def _model_complete(self, model: Model):
time.sleep(random.uniform(0, 1))
if random.uniform(0, 1) < self.failure_prob:
model.status = ModelStatus.Failed
else:
model.metric = random.uniform(0, 1)
model.status = ModelStatus.Trained
self._resource_left += 1
def test_py_execution_engine(self):
advisor = RetiariiAdvisor()
set_execution_engine(PurePythonExecutionEngine())
model = Model._load({
'_model': {
'inputs': None,
'outputs': None,
'nodes': {
'layerchoice_1': {
'operation': {
'type': 'LayerChoice',
'parameters': {
'candidates': ['0', '1']
}
}
}
},
'edges': []
}
})
model.python_class = object
submit_models(model, model)
advisor.stopping = True
advisor.default_worker.join()
advisor.assessor_worker.join()
def test_py_execution_engine(self):
nni.retiarii.integration_api._advisor = None
nni.retiarii.execution.api._execution_engine = None
advisor = RetiariiAdvisor('ws://_unittest_placeholder_')
advisor._channel = LegacyCommandChannel()
advisor.default_worker.start()
advisor.assessor_worker.start()
set_execution_engine(PurePythonExecutionEngine())
model = Model._load({
'_model': {
'inputs': None,
'outputs': None,
'nodes': {
'layerchoice_1': {
'operation': {
'type': 'LayerChoice',
'parameters': {
'candidates': ['0', '1']
}
}
}
},
'edges': []
}
})
model.evaluator = DebugEvaluator()
model.python_class = object
submit_models(model, model)
advisor.stopping = True
advisor.default_worker.join()
advisor.assessor_worker.join()
def test_codegen(self):
with open(self.enclosing_dir / 'mnist_pytorch.json') as f:
model = Model._load(json.load(f))
script = model_to_pytorch_script(model)
with open(self.enclosing_dir / 'debug_mnist_pytorch.py') as f:
reference_script = f.read()
self.assertEqual(script.strip(), reference_script.strip())
def test_mnist_example_pytorch(self):
with open('mnist_pytorch.json') as f:
model = Model._load(json.load(f))
script = model_to_pytorch_script(model)
with open('debug_mnist_pytorch.py') as f:
reference_script = f.read()
self.assertEqual(script.strip(), reference_script.strip())
def test_base_execution_engine(self):
advisor = RetiariiAdvisor()
set_execution_engine(BaseExecutionEngine())
with open(self.enclosing_dir / 'mnist_pytorch.json') as f:
model = Model._load(json.load(f))
submit_models(model, model)
advisor.stopping = True
advisor.default_worker.join()
advisor.assessor_worker.join()
def _load_mnist(n_models: int = 1):
path = Path(__file__).parent / 'converted_mnist_pytorch.json'
with open(path) as f:
mnist_model = Model._load(json.load(f))
if n_models == 1:
return mnist_model
else:
models = [mnist_model]
for i in range(n_models - 1):
models.append(mnist_model.fork())
return models
def test_trainer(self):
sys.path.insert(0, Path(__file__).parent.as_posix())
Model = import_('debug_mnist_pytorch._model')
trainer = PyTorchImageClassificationTrainer(
Model(),
dataset_kwargs={
'root': (Path(__file__).parent / 'data' / 'mnist').as_posix(),
'download': True
},
dataloader_kwargs={'batch_size': 32},
optimizer_kwargs={'lr': 1e-3},
trainer_kwargs={'max_epochs': 1})
trainer.fit()
def test_submit_models(self):
os.makedirs('generated', exist_ok=True)
from nni.runtime import protocol
protocol._out_file = open(
Path(__file__).parent / 'generated/debug_protocol_out_file.py',
'wb')
advisor = RetiariiAdvisor()
with open('mnist_pytorch.json') as f:
model = Model._load(json.load(f))
submit_models(model, model)
advisor.stopping = True
advisor.default_worker.join()
advisor.assessor_worker.join()
def _load_mnist(n_models: int = 1):
path = Path(__file__).parent / 'mnist_pytorch.json'
with open(path) as f:
mnist_model = Model._load(json.load(f))
mnist_model.evaluator = _new_trainer()
if n_models == 1:
return mnist_model
else:
models = [mnist_model]
for i in range(n_models - 1):
forked_model = mnist_model.fork()
forked_model.evaluator = _new_trainer()
models.append(forked_model)
return models
def _load_mnist(n_models: int = 1):
path = Path('ut/nas/mnist_pytorch.json')
with open(path) as f:
mnist_model = Model._load(nni.load(fp=f))
mnist_model.evaluator = _new_trainer()
if n_models == 1:
return mnist_model
else:
models = [mnist_model]
for i in range(n_models - 1):
forked_model = mnist_model.fork()
forked_model.evaluator = _new_trainer()
models.append(forked_model)
return models
def test_base_execution_engine(self):
nni.retiarii.integration_api._advisor = None
nni.retiarii.execution.api._execution_engine = None
advisor = RetiariiAdvisor('ws://_unittest_placeholder_')
advisor._channel = LegacyCommandChannel()
advisor.default_worker.start()
advisor.assessor_worker.start()
set_execution_engine(BaseExecutionEngine())
with open(self.enclosing_dir / 'mnist_pytorch.json') as f:
model = Model._load(json.load(f))
submit_models(model, model)
advisor.stopping = True
advisor.default_worker.join()
advisor.assessor_worker.join()