def run(self) -> Tuple[Optional[DeepLearningConfig], Optional[Run]]:
"""
The main entry point for training and testing models from the commandline. This chooses a model to train
via a commandline argument, runs training or testing, and writes all required info to disk and logs.
:return: If submitting to AzureML, returns the model configuration that was used for training,
including commandline overrides applied (if any).
"""
# Usually, when we set logging to DEBUG, we want diagnostics about the model
# build itself, but not the tons of debug information that AzureML submissions create.
logging_to_stdout(logging.INFO if is_local_rank_zero() else "ERROR")
initialize_rpdb()
user_agent.append(azure_util.INNEREYE_SDK_NAME, azure_util.INNEREYE_SDK_VERSION)
self.parse_and_load_model()
if self.lightning_container.perform_cross_validation:
if self.model_config is None:
raise NotImplementedError("Cross validation for LightingContainer models is not yet supported.")
# force hyperdrive usage if performing cross validation
self.azure_config.hyperdrive = True
run_object: Optional[Run] = None
if self.azure_config.azureml:
run_object = self.submit_to_azureml()
else:
self.run_in_situ()
if self.model_config is None:
return self.lightning_container, run_object
return self.model_config, run_object
def __append_telemetry(self):
if not self.telemetry_set:
self.telemetry_set = True
try:
from azureml._base_sdk_common.user_agent import append
append("AzureMLCluster-DASK", "0.1")
except ImportError:
pass
import threading
import dask
import pathlib
from distributed.utils import (
LoopRunner,
log_errors,
format_bytes,
)
from tornado.ioloop import PeriodicCallback
logger = logging.getLogger(__name__)
try:
from azureml._base_sdk_common.user_agent import append
append('AzureMLCluster-DASK', '0.1')
except:
pass
class AzureMLCluster(Cluster):
""" Deploy a Dask cluster using Azure ML
This creates a dask scheduler and workers on an Azure ML Compute Target.
Parameters
----------
workspace: azureml.core.Workspace (required)
Azure ML Workspace - see https://aka.ms/azureml/workspace
compute_target: azureml.core.ComputeTarget (required)
Azure ML Compute Target - see https://aka.ms/azureml/computetarget
environment_definition: azureml.core.Environment (required)
Azure ML Environment - see https://aka.ms/azureml/environments
"Model",
"PrivateEndPoint",
"PrivateEndPointConfig",
"Run",
"RunConfiguration",
"ScriptRun",
"ScriptRunConfig",
"Webservice",
"Workspace",
"attach_legacy_compute_target",
"get_run",
"is_compute_target_prepared",
"prepare_compute_target",
"remove_legacy_compute_target",
"LinkedService",
"SynapseWorkspaceLinkedServiceConfiguration"
]
user_agent.append("azureml-sdk-core", __version__)
# Appending the Arcadia environment variable to user agent string to indicate request origin.
if _ArcadiaAuthentication._is_arcadia_environment():
user_agent.append(_ArcadiaAuthentication._ARCADIA_ENVIRONMENT_VARIABLE_VALUE)
RUN_TYPE_PROVIDERS_ENTRYPOINT_KEY = "azureml_run_type_providers"
for entrypoint in pkg_resources.iter_entry_points(RUN_TYPE_PROVIDERS_ENTRYPOINT_KEY):
try:
Run.add_type_provider(entrypoint.name, entrypoint.load())
except Exception as e:
module_logger.warning("Failure while loading {}. Failed to load entrypoint {} with exception {}.".format(
RUN_TYPE_PROVIDERS_ENTRYPOINT_KEY, entrypoint, e))
