518 lines
17 KiB
Python
518 lines
17 KiB
Python
# -*- coding: utf-8 -*-
|
|
#
|
|
# Copyright 2019 Google LLC
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# https://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
"""Define resources for the BigQuery ML Models API."""
|
|
|
|
from __future__ import annotations # type: ignore
|
|
|
|
import copy
|
|
import datetime
|
|
import typing
|
|
from typing import Any, Dict, Optional, Sequence, Union
|
|
|
|
import google.cloud._helpers # type: ignore
|
|
from google.cloud.bigquery import _helpers
|
|
from google.cloud.bigquery import standard_sql
|
|
from google.cloud.bigquery.encryption_configuration import EncryptionConfiguration
|
|
|
|
|
|
class Model:
|
|
"""Model represents a machine learning model resource.
|
|
|
|
See
|
|
https://cloud.google.com/bigquery/docs/reference/rest/v2/models
|
|
|
|
Args:
|
|
model_ref:
|
|
A pointer to a model. If ``model_ref`` is a string, it must
|
|
included a project ID, dataset ID, and model ID, each separated
|
|
by ``.``.
|
|
"""
|
|
|
|
_PROPERTY_TO_API_FIELD = {
|
|
"expires": "expirationTime",
|
|
"friendly_name": "friendlyName",
|
|
# Even though it's not necessary for field mapping to map when the
|
|
# property name equals the resource name, we add these here so that we
|
|
# have an exhaustive list of all mutable properties.
|
|
"labels": "labels",
|
|
"description": "description",
|
|
"encryption_configuration": "encryptionConfiguration",
|
|
}
|
|
|
|
def __init__(self, model_ref: Union["ModelReference", str, None]):
|
|
# Use _properties on read-write properties to match the REST API
|
|
# semantics. The BigQuery API makes a distinction between an unset
|
|
# value, a null value, and a default value (0 or ""), but the protocol
|
|
# buffer classes do not.
|
|
self._properties: Dict[str, Any] = {}
|
|
|
|
if isinstance(model_ref, str):
|
|
model_ref = ModelReference.from_string(model_ref)
|
|
|
|
if model_ref:
|
|
self._properties["modelReference"] = model_ref.to_api_repr()
|
|
|
|
@property
|
|
def reference(self) -> Optional["ModelReference"]:
|
|
"""A model reference pointing to this model.
|
|
|
|
Read-only.
|
|
"""
|
|
resource = self._properties.get("modelReference")
|
|
if resource is None:
|
|
return None
|
|
else:
|
|
return ModelReference.from_api_repr(resource)
|
|
|
|
@property
|
|
def project(self) -> Optional[str]:
|
|
"""Project bound to the model."""
|
|
ref = self.reference
|
|
return ref.project if ref is not None else None
|
|
|
|
@property
|
|
def dataset_id(self) -> Optional[str]:
|
|
"""ID of dataset containing the model."""
|
|
ref = self.reference
|
|
return ref.dataset_id if ref is not None else None
|
|
|
|
@property
|
|
def model_id(self) -> Optional[str]:
|
|
"""The model ID."""
|
|
ref = self.reference
|
|
return ref.model_id if ref is not None else None
|
|
|
|
@property
|
|
def path(self) -> Optional[str]:
|
|
"""URL path for the model's APIs."""
|
|
ref = self.reference
|
|
return ref.path if ref is not None else None
|
|
|
|
@property
|
|
def location(self) -> Optional[str]:
|
|
"""The geographic location where the model resides.
|
|
|
|
This value is inherited from the dataset.
|
|
|
|
Read-only.
|
|
"""
|
|
return typing.cast(Optional[str], self._properties.get("location"))
|
|
|
|
@property
|
|
def etag(self) -> Optional[str]:
|
|
"""ETag for the model resource (:data:`None` until set from the server).
|
|
|
|
Read-only.
|
|
"""
|
|
return typing.cast(Optional[str], self._properties.get("etag"))
|
|
|
|
@property
|
|
def created(self) -> Optional[datetime.datetime]:
|
|
"""Datetime at which the model was created (:data:`None` until set from the server).
|
|
|
|
Read-only.
|
|
"""
|
|
value = typing.cast(Optional[float], self._properties.get("creationTime"))
|
|
if value is None:
|
|
return None
|
|
else:
|
|
# value will be in milliseconds.
|
|
return google.cloud._helpers._datetime_from_microseconds(
|
|
1000.0 * float(value)
|
|
)
|
|
|
|
@property
|
|
def modified(self) -> Optional[datetime.datetime]:
|
|
"""Datetime at which the model was last modified (:data:`None` until set from the server).
|
|
|
|
Read-only.
|
|
"""
|
|
value = typing.cast(Optional[float], self._properties.get("lastModifiedTime"))
|
|
if value is None:
|
|
return None
|
|
else:
|
|
# value will be in milliseconds.
|
|
return google.cloud._helpers._datetime_from_microseconds(
|
|
1000.0 * float(value)
|
|
)
|
|
|
|
@property
|
|
def model_type(self) -> str:
|
|
"""Type of the model resource.
|
|
|
|
Read-only.
|
|
"""
|
|
return typing.cast(
|
|
str, self._properties.get("modelType", "MODEL_TYPE_UNSPECIFIED")
|
|
)
|
|
|
|
@property
|
|
def training_runs(self) -> Sequence[Dict[str, Any]]:
|
|
"""Information for all training runs in increasing order of start time.
|
|
|
|
Dictionaries are in REST API format. See:
|
|
https://cloud.google.com/bigquery/docs/reference/rest/v2/models#trainingrun
|
|
|
|
Read-only.
|
|
"""
|
|
return typing.cast(
|
|
Sequence[Dict[str, Any]], self._properties.get("trainingRuns", [])
|
|
)
|
|
|
|
@property
|
|
def feature_columns(self) -> Sequence[standard_sql.StandardSqlField]:
|
|
"""Input feature columns that were used to train this model.
|
|
|
|
Read-only.
|
|
"""
|
|
resource: Sequence[Dict[str, Any]] = typing.cast(
|
|
Sequence[Dict[str, Any]], self._properties.get("featureColumns", [])
|
|
)
|
|
return [
|
|
standard_sql.StandardSqlField.from_api_repr(column) for column in resource
|
|
]
|
|
|
|
@property
|
|
def transform_columns(self) -> Sequence[TransformColumn]:
|
|
"""The input feature columns that were used to train this model.
|
|
The output transform columns used to train this model.
|
|
|
|
See REST API:
|
|
https://cloud.google.com/bigquery/docs/reference/rest/v2/models#transformcolumn
|
|
|
|
Read-only.
|
|
"""
|
|
resources: Sequence[Dict[str, Any]] = typing.cast(
|
|
Sequence[Dict[str, Any]], self._properties.get("transformColumns", [])
|
|
)
|
|
return [TransformColumn(resource) for resource in resources]
|
|
|
|
@property
|
|
def label_columns(self) -> Sequence[standard_sql.StandardSqlField]:
|
|
"""Label columns that were used to train this model.
|
|
|
|
The output of the model will have a ``predicted_`` prefix to these columns.
|
|
|
|
Read-only.
|
|
"""
|
|
resource: Sequence[Dict[str, Any]] = typing.cast(
|
|
Sequence[Dict[str, Any]], self._properties.get("labelColumns", [])
|
|
)
|
|
return [
|
|
standard_sql.StandardSqlField.from_api_repr(column) for column in resource
|
|
]
|
|
|
|
@property
|
|
def best_trial_id(self) -> Optional[int]:
|
|
"""The best trial_id across all training runs.
|
|
|
|
.. deprecated::
|
|
This property is deprecated!
|
|
|
|
Read-only.
|
|
"""
|
|
value = typing.cast(Optional[int], self._properties.get("bestTrialId"))
|
|
if value is not None:
|
|
value = int(value)
|
|
return value
|
|
|
|
@property
|
|
def expires(self) -> Optional[datetime.datetime]:
|
|
"""The datetime when this model expires.
|
|
|
|
If not present, the model will persist indefinitely. Expired models will be
|
|
deleted and their storage reclaimed.
|
|
"""
|
|
value = typing.cast(Optional[float], self._properties.get("expirationTime"))
|
|
if value is None:
|
|
return None
|
|
else:
|
|
# value will be in milliseconds.
|
|
return google.cloud._helpers._datetime_from_microseconds(
|
|
1000.0 * float(value)
|
|
)
|
|
|
|
@expires.setter
|
|
def expires(self, value: Optional[datetime.datetime]):
|
|
if value is None:
|
|
value_to_store: Optional[str] = None
|
|
else:
|
|
value_to_store = str(google.cloud._helpers._millis_from_datetime(value))
|
|
# TODO: Consider using typing.TypedDict when only Python 3.8+ is supported.
|
|
self._properties["expirationTime"] = value_to_store # type: ignore
|
|
|
|
@property
|
|
def description(self) -> Optional[str]:
|
|
"""Description of the model (defaults to :data:`None`)."""
|
|
return typing.cast(Optional[str], self._properties.get("description"))
|
|
|
|
@description.setter
|
|
def description(self, value: Optional[str]):
|
|
# TODO: Consider using typing.TypedDict when only Python 3.8+ is supported.
|
|
self._properties["description"] = value # type: ignore
|
|
|
|
@property
|
|
def friendly_name(self) -> Optional[str]:
|
|
"""Title of the table (defaults to :data:`None`)."""
|
|
return typing.cast(Optional[str], self._properties.get("friendlyName"))
|
|
|
|
@friendly_name.setter
|
|
def friendly_name(self, value: Optional[str]):
|
|
# TODO: Consider using typing.TypedDict when only Python 3.8+ is supported.
|
|
self._properties["friendlyName"] = value # type: ignore
|
|
|
|
@property
|
|
def labels(self) -> Dict[str, str]:
|
|
"""Labels for the table.
|
|
|
|
This method always returns a dict. To change a model's labels, modify the dict,
|
|
then call ``Client.update_model``. To delete a label, set its value to
|
|
:data:`None` before updating.
|
|
"""
|
|
return self._properties.setdefault("labels", {})
|
|
|
|
@labels.setter
|
|
def labels(self, value: Optional[Dict[str, str]]):
|
|
if value is None:
|
|
value = {}
|
|
self._properties["labels"] = value
|
|
|
|
@property
|
|
def encryption_configuration(self) -> Optional[EncryptionConfiguration]:
|
|
"""Custom encryption configuration for the model.
|
|
|
|
Custom encryption configuration (e.g., Cloud KMS keys) or :data:`None`
|
|
if using default encryption.
|
|
|
|
See `protecting data with Cloud KMS keys
|
|
<https://cloud.google.com/bigquery/docs/customer-managed-encryption>`_
|
|
in the BigQuery documentation.
|
|
"""
|
|
prop = self._properties.get("encryptionConfiguration")
|
|
if prop:
|
|
prop = EncryptionConfiguration.from_api_repr(prop)
|
|
return typing.cast(Optional[EncryptionConfiguration], prop)
|
|
|
|
@encryption_configuration.setter
|
|
def encryption_configuration(self, value: Optional[EncryptionConfiguration]):
|
|
api_repr = value.to_api_repr() if value else value
|
|
self._properties["encryptionConfiguration"] = api_repr
|
|
|
|
@classmethod
|
|
def from_api_repr(cls, resource: Dict[str, Any]) -> "Model":
|
|
"""Factory: construct a model resource given its API representation
|
|
|
|
Args:
|
|
resource:
|
|
Model resource representation from the API
|
|
|
|
Returns:
|
|
Model parsed from ``resource``.
|
|
"""
|
|
this = cls(None)
|
|
resource = copy.deepcopy(resource)
|
|
this._properties = resource
|
|
return this
|
|
|
|
def _build_resource(self, filter_fields):
|
|
"""Generate a resource for ``update``."""
|
|
return _helpers._build_resource_from_properties(self, filter_fields)
|
|
|
|
def __repr__(self):
|
|
return f"Model(reference={self.reference!r})"
|
|
|
|
def to_api_repr(self) -> Dict[str, Any]:
|
|
"""Construct the API resource representation of this model.
|
|
|
|
Returns:
|
|
Model reference represented as an API resource
|
|
"""
|
|
return copy.deepcopy(self._properties)
|
|
|
|
|
|
class ModelReference:
|
|
"""ModelReferences are pointers to models.
|
|
|
|
See
|
|
https://cloud.google.com/bigquery/docs/reference/rest/v2/models#modelreference
|
|
"""
|
|
|
|
def __init__(self):
|
|
self._properties = {}
|
|
|
|
@property
|
|
def project(self):
|
|
"""str: Project bound to the model"""
|
|
return self._properties.get("projectId")
|
|
|
|
@property
|
|
def dataset_id(self):
|
|
"""str: ID of dataset containing the model."""
|
|
return self._properties.get("datasetId")
|
|
|
|
@property
|
|
def model_id(self):
|
|
"""str: The model ID."""
|
|
return self._properties.get("modelId")
|
|
|
|
@property
|
|
def path(self) -> str:
|
|
"""URL path for the model's APIs."""
|
|
return f"/projects/{self.project}/datasets/{self.dataset_id}/models/{self.model_id}"
|
|
|
|
@classmethod
|
|
def from_api_repr(cls, resource: Dict[str, Any]) -> "ModelReference":
|
|
"""Factory: construct a model reference given its API representation.
|
|
|
|
Args:
|
|
resource:
|
|
Model reference representation returned from the API
|
|
|
|
Returns:
|
|
Model reference parsed from ``resource``.
|
|
"""
|
|
ref = cls()
|
|
ref._properties = resource
|
|
return ref
|
|
|
|
@classmethod
|
|
def from_string(
|
|
cls, model_id: str, default_project: Optional[str] = None
|
|
) -> "ModelReference":
|
|
"""Construct a model reference from model ID string.
|
|
|
|
Args:
|
|
model_id:
|
|
A model ID in standard SQL format. If ``default_project``
|
|
is not specified, this must included a project ID, dataset
|
|
ID, and model ID, each separated by ``.``.
|
|
default_project:
|
|
The project ID to use when ``model_id`` does not include
|
|
a project ID.
|
|
|
|
Returns:
|
|
Model reference parsed from ``model_id``.
|
|
|
|
Raises:
|
|
ValueError:
|
|
If ``model_id`` is not a fully-qualified table ID in
|
|
standard SQL format.
|
|
"""
|
|
proj, dset, model = _helpers._parse_3_part_id(
|
|
model_id, default_project=default_project, property_name="model_id"
|
|
)
|
|
return cls.from_api_repr(
|
|
{"projectId": proj, "datasetId": dset, "modelId": model}
|
|
)
|
|
|
|
def to_api_repr(self) -> Dict[str, Any]:
|
|
"""Construct the API resource representation of this model reference.
|
|
|
|
Returns:
|
|
Model reference represented as an API resource.
|
|
"""
|
|
return copy.deepcopy(self._properties)
|
|
|
|
def _key(self):
|
|
"""Unique key for this model.
|
|
|
|
This is used for hashing a ModelReference.
|
|
"""
|
|
return self.project, self.dataset_id, self.model_id
|
|
|
|
def __eq__(self, other):
|
|
if not isinstance(other, ModelReference):
|
|
return NotImplemented
|
|
return self._properties == other._properties
|
|
|
|
def __ne__(self, other):
|
|
return not self == other
|
|
|
|
def __hash__(self):
|
|
return hash(self._key())
|
|
|
|
def __repr__(self):
|
|
return "ModelReference(project_id='{}', dataset_id='{}', model_id='{}')".format(
|
|
self.project, self.dataset_id, self.model_id
|
|
)
|
|
|
|
|
|
class TransformColumn:
|
|
"""TransformColumn represents a transform column feature.
|
|
|
|
See
|
|
https://cloud.google.com/bigquery/docs/reference/rest/v2/models#transformcolumn
|
|
|
|
Args:
|
|
resource:
|
|
A dictionary representing a transform column feature.
|
|
"""
|
|
|
|
def __init__(self, resource: Dict[str, Any]):
|
|
self._properties = resource
|
|
|
|
@property
|
|
def name(self) -> Optional[str]:
|
|
"""Name of the column."""
|
|
return self._properties.get("name")
|
|
|
|
@property
|
|
def type_(self) -> Optional[standard_sql.StandardSqlDataType]:
|
|
"""Data type of the column after the transform.
|
|
|
|
Returns:
|
|
Optional[google.cloud.bigquery.standard_sql.StandardSqlDataType]:
|
|
Data type of the column.
|
|
"""
|
|
type_json = self._properties.get("type")
|
|
if type_json is None:
|
|
return None
|
|
return standard_sql.StandardSqlDataType.from_api_repr(type_json)
|
|
|
|
@property
|
|
def transform_sql(self) -> Optional[str]:
|
|
"""The SQL expression used in the column transform."""
|
|
return self._properties.get("transformSql")
|
|
|
|
@classmethod
|
|
def from_api_repr(cls, resource: Dict[str, Any]) -> "TransformColumn":
|
|
"""Constructs a transform column feature given its API representation
|
|
|
|
Args:
|
|
resource:
|
|
Transform column feature representation from the API
|
|
|
|
Returns:
|
|
Transform column feature parsed from ``resource``.
|
|
"""
|
|
this = cls({})
|
|
resource = copy.deepcopy(resource)
|
|
this._properties = resource
|
|
return this
|
|
|
|
|
|
def _model_arg_to_model_ref(value, default_project=None):
|
|
"""Helper to convert a string or Model to ModelReference.
|
|
|
|
This function keeps ModelReference and other kinds of objects unchanged.
|
|
"""
|
|
if isinstance(value, str):
|
|
return ModelReference.from_string(value, default_project=default_project)
|
|
if isinstance(value, Model):
|
|
return value.reference
|
|
return value
|