def log_param(self, run_id, param):
"""
Log a param for the specified run
:param run_id: String id for the run
:param param: :py:class:`mlflow.entities.Param` instance to log
"""
_validate_run_id(run_id)
_validate_param_name(param.key)
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
dynamodb = self._get_dynamodb_resource()
table_name = "_".join([self.table_prefix, DynamodbStore.PARAMS_TABLE])
table = dynamodb.Table(table_name)
response = table.put_item(
Item={
"run_id": run_id,
"key": param.key,
"value": param.value
},
ReturnConsumedCapacity="TOTAL",
ReturnValues="ALL_OLD",
)
if response["ResponseMetadata"]["HTTPStatusCode"] != 200:
raise MlflowException("DynamoDB connection error")
return True
def record_logged_model(self, run_id, mlflow_model):
from mlflow.models import Model
if not isinstance(mlflow_model, Model):
raise TypeError(
"Argument 'mlflow_model' should be mlflow.models.Model, got '{}'"
.format(type(mlflow_model)))
_validate_run_id(run_id)
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
model_dict = mlflow_model.to_dict()
run_info = self._get_run_info(run_id)
path = self._get_tag_path(run_info.experiment_id, run_info.run_id,
MLFLOW_LOGGED_MODELS)
if os.path.exists(path):
with open(path, "r") as f:
model_list = json.loads(f.read())
else:
model_list = []
tag = RunTag(MLFLOW_LOGGED_MODELS,
json.dumps(model_list + [model_dict]))
try:
self._set_run_tag(run_info, tag)
except Exception as e:
raise MlflowException(e, INTERNAL_ERROR)
def set_tag(self, run_id, tag):
"""
Set a tag for the specified run
:param run_id: String id for the run
:param tag: :py:class:`mlflow.entities.RunTag` instance to set
"""
_validate_run_id(run_id)
_validate_tag_name(tag.key)
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
dynamodb = self._get_dynamodb_resource()
table_name = "_".join([self.table_prefix, DynamodbStore.TAGS_TABLE])
table = dynamodb.Table(table_name)
response = table.put_item(
Item={
"run_id": run_id,
"key": tag.key,
"value": tag.value
},
ReturnConsumedCapacity="TOTAL",
)
if response["ResponseMetadata"]["HTTPStatusCode"] != 200:
raise MlflowException("DynamoDB connection error")
return True
def delete_tag(self, run_id, key):
"""
Delete a tag from a run. This is irreversible.
:param run_id: String ID of the run
:param key: Name of the tag
"""
_validate_run_id(run_id)
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
dynamodb = self._get_dynamodb_resource()
table_name = "_".join([self.table_prefix, DynamodbStore.TAGS_TABLE])
table = dynamodb.Table(table_name)
response = table.get_item(Key={"run_id": run_id, "key": key})
if not "Item" in response:
raise MlflowException(
"No tag with name: {} in run with id {}".format(key, run_id),
error_code=RESOURCE_DOES_NOT_EXIST,
)
response = table.delete_item(Key={
"run_id": run_id,
"key": key
},
ReturnConsumedCapacity="TOTAL")
if response["ResponseMetadata"]["HTTPStatusCode"] != 200:
raise MlflowException("DynamoDB connection error")
return True
def update_run_info(self, run_id, run_status, end_time):
_validate_run_id(run_id)
run_info = self.get_run(run_id).info
check_run_is_active(run_info)
new_info = run_info._copy_with_overrides(run_status, end_time)
self._overwrite_run_info(new_info)
return new_info
def log_metric(self, run_id, metric):
_validate_run_id(run_id)
_validate_metric_name(metric.key)
run = self.get_run(run_id)
check_run_is_active(run.info)
metric_path = self._get_metric_path(run.info.experiment_id, run_id, metric.key)
make_containing_dirs(metric_path)
append_to(metric_path, "%s %s %s\n" % (metric.timestamp, metric.value, metric.step))
def delete_run(self, run_id):
run_info = self._get_run_info(run_id)
if run_info is None:
raise MlflowException("Run '%s' metadata is in invalid state." % run_id,
databricks_pb2.INVALID_STATE)
check_run_is_active(run_info)
new_info = run_info._copy_with_overrides(lifecycle_stage=LifecycleStage.DELETED)
self._overwrite_run_info(new_info)
def log_param(self, run_id, param):
_validate_run_id(run_id)
_validate_param_name(param.key)
run = self.get_run(run_id)
check_run_is_active(run.info)
param_path = self._get_param_path(run.info.experiment_id, run_id, param.key)
make_containing_dirs(param_path)
write_to(param_path, self._writeable_value(param.value))
def set_tag(self, run_id, tag):
_validate_run_id(run_id)
_validate_tag_name(tag.key)
run = self.get_run(run_id)
check_run_is_active(run.info)
tag_path = self._get_tag_path(run.info.experiment_id, run_id, tag.key)
make_containing_dirs(tag_path)
# Don't add trailing newline
write_to(tag_path, self._writeable_value(tag.value))
def update_run_info(self, run_id, run_status, end_time):
"""
Update the metadata of the specified run.
:return: :py:class:`mlflow.entities.RunInfo` describing the updated run.
"""
_validate_run_id(run_id)
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
return self._update_run_info(run_id, run_status, end_time)
def delete_run(self, run_id):
"""
Delete a run.
:param run_id
"""
_validate_run_id(run_id)
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
return self._update_run_status(run_id, LifecycleStage.ACTIVE,
LifecycleStage.DELETED)
def log_batch(self, run_id, metrics, params, tags):
_validate_run_id(run_id)
run = self.get_run(run_id)
check_run_is_active(run.info)
try:
for param in params:
self.log_param(run_id, param)
for metric in metrics:
self.log_metric(run_id, metric)
for tag in tags:
self.set_tag(run_id, tag)
except Exception as e:
raise MlflowException(e, INTERNAL_ERROR)
def delete_tag(self, run_id, key):
"""
Delete a tag from a run. This is irreversible.
:param run_id: String ID of the run
:param key: Name of the tag
"""
_validate_run_id(run_id)
run = self.get_run(run_id)
check_run_is_active(run.info)
if key not in run.data.tags.keys():
raise MlflowException(
"No tag with name: {} in run with id {}".format(key, run_id),
error_code=RESOURCE_DOES_NOT_EXIST)
tag_path = self._get_tag_path(run.info.experiment_id, run_id, key)
os.remove(tag_path)
def log_batch(self, run_id, metrics, params, tags):
_validate_run_id(run_id)
_validate_batch_log_data(metrics, params, tags)
_validate_batch_log_limits(metrics, params, tags)
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
try:
for param in params:
self._log_run_param(run_info, param)
for metric in metrics:
self._log_run_metric(run_info, metric)
for tag in tags:
self._set_run_tag(run_info, tag)
except Exception as e:
raise MlflowException(e, INTERNAL_ERROR)
def log_metric(self, run_id, metric):
"""
Log a metric for the specified run
:param run_id: String id for the run
:param metric: :py:class:`mlflow.entities.Metric` instance to log
"""
_validate_run_id(run_id)
_validate_metric_name(metric.key)
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
dynamodb = self._get_dynamodb_resource()
table_name = "_".join([self.table_prefix, DynamodbStore.METRICS_TABLE])
table = dynamodb.Table(table_name)
# Append metrics to head of list, so the first element is most recent
response = table.update_item(
Key={
"run_id": run_id,
"key": metric.key
},
UpdateExpression="SET #m = list_append(:m, if_not_exists(#m, :e))",
ExpressionAttributeNames={"#m": "metrics"},
ExpressionAttributeValues={
":e": [],
":m": [{
"value": Decimal(str(metric.value)),
"timestamp": metric.timestamp,
"step": metric.step,
}],
},
ReturnValues="NONE",
ReturnConsumedCapacity="TOTAL",
)
if response["ResponseMetadata"]["HTTPStatusCode"] != 200:
raise MlflowException("DynamoDB connection error")
return True
def delete_run(self, run_id):
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
new_info = run_info._copy_with_overrides(
lifecycle_stage=RunInfo.DELETED_LIFECYCLE)
self._overwrite_run_info(new_info)
def log_metric(self, run_id, metric):
_validate_run_id(run_id)
_validate_metric_name(metric.key)
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
self._log_run_metric(run_info, metric)
def set_tag(self, run_id, tag):
_validate_run_id(run_id)
_validate_tag_name(tag.key)
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
self._set_run_tag(run_info, tag)
def log_param(self, run_id, param):
_validate_run_id(run_id)
_validate_param_name(param.key)
run_info = self._get_run_info(run_id)
check_run_is_active(run_info)
self._log_run_param(run_info, param)
