def test_tuning_serializer(self) -> None:
"""Test Tuning config serializer."""
data = {
"strategy": {
"name": "mse",
"accuracy_weight": 0.5,
"latency_weight": 1.0,
},
"accuracy_criterion": {
"relative": 0.01,
"absolute": 0.02,
},
"objective": "performance",
"exit_policy": {
"timeout": 60,
"max_trials": 200,
},
"random_seed": 12345,
"tensorboard": True,
"workspace": {
"path": "/path/to/workspace",
"resume": "/path/to/snapshot/file",
},
"additional_field": {
"key": "val"
},
}
tuning = Tuning(data)
result = tuning.serialize()
self.assertDictEqual(
result,
{
"strategy": {
"name": "mse",
"accuracy_weight": 0.5,
"latency_weight": 1.0,
},
"accuracy_criterion": {
"relative": 0.01,
"absolute": 0.02,
},
"objective": "performance",
"exit_policy": {
"timeout": 60,
"max_trials": 200,
},
"random_seed": 12345,
"tensorboard": True,
"workspace": {
"path": "/path/to/workspace",
"resume": "/path/to/snapshot/file",
},
},
)
def test_set_timeout_with_exit_policy(self) -> None:
"""Test overwriting timeout in Tuning config."""
tuning = Tuning({
"exit_policy": {
"timeout": 60,
},
}, )
self.assertIsNotNone(tuning.exit_policy)
self.assertEqual(tuning.exit_policy.timeout, 60)
tuning.set_timeout(10)
self.assertIsNotNone(tuning.exit_policy)
self.assertEqual(tuning.exit_policy.timeout, 10)
def test_set_max_trials_with_exit_policy(self) -> None:
"""Test overwriting max_trials in Tuning config."""
tuning = Tuning({
"exit_policy": {
"max_trials": 60,
},
}, )
self.assertIsNotNone(tuning.exit_policy)
self.assertEqual(tuning.exit_policy.max_trials, 60)
tuning.set_max_trials(10)
self.assertIsNotNone(tuning.exit_policy)
self.assertEqual(tuning.exit_policy.max_trials, 10)
def initialize(self, data: Dict[str, Any] = {}) -> None:
"""Initialize config from dict."""
self.model_path = data.get("model_path", self.model_path)
self.domain = data.get("domain", self.domain)
if isinstance(data.get("model"), dict):
self.model = Model(data.get("model", {}))
# [Optional] One of "cpu", "gpu"; default cpu
self.device = data.get("device", None)
if isinstance(data.get("quantization"), dict):
self.quantization = Quantization(data.get("quantization", {}))
if isinstance(data.get("tuning"), dict):
self.tuning = Tuning(data.get("tuning", {}))
if isinstance(data.get("evaluation"), dict):
self.evaluation = Evaluation(data.get("evaluation", {}))
if isinstance(data.get("pruning"), dict):
self.pruning = Pruning(data.get("pruning", {}))
if isinstance(data.get("graph_optimization"), dict):
self.graph_optimization = GraphOptimization(
data.get("graph_optimization", {}))
def __init__(self, data: Dict[str, Any] = {}):
"""Initialize Configuration class."""
super().__init__()
self._skip.append("model_path")
self.model_path: str = data.get("model_path", "")
self.model: Model = Model()
self.domain: Optional[str] = data.get("domain", None)
self.device: Optional[str] = None
self.quantization: Optional[Quantization] = None
self.tuning: Tuning = Tuning()
self.evaluation: Optional[Evaluation] = None
self.pruning: Optional[Pruning] = None
self.initialize(data)
def test_tuning_constructor(self) -> None:
"""Test Tuning config constructor."""
data = {
"strategy": {
"name": "mse",
"accuracy_weight": 0.5,
"latency_weight": 1.0,
},
"accuracy_criterion": {
"relative": 0.01,
"absolute": 0.02,
},
"objective": "performance",
"exit_policy": {
"timeout": 60,
"max_trials": 200,
},
"random_seed": 12345,
"tensorboard": True,
"workspace": {
"path": "/path/to/workspace",
"resume": "/path/to/snapshot/file",
},
}
tuning = Tuning(data)
self.assertIsNotNone(tuning.strategy)
self.assertEqual(tuning.strategy.name, "mse")
self.assertEqual(tuning.strategy.accuracy_weight, 0.5)
self.assertEqual(tuning.strategy.latency_weight, 1.0)
self.assertIsNotNone(tuning.accuracy_criterion)
self.assertEqual(tuning.accuracy_criterion.relative, 0.01)
self.assertEqual(tuning.accuracy_criterion.absolute, 0.02)
self.assertEqual(tuning.objective, "performance")
self.assertIsNotNone(tuning.exit_policy)
self.assertEqual(tuning.exit_policy.timeout, 60)
self.assertEqual(tuning.exit_policy.max_trials, 200)
self.assertEqual(tuning.random_seed, 12345)
self.assertTrue(tuning.tensorboard)
self.assertIsNotNone(tuning.workspace)
self.assertEqual(tuning.workspace.path, "/path/to/workspace")
self.assertEqual(tuning.workspace.resume, "/path/to/snapshot/file")
def test_tuning_constructor_defaults(self) -> None:
"""Test Tuning config constructor defaults."""
tuning = Tuning()
self.assertIsNotNone(tuning.strategy)
self.assertEqual(tuning.strategy.name, "basic")
self.assertIsNone(tuning.strategy.accuracy_weight)
self.assertIsNone(tuning.strategy.latency_weight)
self.assertIsNotNone(tuning.accuracy_criterion)
self.assertIsNone(tuning.accuracy_criterion.relative)
self.assertIsNone(tuning.accuracy_criterion.absolute)
self.assertIsNone(tuning.objective)
self.assertIsNone(tuning.exit_policy)
self.assertIsNone(tuning.random_seed)
self.assertIsNone(tuning.tensorboard)
self.assertIsNotNone(tuning.workspace)
self.assertIsNone(tuning.workspace.path)
self.assertIsNone(tuning.workspace.resume)
def test_set_random_seed_invalid_string(self) -> None:
"""Test setting random_seed from invalid string in Tuning config."""
tuning = Tuning()
with self.assertRaises(ClientErrorException):
tuning.set_random_seed("abc")
def test_set_random_seed_from_string(self) -> None:
"""Test setting random_seed from string in Tuning config."""
tuning = Tuning()
tuning.set_random_seed("123456")
self.assertEqual(tuning.random_seed, 123456)
def test_set_random_seed(self) -> None:
"""Test setting random_seed in Tuning config."""
tuning = Tuning()
tuning.set_random_seed(123456)
self.assertEqual(tuning.random_seed, 123456)
def test_set_max_trials_invalid_string(self) -> None:
"""Test overwriting max_trials in Tuning config."""
tuning = Tuning()
with self.assertRaises(ClientErrorException):
tuning.set_max_trials("abc")
def test_set_max_trials_negative(self) -> None:
"""Test overwriting max_trials in Tuning config."""
tuning = Tuning()
with self.assertRaises(ClientErrorException):
tuning.set_max_trials(-1)
def test_set_max_trials_from_string(self) -> None:
"""Test overwriting max_trials in Tuning config."""
tuning = Tuning()
tuning.set_max_trials("10")
self.assertIsNotNone(tuning.exit_policy)
self.assertEqual(tuning.exit_policy.max_trials, 10)
def test_set_max_trials(self) -> None:
"""Test setting max_trials in Tuning config."""
tuning = Tuning()
tuning.set_max_trials(10)
self.assertIsNotNone(tuning.exit_policy)
self.assertEqual(tuning.exit_policy.max_trials, 10)
def test_set_timeout_negative(self) -> None:
"""Test overwriting timeout in Tuning config."""
tuning = Tuning()
with self.assertRaises(ClientErrorException):
tuning.set_timeout(-1)
def test_set_timeout_from_string(self) -> None:
"""Test overwriting timeout in Tuning config."""
tuning = Tuning()
tuning.set_timeout("10")
self.assertIsNotNone(tuning.exit_policy)
self.assertEqual(tuning.exit_policy.timeout, 10)
def test_set_timeout(self) -> None:
"""Test setting timeout in Tuning config."""
tuning = Tuning()
tuning.set_timeout(10)
self.assertIsNotNone(tuning.exit_policy)
self.assertEqual(tuning.exit_policy.timeout, 10)