn(axs[0,2], mean[t].numpy(), latlon_hi[0], latlon_hi[1], title='hi-res predicted mean')32    for i, sample in enumerate(samples[t:t+3]):33        plot_fn(axs[1,i], samples[i].numpy(), latlon_hi[0], latlon_hi[1], title=f'sample {i}')34    return fig35def fit_bmd_maxt(fold, i, epochs, lr, batch_size, buffer_size, validate_freq):36    mlflow.log_param('fold', i+1)37    indices = tdex.indices('Time', convert_units_fn=lambda x: x + 273.15)38    data_fold = maxt.preprocess_fold_maxt(fold)39    train_lo, train_hi = data_fold.train40    test_lo, test_hi = data_fold.test41    N_train, N_test = train_lo.Time.size, test_lo.Time.size42    (ht_lr, wt_lr), (ht_hi, wt_hi) = train_lo.shape[1:3], train_hi.shape[1:3]43    monthly_means_lo, monthly_means_hi = data_fold.monthly_means44    train_ds = data_fold.train_dataset(batch_size=batch_size, buffer_size=buffer_size,45                                       mode='supervised')46    test_ds = data_fold.test_dataset(batch_size=batch_size, buffer_size=N_test,47                                     mode='test')48    scale = wt_hi // wt_lr49    encoder = create_bmd_cnn10(ht_lr, wt_lr, scale=scale, c_out=2)50    model = VariationalModel(encoder, normal(), optimizer=tf.keras.optimizers.Adam(lr=lr), output_shape=(None,ht_hi,wt_hi,1))51    ckpt_dir = f'/tmp/bmd-final'52    os.makedirs(ckpt_dir)53    for j in range(0, epochs, validate_freq):54        hist = model.fit(train_ds, epochs=validate_freq, steps_per_epoch=N_train//batch_size,55                         validation_data=test_ds, validation_steps=N_test//batch_size)56        hist = get_metrics(hist)57        mlflow.log_metrics(hist)58        j += validate_freq59        mlflow.log_metric('epoch', j)60        encoder.save(f'{ckpt_dir}/bmd-epoch{j}.h5')61        mlflow.log_artifact(f'{ckpt_dir}/bmd-epoch{j}.h5', artifact_path=f'model/')62        x_true = []63        y_true = []64        y_mean = []65        y_samples = []66        for x, y in test_ds:67            x_true.append(x)68            y_true.append(y)69            y_mean.append(model.mean(x))70            y_samples.append(model.sample(x))71        x_true = tf.concat(x_true, axis=0)72        y_true = tf.concat(y_true, axis=0)73        y_mean = tf.concat(y_mean, axis=0)74        y_samples = tf.concat(y_samples, axis=0)75        fig = bmd_plot(x_true, y_true, y_mean, y_samples, (test_lo.lat, test_lo.lon), (test_hi.lat, test_hi.lon))76        plt.savefig(f'/tmp/samples-epoch{j}.png')77        mlflow.log_artifact(f'/tmp/samples-epoch{j}.png', 'figures')78        metrics = maxt.eval_metrics(indices, y_true, y_mean, test_hi.coords, monthly_means_hi)79        np.savez(f'/tmp/metrics-epoch{j}.npz', **metrics)80        mlflow.log_artifact(f'/tmp/metrics-epoch{j}.npz', 'data')81        avg_metrics = {k: float(np.mean(v)) for k,v in metrics.items()}82        mlflow.log_metrics(avg_metrics)83        # create plots84        fig = maxt.plot_indices(metrics)85        plt.savefig(f'/tmp/indices-epoch{j}.png')86        mlflow.log_artifact(f'/tmp/indices-epoch{j}.png', 'figures')87        fig = maxt.plot_error_maps(metrics, test_hi.lat, test_hi.lon)88        plt.savefig(f'/tmp/error-maps-epoch{j}.png')89        mlflow.log_artifact(f'/tmp/error-maps-epoch{j}.png', 'figures')90    shutil.rmtree(ckpt_dir)91        92def fit_bmd_prcp(fold, i, epochs, lr, batch_size, buffer_size, validate_freq):93    mlflow.log_param('fold', i+1)94    indices = pdex.indices('Time')95    data_fold = prcp.preprocess_fold_prcp(fold)96    train_lo, train_hi = data_fold.train97    test_lo, test_hi = data_fold.test98    N_train, N_test = train_lo.Time.size, test_lo.Time.size99    (ht_lr, wt_lr), (ht_hi, wt_hi) = train_lo.shape[1:3], train_hi.shape[1:3]100    train_ds = data_fold.train_dataset(batch_size=batch_size, buffer_size=buffer_size,101                                       mode='supervised')102    test_ds = data_fold.test_dataset(batch_size=batch_size, buffer_size=N_test,103                                     mode='test')104    scale = wt_hi // wt_lr105    encoder = create_bmd_cnn10(ht_lr, wt_lr, scale=scale, c_out=3)106    model = VariationalModel(encoder, bernoulli_gamma(), optimizer=tf.keras.optimizers.Adam(lr=lr), output_shape=(None,ht_hi,wt_hi,1))107    ckpt_dir = f'/tmp/bmd-prcp-final'108    os.makedirs(ckpt_dir)109    for j in range(0, epochs, validate_freq):110        hist = model.fit(train_ds, epochs=validate_freq, steps_per_epoch=N_train//batch_size,111                         validation_data=test_ds, validation_steps=N_test//batch_size)112        hist = get_metrics(hist)113        mlflow.log_metrics(hist)114        j += validate_freq115        mlflow.log_metric('epoch', j)116        encoder.save(f'{ckpt_dir}/bmd-epoch{j}.h5')117        mlflow.log_artifact(f'{ckpt_dir}/bmd-epoch{j}.h5', artifact_path=f'model/')118        x_true = []119        y_true = []120        y_mean = []121        y_samples = []122        for x, y in test_ds:123            x_true.append(x)124            y_true.append(y)125            y_mean.append(model.mean(x))126            y_samples.append(model.sample(x))127        x_true = tf.concat(x_true, axis=0)128        y_true = tf.concat(y_true, axis=0)129        y_mean = tf.concat(y_mean, axis=0)130        y_samples = tf.concat(y_samples, axis=0)131        fig = bmd_plot(x_true, y_true, y_mean, y_samples, (test_lo.lat, test_lo.lon), (test_hi.lat, test_hi.lon))132        plt.savefig(f'/tmp/samples-epoch{j}.png')133        plt.close(fig)134        mlflow.log_artifact(f'/tmp/samples-epoch{j}.png', 'figures')135        y_true = tf.math.pow(y_true, 3.0)136        y_mean = tf.math.pow(y_mean, 3.0)137        metrics = prcp.eval_metrics(indices, y_true, y_mean, test_hi.coords)138        np.savez(f'/tmp/metrics-epoch{j}.npz', **metrics)139        mlflow.log_artifact(f'/tmp/metrics-epoch{j}.npz', 'data')140        avg_metrics = {k: float(np.mean(v)) for k,v in metrics.items()}141        mlflow.log_metrics(avg_metrics)142        # create plots143        fig = prcp.plot_indices(metrics)144        plt.savefig(f'/tmp/indices-epoch{j}.png')145        plt.close(fig)146        mlflow.log_artifact(f'/tmp/indices-epoch{j}.png', 'figures')147        fig = prcp.plot_error_maps(metrics, test_hi.lat, test_hi.lon)148        plt.savefig(f'/tmp/error-maps-epoch{j}.png')149        plt.close(fig)150        mlflow.log_artifact(f'/tmp/error-maps-epoch{j}.png', 'figures')151    shutil.rmtree(ckpt_dir)152@click.command(help="Fits and evaluates the Bano-Medina CNN10 on the ERA-I/Rasmussen dataset")153@click.option("--scale", type=click.INT, required=True, help="Downscaling factor")154@click.option("--epochs", type=click.INT, default=50)155@click.option("--learning-rate", type=click.FLOAT, default=1.0E-4)156@click.option("--batch-size", type=click.INT, default=100)157@click.option("--buffer-size", type=click.INT, default=2400)158@click.option("--validate-freq", type=click.INT, default=10)159@click.option("--region", type=click.STRING, default='southeast_us')160@click.option("--var", type=click.STRING, default='MAXT', help="Dataset var name")161@click.option("--test-size", type=click.INT, default=146, help='size of the test set for each fold')162@click.option("--splits", type=click.INT, default=5, help="Number of CV splits to use")163@click.option("--auth", type=click.STRING, default='gcs.secret.json', help="GCS keyfile")164@click.argument("data_lr", type=click.STRING, default="ras/daily-1deg")165def bmd(data_lr, scale, epochs, learning_rate, batch_size, buffer_size, validate_freq, region,166        var, test_size, splits, auth, **kwargs):167    mlflow.log_param('region', region)168    mlflow.log_param('var', var)169    if scale == 2:170        data_hr = 'daily-1-2deg'171    elif scale == 4:172        data_hr = 'daily-1-4deg'173    elif scale == 8:174        data_hr = 'daily-1-8deg'175    else:176        raise NotImplementedError(f'unsupported downscaling factor {scale}')177    logging.info(f'==== Starting run ====')178    data_lo, data_hi = load_data(data_lr, data_hr, region, auth, scale=scale)179    data_lo = data_lo[[var]].fillna(0.).clip(min=0.0, max=np.inf)180    data_hi = data_hi[[var]].fillna(0.).clip(min=0.0, max=np.inf)181    if var == 'PRCP':182        data_lo = prcp.preprocess_dataset(data_lo, (data_lo.Time.size, data_lo.lat.size, data_lo.lon.size), epsilon=1.0)183        data_lo = xr.where(data_lo > 1.0, data_lo, 0.0)184        data_hi = prcp.preprocess_dataset(data_hi, (data_hi.Time.size, data_hi.lat.size, data_hi.lon.size), epsilon=1.0)185        data_hi = xr.where(data_hi > 1.0, data_hi, 0.0)186    split_fn = create_time_series_train_test_generator_v2(n_splits=splits, test_size=test_size)187    folds = list(split_fn(data_lo, data_hi))188    for i, fold in enumerate(folds):189        if i < 4:190            continue191        logging.info(f'Fold {i+1}/{len(folds)}')192        with mlflow.start_run(nested=True):193            if var == 'MAXT':194                fit_bmd_maxt(fold, i, epochs, learning_rate,195                             batch_size, buffer_size, validate_freq)196            elif var == 'PRCP':197                fit_bmd_prcp(fold, i, epochs, learning_rate,198                             batch_size, buffer_size, validate_freq)199            else:...

cv_split.py

Source:cv_split.py

1#!/usr/bin/env python2# coding=utf-83"""4Cross-validation split of BAGEL data5------------------------------------6Will produce train-test splits for all cross-validation folds, each7in its own subdirectory.8Usage: ./cv_split.py [-f 10] [-c 2] [-d cv] all-xxx.txt all-yyy.txt9-f = number of folds (default to 10)10-c = number of chunks (alternative realizations, defaults to 2)11-d = directory prefix (defaults to "cv")12"""13from __future__ import unicode_literals14from pytreex.core.util import file_stream15from getopt import getopt16import os17from tgen.logf import log_warn18import random19import sys20import re21def write_data(dir, fname_base, fname_repl, data):22    chunk_size = len(data[0])23    for chunk_idx in xrange(chunk_size):24        fname_suff = ".%d" % chunk_idx if chunk_size > 1 else ''25        file_name = os.path.join(dir, re.sub(r'^[^-._]*', fname_repl + fname_suff, fname_base))26        print 'WRITING ' + file_name27        with file_stream(file_name, 'w') as fh:28            for chunk in data:29                print >> fh, chunk[chunk_idx],30def main(argv):31    opts, files = getopt(argv, 'f:c:d:')32    folds = 1033    chunk_size = 234    dir_prefix = 'cv'35    for opt, arg in opts:36        if opt == '-f':37            folds = int(arg)38        elif opt == '-c':39            chunk_size = int(arg)40        elif opt == '-d':41            dir_prefix = arg42    if not files:43        sys.exit(__doc__)44    45    random.seed(1206)46    ordering = None47    for file in files:48        # read all data49        data = []50        with file_stream(file) as fh:51            chunk = []52            for line in fh:53                chunk.append(line)54                if len(chunk) == chunk_size:55                    data.append(chunk)56                    chunk = []57            if chunk:58                log_warn('Incomplete chunk at end of file %s, size %d' % (file, len(chunk)))59        if ordering is None:60            # create ordering61            ordering = range(len(data))62            random.shuffle(ordering)63            # create directories64            for fold_no in xrange(folds):65                os.mkdir(dir_prefix + "%02d" % fold_no)66            67        # output as train and test into all CV portions68        fold_size, bigger_folds = divmod(len(data), folds)69        for fold_no in xrange(folds):70            # compute test data bounds71            if fold_no < bigger_folds:72                test_lo = (fold_size + 1) * fold_no73                test_hi = (fold_size + 1) * (fold_no + 1)74            else:75                test_lo = fold_size * fold_no + bigger_folds76                test_hi = fold_size * (fold_no + 1) + bigger_folds77            # select train and test data instances78            train_data = [data[idx] for ord, idx in enumerate(ordering)79                          if ord < test_lo or ord >= test_hi]80            test_data = [data[idx] for ord, idx in enumerate(ordering)81                         if ord >= test_lo and ord < test_hi]82            # write them out to a file (replace `all' in name with train/test)83            fname_base = os.path.basename(file)84            write_data(dir_prefix + "%02d" % fold_no, fname_base, 'train', train_data)85            write_data(dir_prefix + "%02d" % fold_no, fname_base, 'test', test_data)86            87if __name__ == '__main__':...

main.py

Source:main.py

1from demo import test_foo,test_hi2from plugin import pytest_depend3@pytest_depend([test_foo,test_hi])4def test_bar():5    pass6@pytest_depend([test_foo,test_hi])7def test_bar2():...

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.