Test your AI Agents with the all-new Agent to Agent Testing Platform.Learn More

How to use test_output_diff method in avocado

Best Python code snippet using avocado_python

test_output_check.py

Source:test_output_check.py

...124        self.assertEqual(result.exit_status, expected_rc,125                         "Avocado did not return rc %d:\n%s" %126                         (expected_rc, result))127        self.assertIn(tampered_msg, result.stdout)128    def test_output_diff(self):129        self._check_output_record_all()130        tampered_msg_stdout = b"I PITY THE FOOL THAT STANDS ON STDOUT!"131        tampered_msg_stderr = b"I PITY THE FOOL THAT STANDS ON STDERR!"132        stdout_file = "%s.data/stdout.expected" % self.output_script.path133        with open(stdout_file, 'wb') as stdout_file_obj:134            stdout_file_obj.write(tampered_msg_stdout)135        stderr_file = "%s.data/stderr.expected" % self.output_script.path136        with open(stderr_file, 'wb') as stderr_file_obj:137            stderr_file_obj.write(tampered_msg_stderr)138        cmd_line = ('%s run --job-results-dir %s --sysinfo=off %s --json -'139                    % (AVOCADO, self.tmpdir, self.output_script.path))140        result = process.run(cmd_line, ignore_status=True)141        expected_rc = exit_codes.AVOCADO_TESTS_FAIL142        self.assertEqual(result.exit_status, expected_rc,...

train.py

Source:train.py

1import torch2import torch.nn as nn3import os4import time5import pickle6import argparse7import numpy as np8from layer import QRNNLayer9from model import QRNNModel10import data.dataloader as data_loader11from torch.utils.data import DataLoader12from torch.utils.data.sampler import SubsetRandomSampler13"""14Train the model15"""16device = torch.device('cuda:3')17def create_model(config):  # not to use config.num_symbols, but actual dict size18    print('Creating new model parameters..')19    model = QRNNModel(config.dim_in, config.dim_hid, config.dim_out,20                      QRNNLayer, config.num_layers, config.kernel_size,21                      config.hidden_size, config.batch_size,22                      config.frames, config.dec_size, config.out_size, device)23    # Initialize a model state24    model_state = vars(config)25    model_state['epoch'], model_state['train_steps'] = 0, 026    model_state['state_dict'] = None27    model_path = os.path.join(config.model_dir, config.model_name)28    # If training stops half way, restart training from last checkpoint of previous training.29    if os.path.exists(model_path):30        print('Reloading model parameters..')31        checkpoint = torch.load(model_path)32        model_state['epoch'] = checkpoint['epoch']33        model_state['train_steps'] = checkpoint['train_steps']34        model.load_state_dict(checkpoint['state_dict'])35    print('Using gpu..')36    model.train().to(device)37    print(next(model.parameters()).is_cuda)38    return model, model_state39def diff_trans(output_diff, index):40    output = torch.zeros_like(output_diff).to(device)41    for i in range(len(output_diff)):42        ind = index[i][-1]43        with open('/home/tfukuda/qrnn/data/labels/'+str(int(ind[0])).zfill(4)+'_ped'+str(int(ind[2]))+'.pkl', 'rb') as f:44            bbox = pickle.load(f)45            46        p_bbox = torch.tensor(bbox[int(ind[1])]).to(device)47        for j in range(len(output_diff[i])):48            output[i][j] = p_bbox + output_diff[i][j]49            p_bbox = output[i][j]50    return output51def train(config):52    print('Loading data..')53    # Creating data for training and test splits54    dataset = data_loader.Seq_Data(config.frames, config.dim_in, config.dim_out)55    # Get random datas by using SubsetRandomSampler56    batch_size = config.batch_size57    test_split = 0.258    shuffle_dataset = True59    random_seed= 4260    dataset_size = len(dataset)61    indices = list(range(dataset_size))62    split = int(np.floor(test_split * dataset_size))63    if shuffle_dataset :64        np.random.seed(random_seed)65        np.random.shuffle(indices)66    train_indices, test_indices = indices[split:], indices[:split]67    train_sampler = SubsetRandomSampler(train_indices)68    test_sampler = SubsetRandomSampler(test_indices)69    train_datas = DataLoader(dataset, batch_size=batch_size, sampler=train_sampler, drop_last=True, num_workers=2)70    test_datas = DataLoader(dataset, batch_size=batch_size, sampler=test_sampler, drop_last=True, num_workers=2)71    print('train data length: ', len(train_datas))72    print('test data length: ', len(test_datas))73    model, model_state = create_model(config)74    criterion = nn.L1Loss()75    optimizer = torch.optim.Adam(model.parameters(), lr=config.lr)76    loss = 0.077    start_time = time.time()78    # Training loop79    print('Training..')80    for epoch in range(config.epochs):81        model_state['epoch'] += 182        for data, label, index in train_datas:83            data = data.requires_grad_().to(device)84            label = label.to(device)85            index = index.to(device)86            # Execute a single training step87            optimizer.zero_grad()88            output_diff = model(data)89            output = diff_trans(output_diff, index)90            step_loss = criterion(output, label)91            step_loss.backward()92            nn.utils.clip_grad_norm(model.parameters(), config.max_grad_norm)93            optimizer.step()94            loss += float(step_loss) / config.display_freq95            model_state['train_steps'] += 196            # Display the average loss of training97            if model_state['train_steps'] % config.display_freq == 0:98                avg_loss = float(loss)99                time_elapsed = time.time() - start_time100                step_time = time_elapsed / config.display_freq101                print('Epoch ', model_state['epoch'], 'Step ', model_state['train_steps'], \102                      'Loss {0:.2f}'.format(avg_loss), 'Step-time {0:.2f}'.format(step_time))103                loss = 0.0104                start_time = time.time()105            # Test step start106            if model_state['train_steps'] % config.test_freq == 0:107                model.eval()108                print('Test step')109                test_steps = 0110                test_loss = 0.0111                for test_data, test_label, test_index in test_datas:112                    test_data = test_data.requires_grad_().to(device)113                    test_label = test_label.to(device)114                    test_index = test_index.to(device)115                    test_output_diff = model(test_data)116                    test_output = diff_trans(test_output_diff, test_index)117                    step_loss = criterion(test_output, test_label)118                    test_steps += 1 119                    test_loss += float(step_loss)120                model.train()121                #Display loss of test steps122                print('Test Loss: {0:.2f}'.format(test_loss / test_steps))123            # Save the model checkpoint124            if model_state['train_steps'] % config.save_freq == 0:125                print('Saving the model..')126                model_state['state_dict'] = model.state_dict()127                model_path = os.path.join(config.model_dir, config.model_name)128                torch.save(model_state, model_path)129        # Increase the epoch index of the model130        print('Epoch {0:} DONE'.format(model_state['epoch']))131if __name__ == "__main__":132    parser = argparse.ArgumentParser()133    # Network parameters134    parser.add_argument('--dim_in', type=float, default=17)135    parser.add_argument('--dim_hid', type=float, default=8)136    parser.add_argument('--dim_out', type=float, default=4)137    parser.add_argument('--kernel_size', type=int, default=2)138    parser.add_argument('--hidden_size', type=int, default=1024)139    parser.add_argument('--num_layers', type=int, default=2)140    parser.add_argument('--dec_size', type=int, default=2)141    parser.add_argument('--out_size', type=int, default=4)142    # Training parameters143    parser.add_argument('--lr', type=float, default=0.001)144    parser.add_argument('--max_grad_norm', type=float, default=1.0)145    parser.add_argument('--batch_size', type=int, default=16)146    parser.add_argument('--frames', type=int, default=30)147    parser.add_argument('--epochs', type=int, default=16)148    parser.add_argument('--display_freq', type=int, default=100)149    parser.add_argument('--save_freq', type=int, default=200)150    parser.add_argument('--test_freq', type=int, default=200)151    parser.add_argument('--model_dir', type=str, default='model/')152    parser.add_argument('--model_name', type=str, default='model_diff_hid-8_dec-2_ep-16.pkl')153    config = parser.parse_args()154    print(config)155    train(config)...

test_tflite_debugger.py

Source:test_tflite_debugger.py

...92        )93        self.assertEqual(len(output_diffs), len(debug_tflite_model.output_names))94        self.assertEqual(len(output_diffs[0].metrics["mae"]), 2)95class OutputDiffTest(tf.test.TestCase):96    def test_output_diff(self):97        diff = OutputDiff(98            left_name="a",99            right_name="b",100            shape=(4, 5),101            left_dtype=np.float32,102            right_dtype=np.float32,103            metrics={"mse": [0, 1, 2, 1]},104        )105        self.assertEqual(diff.as_flat_dict()["metric/mse"], 1.0)106        df = OutputDiff.to_df([diff, diff])107        df = pd.DataFrame(df)108        df = df.set_index("left_name")...

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.