How to use get_predict_point method in Airtest

Best Python code snippet using Airtest

Evaluation.py

Source:Evaluation.py

...9    df = pd.DataFrame()10    df['y_xgb_test'] = y_test11    df['y_xgb_pred'] = y_pred12    kstable = ks(df, 'y_xgb_test', 'y_xgb_pred')13    get_pos_neg_cnt(y_test, y_pred, get_predict_point(kstable))14def ks(data=None, target=None, prob=None):15    data['target0'] = 1 - data[target]16    data['bucket'] = pd.qcut(data[prob], 10)17    grouped = data.groupby('bucket', as_index=False)18    kstable = pd.DataFrame()19    kstable['min_prob'] = grouped.min()[prob]20    kstable['max_prob'] = grouped.max()[prob]21    kstable['events'] = grouped.sum()[target]22    kstable['nonevents'] = grouped.sum()['target0']23    kstable = kstable.sort_values(by="min_prob", ascending=False).reset_index(drop=True)24    kstable['event_rate'] = (kstable.events / data[target].sum()).apply('{0:.2%}'.format)25    kstable['nonevent_rate'] = (kstable.nonevents / data['target0'].sum()).apply('{0:.2%}'.format)26    kstable['cum_eventrate'] = (kstable.events / data[target].sum()).cumsum()27    kstable['cum_noneventrate'] = (kstable.nonevents / data['target0'].sum()).cumsum()28    kstable['KS'] = np.round(kstable['cum_eventrate'] - kstable['cum_noneventrate'], 3) * 10029    # Formating30    kstable['cum_eventrate'] = kstable['cum_eventrate'].apply('{0:.2%}'.format)31    kstable['cum_noneventrate'] = kstable['cum_noneventrate'].apply('{0:.2%}'.format)32    kstable.index = range(1, 11)33    kstable.index.rename('Decile', inplace=True)34    pd.set_option('display.max_columns', 9)35    print(kstable)36    # Display KS37    from colorama import Fore38    print(Fore.RED + "KS is " + str(max(kstable['KS'])) + "%" + " at decile " + str(39        (kstable.index[kstable['KS'] == max(kstable['KS'])][0])))40    return kstable41# è¿åæ£è´é¢æµ array42def get_pos_neg_predict(y_test, y_pred):43    print('in %s' % sys._getframe().f_code.co_name)44    y = pd.DataFrame(np.array([list(y_test), list(y_pred)]).T, columns=['true', 'pred'])45    pos_pred = y[y['true'] == 1]['pred']46    neg_pred = y[y['true'] == 0]['pred']47    print(pos_pred.shape)48    print(neg_pred.shape)49    return pos_pred, neg_pred50def draw_pos_neg_picture(pos_pred, neg_pred):51    print('in %s' % sys._getframe().f_code.co_name)52    # ä½¿æ£è´æ ·æ¬éä¸è´ ç¨nanå¡«å53    pos_cnt = pos_pred.shape[0]54    neg_cnt = neg_pred.shape[0]55    if neg_cnt > pos_cnt:56        pos_pred_fill = np.array([np.nan] * (neg_cnt - pos_cnt))57        pos_pred_fill = np.hstack((pos_pred, pos_pred_fill))58        neg_pred_fill = neg_pred59    else:60        neg_pred_fill = np.array([np.nan] * (pos_cnt - neg_cnt))61        neg_pred_fill = np.hstack((neg_pred, neg_pred_fill))62        pos_pred_fill = pos_pred63    dist = pd.DataFrame(np.array([pos_pred_fill, neg_pred_fill]).T, columns=['good', 'bad'])64    #     dist= pd.DataFrame(np.array([[0.1,0.2,0.2,np.nan],[0.7,0.8,0.8,0.9]]).T,columns=['a','b'])65    fig, ax = plt.subplots(figsize=(15, 6))66    dist.plot.kde(ax=ax, legend=False, title='Histogram: good vs. bad')67    dist.plot.hist(density=False, ax=ax, color=['red', 'blue'], histtype='barstacked')  # density=Frueç¨äºå è½®å»68    ax.set_ylabel('Frequency')69    ax.grid(axis='sample')70    ax.set_facecolor('#d8dcd6')71    plt.xlim(0, 1)72    plt.show()73# ä½ç»æåå¸å¾ è°ç¨save_test_resultãget_test_resultãget_pos_neg_predictãdraw_pos_neg_pictureçå½æ°74def get_pos_neg_picture(y_test, y_pred):75    import ParseData76    print('in %s' % sys._getframe().f_code.co_name)77    ParseData.save_test_result(y_test, y_pred)78    y_test, y_pred = ParseData.read_test_result()79    pos_pred, neg_pred = get_pos_neg_predict(y_pred, y_test)80    draw_pos_neg_picture(pos_pred, neg_pred)81# è·å¾é¢æµååç¹ï¼ksæå¤§çç¹çæ¦çå¼ï¼82def get_predict_point(kstable):83    max_decile = kstable.index[kstable['KS'] == max(kstable['KS'])][0]84    predict_point = kstable['min_prob'][max_decile]85    return predict_point86# è·å¾æ£è´æ ·æ¬çå®ãé¢æµä¸ªæ° ä¼ å¥ä¸ºarrayæSeries87def get_pos_neg_cnt(y_true, y_pred, predict_point):88    df = pd.DataFrame()89    df['true'] = y_true90    df['pred'] = y_pred91    true_pos = df[df['true'] == 1].shape[0]92    true_neg = df[df['true'] == 0].shape[0]93    pred_pos = df[df['pred'] > predict_point].shape[0]  # ksæ¯å·¦å¼å³éåºé´94    pred_neg = df[df['pred'] <= predict_point].shape[0]95    assert true_pos + true_neg == pred_pos + pred_neg, 'çå¼åé¢æµå¼çæ°éä¸ä¸è´ï¼'96    print('å¨å®éçæ ·æ¬ä¸ï¼%dä¸ºæ£æ ·æ¬ï¼%dä¸ºè´æ ·æ¬ï¼æ£è´æ¯ä¾ä¸º%f' % (true_pos, true_neg, (true_pos / true_neg)))...

callbacks.py

Source:callbacks.py

...22        self.row=row23        self.count = count24        self.point = point25    26    def get_predict_point(self, transform_mat):27        """28        transform_mat: (batch, 6)29        """30        if self.point == 4:31            zeros=tf.zeros_like(transform_mat)[:,:1]32            # affine_transforms=(batch, 6)33            transform_mat = tf.concat([transform_mat[:,0:1], zeros, transform_mat[:,1:2], zeros, transform_mat[:,2:4]],1)34        35        #print("get_predict_point : transform_mat  ",end = "")36        #print(type(transform_mat[1]), transform_mat.shape,transform_mat) 37        #tf.print(tf.shape(transform_mat))38        transform_mat = np.array(transform_mat)39        #print("\n\nafter\n\n")40        #print(type(transform_mat[1]), transform_mat.shape,transform_mat)41        transform_mat = transform_mat.reshape((-1, 2, 3))42        my_coord = np.array([[43            [-1,-1,1],44            [ 1, 1,1],45            [ 1,-1,1],46            [-1, 1,1]47        ]])48        my_coord = my_coord.transpose((0,2,1))49        new_coord = np.matmul(transform_mat, my_coord)50        new_coord = new_coord.transpose((0,2,1))51        return new_coord52    def coord_to_int(self, coords, imgshape):53        b, ih, iw = imgshape[:3]54        n_points=coords.shape[-1] // 255        ncoords = (coords + 1.0) / 2.0 * np.array([[iw, ih]*n_points])56        ncoords = ncoords.astype(np.int32).reshape((b, -1))57        return ncoords58    def on_epoch_end(self, epoch, logs={}):59        for i, (images,labels) in enumerate(self.dataset, 1):60            if self.require_coords:61                label, coords=labels62            stn_result, transform_mat = self.stn_model(images, training=False)63            # process origin image64            images = images.numpy()[:self.row]65            images = (images*255.).astype(np.uint8)66            transform_mat = transform_mat.numpy()[:self.row]67            #print("on_epoch_end : transform_mat  ",end = "")68            #print(type(transform_mat[1]), transform_mat.shape,transform_mat) 69            #tf.print(tf.shape(transform_mat))70            pcoords = self.coord_to_int(self.get_predict_point(transform_mat), images.shape)71            if self.require_coords: 72                gcoords = self.coord_to_int(coords.numpy()[:self.row], images.shape)73            n_points = pcoords.shape[-1] // 274            for ii in range(len(images)):75                img = images[ii].copy()76                for iii in range(n_points):77                    if self.require_coords: 78                        images[ii] = cv2.circle(img, tuple(gcoords[ii,2*iii:2*(iii+1)]), 3, (int(127+128/4*iii), 0, 0), -1)79                    images[ii] = cv2.circle(img, tuple(pcoords[ii,2*iii:2*(iii+1)]), 3, (0, 0, int(127+128/4*iii)), -1)80                    81            images = np.vstack(images)82            # process stn_result83            stn_result = stn_result.numpy()[:self.row]84            stn_result = (stn_result*255.).astype(np.uint8)...

server.py

Source:server.py

...24@app.route("/predict", methods=["GET","POST"])25def predict():26    if not request.json or not 'data' in request.json:27        abort(400)28    result = get_predict_point(embeddings,29                            request.json['p0x'], 30                            request.json['p0y'],31                            request.json['p0r'],32                            request.json['p1x'],33                            request.json['p1y'],34                            request.json['p1r'])35    return result, 20036def get_predict_point(embeddings, p0x, p0y, p0r, p1x, p1y, p1r):37    #max_p0x, max_p0y, max_p1x, max_p1y = get_corners(p0x, p0y, p0r, p1x, p1y, p1r)38    39    #area_embeddings = get_embaddings_of_tiles_in_area(max_p0x, max_p0y, max_p1x, max_p1y)40    41    id0 = get_tile_id_by_geolocation(p0x, p0y)42    id1 = get_tile_id_by_geolocation(p1x, p1y)43    grapf = build_grapf(area_embeddings)44    predict_point = predict_on_tree(id0, id1, grapf)45def predict_on_tree(id0, id1, grapf):46    interpolation_size = 547    # interpolate from 185175 to 75168848    # These numbers are cherry picked. Once we train for longer, we can remove it49    begin_index = id0 #8421250    end_index = id1...

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.