How to use test_window_size method in Airtest

Best Python code snippet using Airtest

main.py

Source:main.py

1# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.2#3# Licensed under the Apache License, Version 2.0 (the "License");4# you may not use this file except in compliance with the License.5# You may obtain a copy of the License at6#7#     http://www.apache.org/licenses/LICENSE-2.08#9# Unless required by applicable law or agreed to in writing, software10# distributed under the License is distributed on an "AS IS" BASIS,11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.12# See the License for the specific language governing permissions and13# limitations under the License.14import paddle15import argparse16from paddlevideo.utils import get_config17from paddlevideo.tasks import train_model, train_model_multigrid, test_model, train_dali18from paddlevideo.utils import get_dist_info19import numpy as np20import random21from paddlevideo.modeling.backbones.effigcn_init import EffigcnInit22def predefined_model_list():23    agcn_model_list = [24        {'cfg_path': 'configs/recognition/agcn_fold/agcn_fsd0.yaml',25         'weight_path': './output/AGCN_fold0/AGCN_best.pdparams',26         'test_window_size': [300, 350, 500]},27        {'cfg_path': 'configs/recognition/agcn_fold/agcn_fsd1.yaml',28         'weight_path': './output/AGCN_fold1/AGCN_best.pdparams',29         'test_window_size': [300, 350, 500]},30        {'cfg_path': 'configs/recognition/agcn_fold/agcn_fsd2.yaml',31         'weight_path': './output/AGCN_fold2/AGCN_best.pdparams',32         'test_window_size': [300, 350, 500]},33        {'cfg_path': 'configs/recognition/agcn_fold/agcn_fsd3.yaml',34         'weight_path': './output/AGCN_fold3/AGCN_best.pdparams',35         'test_window_size': [300, 350, 500]},36        {'cfg_path': 'configs/recognition/agcn_fold/agcn_fsd4.yaml',37         'weight_path': './output/AGCN_fold4/AGCN_best.pdparams',38         'test_window_size': [300, 350, 500]}39    ]40    ctrgcn_model_list = [41        {'cfg_path': 'configs/recognition/ctrgcn_fold/ctrgcn_fsd0.yaml',42         'weight_path': './output/CTRGCN_fold0/CTRGCN_best.pdparams',43         'test_window_size': [300, 350, 450]},44        {'cfg_path': 'configs/recognition/ctrgcn_fold/ctrgcn_fsd1.yaml',45         'weight_path': './output/CTRGCN_fold1/CTRGCN_best.pdparams',46         'test_window_size': [300, 350, 450]},47        {'cfg_path': 'configs/recognition/ctrgcn_fold/ctrgcn_fsd2.yaml',48         'weight_path': './output/CTRGCN_fold2/CTRGCN_best.pdparams',49         'test_window_size': [300, 350, 450]},50        {'cfg_path': 'configs/recognition/ctrgcn_fold/ctrgcn_fsd3.yaml',51         'weight_path': './output/CTRGCN_fold3/CTRGCN_best.pdparams',52         'test_window_size': [300, 350, 450]},53        {'cfg_path': 'configs/recognition/ctrgcn_fold/ctrgcn_fsd4.yaml',54         'weight_path': './output/CTRGCN_fold4/CTRGCN_best.pdparams',55         'test_window_size': [300, 350, 450]}56    ]57    ctrgcn2_model_list = [58        {'cfg_path': 'configs/recognition/ctrgcn2_fold/ctrgcn_fsd0.yaml',59         'weight_path': './output/CTRGCN2_fold0/CTRGCN_best.pdparams',60         'test_window_size': [350]},61        {'cfg_path': 'configs/recognition/ctrgcn2_fold/ctrgcn_fsd1.yaml',62         'weight_path': './output/CTRGCN2_fold1/CTRGCN_best.pdparams',63         'test_window_size': [350]},64        {'cfg_path': 'configs/recognition/ctrgcn2_fold/ctrgcn_fsd2.yaml',65         'weight_path': './output/CTRGCN2_fold2/CTRGCN_best.pdparams',66         'test_window_size': [350]},67        {'cfg_path': 'configs/recognition/ctrgcn2_fold/ctrgcn_fsd3.yaml',68         'weight_path': './output/CTRGCN2_fold3/CTRGCN_best.pdparams',69         'test_window_size': [350]},70        {'cfg_path': 'configs/recognition/ctrgcn2_fold/ctrgcn_fsd4.yaml',71         'weight_path': './output/CTRGCN2_fold4/CTRGCN_best.pdparams',72         'test_window_size': [350]}73    ]74    stgcn_model_list = [75        {'cfg_path': 'configs/recognition/stgcn_fold/stgcn_fsd0.yaml',76         'weight_path': './output/STGCN_fold0/STGCN_best.pdparams',77         'test_window_size': [350]},78        {'cfg_path': 'configs/recognition/stgcn_fold/stgcn_fsd1.yaml',79         'weight_path': './output/STGCN_fold1/STGCN_best.pdparams',80         'test_window_size': [350]},81        {'cfg_path': 'configs/recognition/stgcn_fold/stgcn_fsd2.yaml',82         'weight_path': './output/STGCN_fold2/STGCN_best.pdparams',83         'test_window_size': [350]},84        {'cfg_path': 'configs/recognition/stgcn_fold/stgcn_fsd3.yaml',85         'weight_path': './output/STGCN_fold3/STGCN_best.pdparams',86         'test_window_size': [350]},87        {'cfg_path': 'configs/recognition/stgcn_fold/stgcn_fsd4.yaml',88         'weight_path': './output/STGCN_fold4/STGCN_best.pdparams',89         'test_window_size': [350]},90    ]91    effigcn_model_list = [92        {'cfg_path': 'configs/recognition/effigcn_fold/effigcn_fsd0.yaml',93         'weight_path': './output/EFFGCN_fold0/EFFGCN_best.pdparams',94         'test_window_size': [500]},95        {'cfg_path': 'configs/recognition/effigcn_fold/effigcn_fsd1.yaml',96         'weight_path': './output/EFFGCN_fold1/EFFGCN_best.pdparams',97         'test_window_size': [500]},98        {'cfg_path': 'configs/recognition/effigcn_fold/effigcn_fsd2.yaml',99         'weight_path': './output/EFFGCN_fold2/EFFGCN_best.pdparams',100         'test_window_size': [500]},101        {'cfg_path': 'configs/recognition/effigcn_fold/effigcn_fsd3.yaml',102         'weight_path': './output/EFFGCN_fold3/EFFGCN_best.pdparams',103         'test_window_size': [500]},104        {'cfg_path': 'configs/recognition/effigcn_fold/effigcn_fsd4.yaml',105         'weight_path': './output/EFFGCN_fold4/EFFGCN_best.pdparams',106         'test_window_size': [500]},107    ]108    models = agcn_model_list + ctrgcn_model_list + ctrgcn2_model_list + stgcn_model_list + effigcn_model_list109    return models110def parse_args():111    parser = argparse.ArgumentParser("PaddleVideo train script")112    # parser.add_argument('-c',113    #                     '--config',114    #                     type=str,115    #                     default='configs/example.yaml',116    #                     help='config file path')117    parser.add_argument('-o',118                        '--override',119                        action='append',120                        default=[],121                        help='config options to be overridden')122    parser.add_argument('--test',123                        action='store_true',124                        help='whether to test a model')125    parser.add_argument('--train_dali',126                        action='store_true',127                        help='whether to use dali to speed up training')128    parser.add_argument('--multigrid',129                        action='store_true',130                        help='whether to use multigrid training')131    parser.add_argument('-w',132                        '--weights',133                        type=str,134                        help='weights for finetuning or testing')135    parser.add_argument('--fleet',136                        action='store_true',137                        help='whether to use fleet run distributed training')138    parser.add_argument('--amp',139                        action='store_true',140                        help='whether to open amp training.')141    parser.add_argument(142        '--validate',143        action='store_true',144        help='whether to evaluate the checkpoint during training')145    parser.add_argument('-g', '--gpu', default="0", type=str)146    args = parser.parse_args()147    return args148def main():149    args = parse_args()150    # cfg = get_config(args.config, overrides=args.override)151    _, world_size = get_dist_info()152    parallel = world_size != 1153    if parallel:154        paddle.distributed.init_parallel_env()155    model_list = predefined_model_list()156    print("************************************************************")157    for model in model_list:158        print(model['cfg_path'])159    print("************************************************************")160    for model in model_list:161        np.random.seed(42)162        random.seed(2021)163        paddle.seed(102)164        cfg = get_config(model['cfg_path'], overrides=args.override)165        cfg['gpu'] = args.gpu166        print("*************" + model['cfg_path'] + "*******************")167        if (cfg.MODEL.backbone.name == 'EfficientGCN'):168            EffigcnInit(cfg)  # åå§åEfficientGCNåæ°169        if args.test:170            test_model(cfg, weights=model['weight_path'], parallel=parallel, test_window_size=model['test_window_size'])171        else:172            train_model(cfg, weights=args.weights, parallel=parallel, validate=args.validate, use_fleet=args.fleet,173                        amp=args.amp)174    # if args.test:175    #     test_model(cfg, weights=args.weights, parallel=parallel)176    # elif args.train_dali:177    #178    #     train_dali(cfg, weights=args.weights, parallel=parallel)179    # elif args.multigrid:180    #     train_model_multigrid(cfg, world_size, validate=args.validate)181    # else:182    #     train_model(cfg,183    #                 weights=args.weights,184    #                 parallel=parallel,185    #                 validate=args.validate,186    #                 use_fleet=args.fleet,187    #                 amp=args.amp)188if __name__ == '__main__':...

nightly_torch.py

Source:nightly_torch.py

1import sys2import time3#sys.path.append('../gpt-2/src/')4#import callable_gpt2 as gpt25#import model6#import encoder7#import sample8import torch9import numpy as np10import pandas as pd11from transformers import *12#tf.keras.backend.set_floatx('float64')13'''14error in tensorflow dtype init if tensorflow version = 2.0.0beta115pip uninstall tensorflow 16pip install tensorflow==2.0.017unhashable18tf.compat.v1.enable_tensor_equality()19'''20train_window_size = 40021test_window_size = 10022def pad_corpus_ori(sentences, pad_token, window_size):23    pad_sentences = []24    i = 025    for text in sentences:26        text = text.split()27        extended_text = text[:window_size-1] + [pad_token] * (window_size - len(text)-1)28        29        pad_sentences.append(extended_text)30        i += 131    return pad_sentences32def pad_corpus(sentences, start_token, end_token, pad_token, window_size):33    pad_sentences = []34    i = 035    for text in sentences:36        text = text.split()37        extended_text = text[:window_size-1] + [pad_token] * (window_size - len(text)-1) + [end_token]38        39        pad_sentences.append(extended_text)40        i += 141    return pad_sentences42def pad_corpus_simul(train_sentences, test_sentences, special_tokens, train_window_size, test_window_size):43    pad_sentences = []44    i = 045    for train_text, test_text in zip(train_sentences, test_sentences):46        train_text, test_text = train_text.split(), test_text.split()47        extended_text = [special_tokens['bos_token']] + train_text[:train_window_size-1] + [special_tokens['pad_token']] * (train_window_size - len(train_text)-1) + [special_tokens['cls_token']] + \48                        test_text[:test_window_size-1] + [special_tokens['pad_token']] * (test_window_size - len(test_text)-1) + [special_tokens['eos_token']]49                                                                                          50        pad_sentences.append(extended_text)51        i += 152    #print("length of post-padded text", len(pad_sentences[0]))53    return pad_sentences54def train(model, tokenizer, file_path, special_tokens):55    optimizer = AdamW(model.parameters())56    57    batch_size = 1058    reader = pd.read_json(file_path, precise_float=True, dtype=False, lines=True, chunksize=batch_size)59    train_steps = 060    step_start_time = time.time()61    training_time = [step_start_time]62    63    for section in reader:64        #print("section['normalizedBody']: \n", section['normalizedBody'].tolist(), len(section['normalizedBody'].tolist()))65        train_steps += 166        67        # get paragraph and summary68        train_words = section['normalizedBody'].tolist()69        test_words = section['summary'].tolist()70        #words = np.array(words)71        #print("words as numpy\n", words, "\n", words.shape)72        #print("concatted words", words)73        words = pad_corpus_simul(train_words, test_words, special_tokens, train_window_size, test_window_size)74        #test_words = pad_corpus(test_words, special_tokens['bos_token'], special_tokens['eos_token'], special_tokens['pad_token'], test_window_size)75        #train_words = pad_corpus_ori(train_words, special_tokens['pad_token'], train_window_size)76        #test_words = pad_corpus_ori(test_words, special_tokens['pad_token'], test_window_size)77        '''78        train_words = pad_corpus_ori(train_words, '', train_window_size)79        words = []80        for train, test in zip(train_words, test_words):81            words.append(train + test)82        '''83        words = [ tokenizer.encode(text) for text in words ]84        words = torch.tensor(words)85        test_words = pad_corpus_ori(test_words, special_tokens['pad_token'], test_window_size)86        test_words = [ tokenizer.encode(text) for text in test_words ]87        test_words = torch.tensor(test_words)88        #words = tf.reshape(words, [batch_size, -1])89        print("words as tensor\n", words, "\n", words.shape)90        91        outputs  = model(words, labels=words) # last hidden states92        loss, logits = outputs[0], outputs[1]93        #logit = logit.numpy()94        predicted_index = torch.argmax(logits[0, -1, :])95        #loss = torch.keras.losses.sparse_categorical_crossentropy(words, logits, from_logits=True)96        #print("loss\n", loss, "\n", loss.shape)97        print("probs first\n", logits.shape)98        print("encoded prediction", predicted_index)99        print("decoded sentence first\n", tokenizer.decode(predicted_index))100        exit(0)101if __name__ == '__main__':102    # convert to pretrained to use transformers103    104    tokenizer = GPT2Tokenizer.from_pretrained('gpt2')105    model = GPT2LMHeadModel.from_pretrained('gpt2')106    #print("loaded model", model)107    #add the special tokens into the vocabulary108    special_tokens = {'bos_token': '<bos>', 'cls_token': '<cls>', 'eos_token': '<eos>',109                      'pad_token': '<pad>'}110    # tf transformers module not currently implemented resize_token_embeddings function :(111    tokenizer.add_special_tokens(special_tokens)112    model.resize_token_embeddings(len(tokenizer))113    114    #num_train_optimization_steps = len(self.train_dataset) * options['num_epochs'] // options['train_batchsize']115    #train(model, '../data/tldr_train80.jsonl', special_tokens)116    train(model, tokenizer, '../data/tldr-training-data.jsonl', special_tokens)...

out_score.py

Source:out_score.py

1import numpy as np2import csv3def get_model(model_name, power, test_window_size, power_window_size):4    """5    get score of model6    Args:7        model_name: str, we use model_name to find the folder8        power:  ndarray, power of each fold to fuse9        test_window_size: list, the window_size we selected to fuse10        power_window_size: ndarray, power of each window_size to fuse11    Returns:12        score: score of model13    """14    out = np.zeros((634, 30), dtype='float32')15    power = power / power.sum()16    power_window_size = power_window_size / power_window_size.sum()17    for i in range(0, 5):18        for j in range(0, len(test_window_size)):19            path = "./output/%s_fold%s/score_%s.npy" % (model_name, i, test_window_size[j])20            print(path)21            data = np.load(path)22            out = out + data * power[i] * power_window_size[j]23    return out24def get_agcn():25    model_name = 'AGCN'26    power = np.array([1.5, 1, 1.5, 1, 2], dtype='float32')27    test_window_size = [300, 350, 500]28    power_window_size = np.array([1, 1, 1], dtype='float32')29    return get_model(model_name, power, test_window_size, power_window_size)30def get_ctrgcn():31    model_name = 'CTRGCN'32    power = np.array([1, 1, 2, 2, 3], dtype='float32')33    test_window_size = [300, 350, 450]34    power_window_size = np.array([1, 1, 1], dtype='float32')35    return get_model(model_name, power, test_window_size, power_window_size)36def get_ctrgcn2():37    model_name = 'CTRGCN2'38    power = np.array([1, 1, 2, 2, 3], dtype='float32')39    test_window_size = [350]40    power_window_size = np.array([1], dtype='float32')41    return get_model(model_name, power, test_window_size, power_window_size)42def get_stgcn():43    model_name = 'STGCN'44    power = np.array([1, 1, 2, 1, 2], dtype='float32')45    test_window_size = [350]46    power_window_size = np.array([1], dtype='float32')47    return get_model(model_name, power, test_window_size, power_window_size)48def get_effgcn():49    model_name = 'EFFGCN'50    power = np.array([1, 1, 2, 1, 3], dtype='float32')51    test_window_size = [500]52    power_window_size = np.array([1], dtype='float32')53    return get_model(model_name, power, test_window_size, power_window_size)54def get_all():55    out = np.zeros((634, 30), dtype='float32')56    data = []57    data.append(get_agcn())58    data.append(get_ctrgcn())59    data.append(get_ctrgcn2())60    data.append(get_stgcn())61    data.append(get_effgcn())62    power = np.array([2, 0.75, 1.5, 3.75, 2], dtype='float32')63    power = power / power.sum()64    for i in range(0, len(data)):65        out = out + data[i] * power[i]66    return out67def get_values_for_csv(data):68    values = []69    for i in range(0, data.shape[0]):70        j = np.argmax(data[i], axis=0)71        values.append((i, j))72    return values73def out_file(data, out_file):74    with open(75            out_file,76            'w',77    ) as fp:78        writer = csv.writer(fp)79        if out_file == "submission.csv":80            writer.writerow(['sample_index', 'predict_category'])81        writer.writerows(data)82if __name__ == "__main__":...

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.