def _test_experiment_in_separate_process(rootpath):
try:
base_model, evaluator = _mnist_net('simple', {'max_epochs': 1})
search_strategy = strategy.Random()
exp = RetiariiExperiment(base_model, evaluator, strategy=search_strategy)
exp_config = RetiariiExeConfig('local')
exp_config.experiment_name = 'mnist_unittest'
exp_config.trial_concurrency = 1
exp_config.max_trial_number = 1
exp_config._trial_command_params = nas_experiment_trial_params(rootpath)
exp.run(exp_config)
ensure_success(exp)
assert isinstance(exp.export_top_models()[0], dict)
finally:
# https://stackoverflow.com/questions/34506638/how-to-register-atexit-function-in-pythons-multiprocessing-subprocess
import atexit
atexit._run_exitfuncs()
def test_multi_trial(model, pytestconfig):
evaluator_kwargs = {
'max_epochs': 1
}
base_model, evaluator = _mnist_net(model, evaluator_kwargs)
search_strategy = strategy.Random()
exp = RetiariiExperiment(base_model, evaluator, strategy=search_strategy)
exp_config = RetiariiExeConfig('local')
exp_config.experiment_name = 'mnist_unittest'
exp_config.trial_concurrency = 1
exp_config.max_trial_number = 1
exp_config._trial_command_params = nas_experiment_trial_params(pytestconfig.rootpath)
exp.run(exp_config)
ensure_success(exp)
assert isinstance(exp.export_top_models()[0], dict)
exp.stop()
trainer = cgo.Classification(train_dataloader=pl.DataLoader(
train_dataset, batch_size=100),
val_dataloaders=pl.DataLoader(test_dataset,
batch_size=100),
max_epochs=1,
limit_train_batches=0.2)
applied_mutators = [BlockMutator('mutable_0'), BlockMutator('mutable_1')]
simple_strategy = TPEStrategy()
exp = RetiariiExperiment(base_model, trainer, applied_mutators,
simple_strategy)
exp_config = RetiariiExeConfig('remote')
exp_config.experiment_name = 'darts_search'
exp_config.trial_concurrency = 3
exp_config.max_trial_number = 10
exp_config.trial_gpu_number = 1
exp_config.training_service.reuse_mode = True
rm_conf = RemoteMachineConfig()
rm_conf.host = '127.0.0.1'
rm_conf.user = '******'
rm_conf.password = '******'
rm_conf.port = 22
rm_conf.python_path = '/home/xxx/py38/bin'
rm_conf.gpu_indices = [0, 1, 2]
rm_conf.use_active_gpu = True
rm_conf.max_trial_number_per_gpu = 3
best_val_ppl
) # reports best validation ppl to nni as final result of one trial
if __name__ == "__main__":
train_iter = WikiText2(split='train')
tokenizer = get_tokenizer('basic_english')
vocab = build_vocab_from_iterator(map(tokenizer, train_iter),
specials=['<unk>'])
vocab.set_default_index(vocab['<unk>'])
n_token = len(vocab)
base_model = Transformer(n_token)
evaluator = FunctionalEvaluator(fit)
exp = RetiariiExperiment(base_model, evaluator, [], strategy.Random())
exp_config = RetiariiExeConfig('local')
exp_config.experiment_name = 'transformer tuning'
exp_config.trial_concurrency = 3 # please change configurations accordingly
exp_config.max_trial_number = 25
exp_config.trial_gpu_number = 1
exp_config.training_service.use_active_gpu = False
export_formatter = 'dict'
exp.run(exp_config, 8081)
print('Final model:')
for model_code in exp.export_top_models(optimize_mode='minimize',
formatter=export_formatter):
print(model_code)
transform=transform)
trainer = pl.Classification(train_dataloader=pl.DataLoader(train_dataset,
batch_size=100),
val_dataloaders=pl.DataLoader(test_dataset,
batch_size=100),
max_epochs=2,
gpus=1,
limit_train_batches=0.1,
limit_val_batches=0.1)
simple_strategy = strategy.Random()
exp = RetiariiExperiment(base_model, trainer, [], simple_strategy)
exp_config = RetiariiExeConfig('local')
exp_config.experiment_name = 'mnist_search'
exp_config.trial_concurrency = 2
exp_config.max_trial_number = args.budget
exp_config.trial_gpu_number = 1
exp_config.training_service.use_active_gpu = True # Integration test GPU has a Xorg running
export_formatter = 'dict'
if args.exec == 'graph':
exp_config.execution_engine = 'base'
export_formatter = 'code'
exp.run(exp_config, args.port)
print('Final model:')
for model_code in exp.export_top_models(formatter=export_formatter):
print(model_code)
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
])
valid_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
])
train_dataset = serialize(CIFAR10, root='data/cifar10', train=True, download=True, transform=train_transform)
test_dataset = serialize(CIFAR10, root='data/cifar10', train=False, download=True, transform=valid_transform)
trainer = pl.Classification(train_dataloader=pl.DataLoader(train_dataset, batch_size=100),
val_dataloaders=pl.DataLoader(test_dataset, batch_size=100),
max_epochs=1, limit_train_batches=0.2)
applied_mutators = [
BlockMutator('mutable_0'),
BlockMutator('mutable_1')
]
simple_strategy = TPEStrategy()
exp = RetiariiExperiment(base_model, trainer, applied_mutators, simple_strategy)
exp_config = RetiariiExeConfig('local')
exp_config.experiment_name = 'mnasnet_search'
exp_config.trial_concurrency = 2
exp_config.max_trial_number = 10
exp_config.training_service.use_active_gpu = False
exp.run(exp_config, 8097)