File size: 13,061 Bytes
3193ef1 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 | # Copyright 2024 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
#
# http://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.
#
# pylint: disable=protected-access
"""Classes to support Tuning."""
from typing import Dict, List, Optional, Union
from google.auth import credentials as auth_credentials
from google.cloud import aiplatform
from google.cloud.aiplatform import base as aiplatform_base
from google.cloud.aiplatform import compat
from google.cloud.aiplatform import initializer as aiplatform_initializer
from google.cloud.aiplatform import jobs
from google.cloud.aiplatform import utils as aiplatform_utils
from google.cloud.aiplatform.utils import _ipython_utils
from google.cloud.aiplatform_v1.services import (
gen_ai_tuning_service as gen_ai_tuning_service_v1,
)
from google.cloud.aiplatform_v1beta1.services import (
gen_ai_tuning_service as gen_ai_tuning_service_v1beta1,
)
from google.cloud.aiplatform_v1beta1.types import (
tuning_job as gca_tuning_job_types,
)
from google.cloud.aiplatform_v1beta1 import types as gca_types
from google.rpc import status_pb2 # type: ignore
_LOGGER = aiplatform_base.Logger(__name__)
class TuningJobClientWithOverride(aiplatform_utils.ClientWithOverride):
_is_temporary = True
_default_version = compat.V1BETA1
_version_map = (
(compat.V1, gen_ai_tuning_service_v1.client.GenAiTuningServiceClient),
(compat.V1BETA1, gen_ai_tuning_service_v1beta1.client.GenAiTuningServiceClient),
)
class TuningJob(aiplatform_base._VertexAiResourceNounPlus):
"""Represents a TuningJob that runs with Google owned models."""
_resource_noun = "tuningJobs"
_getter_method = "get_tuning_job"
_list_method = "list_tuning_jobs"
_cancel_method = "cancel_tuning_job"
_delete_method = "delete_tuning_job"
_parse_resource_name_method = "parse_tuning_job_path"
_format_resource_name_method = "tuning_job_path"
_job_type = "tuning/tuningJob"
_has_displayed_experiments_button = False
client_class = TuningJobClientWithOverride
_gca_resource: gca_tuning_job_types.TuningJob
api_client: gen_ai_tuning_service_v1beta1.client.GenAiTuningServiceClient
def __init__(self, tuning_job_name: str):
super().__init__(resource_name=tuning_job_name)
self._gca_resource: gca_tuning_job_types.TuningJob = self._get_gca_resource(
resource_name=tuning_job_name
)
def refresh(self) -> "TuningJob":
"""Refreshed the tuning job from the service."""
self._gca_resource: gca_tuning_job_types.TuningJob = self._get_gca_resource(
resource_name=self.resource_name
)
if self.experiment and not self._has_displayed_experiments_button:
self._has_displayed_experiments_button = True
_ipython_utils.display_experiment_button(self.experiment)
return self
@property
def tuned_model_name(self) -> Optional[str]:
return self._gca_resource.tuned_model.model
@property
def tuned_model_endpoint_name(self) -> Optional[str]:
return self._gca_resource.tuned_model.endpoint
@property
def _experiment_name(self) -> Optional[str]:
return self._gca_resource.experiment
@property
def experiment(self) -> Optional[aiplatform.Experiment]:
if self._experiment_name:
return aiplatform.Experiment(experiment_name=self._experiment_name)
@property
def state(self) -> gca_types.JobState:
return self._gca_resource.state
@property
def service_account(self) -> Optional[str]:
self._assert_gca_resource_is_available()
return self._gca_resource.service_account
@property
def has_ended(self):
return self.state in jobs._JOB_COMPLETE_STATES
@property
def has_succeeded(self):
return self.state == gca_types.JobState.JOB_STATE_SUCCEEDED
@property
def error(self) -> Optional[status_pb2.Status]:
return self._gca_resource.error
@property
def tuning_data_statistics(self) -> gca_tuning_job_types.TuningDataStats:
return self._gca_resource.tuning_data_stats
@classmethod
def _create(
cls,
*,
base_model: str,
tuning_spec: Union[
gca_tuning_job_types.SupervisedTuningSpec,
gca_tuning_job_types.DistillationSpec,
],
tuned_model_display_name: Optional[str] = None,
description: Optional[str] = None,
labels: Optional[Dict[str, str]] = None,
project: Optional[str] = None,
location: Optional[str] = None,
credentials: Optional[auth_credentials.Credentials] = None,
) -> "TuningJob":
r"""Submits TuningJob.
Args:
base_model (str):
Model name for tuning, e.g., "gemini-1.0-pro"
or "gemini-1.0-pro-001".
This field is a member of `oneof`_ ``source_model``.
tuning_spec: Tuning Spec for Fine Tuning.
Supported types: SupervisedTuningSpec, DistillationSpec.
tuned_model_display_name: The display name of the
[TunedModel][google.cloud.aiplatform.v1.Model]. The name can
be up to 128 characters long and can consist of any UTF-8
characters.
description: The description of the `TuningJob`.
labels: The labels with user-defined metadata to organize
[TuningJob][google.cloud.aiplatform.v1.TuningJob] and
generated resources such as
[Model][google.cloud.aiplatform.v1.Model] and
[Endpoint][google.cloud.aiplatform.v1.Endpoint].
Label keys and values can be no longer than 64 characters
(Unicode codepoints), can only contain lowercase letters,
numeric characters, underscores and dashes. International
characters are allowed.
See https://goo.gl/xmQnxf for more information and examples
of labels.
project: Project to run the tuning job in.
Overrides project set in aiplatform.init.
location: Location to run the tuning job in.
Overrides location set in aiplatform.init.
credentials: Custom credentials to use to call tuning job service.
Overrides credentials set in aiplatform.init.
Returns:
Submitted TuningJob.
Raises:
RuntimeError is tuning_spec kind is unsupported
"""
_LOGGER.log_create_with_lro(cls)
if not tuned_model_display_name:
tuned_model_display_name = cls._generate_display_name()
gca_tuning_job = gca_tuning_job_types.TuningJob(
base_model=base_model,
tuned_model_display_name=tuned_model_display_name,
description=description,
labels=labels,
# The tuning_spec one_of is set later
)
if isinstance(tuning_spec, gca_tuning_job_types.SupervisedTuningSpec):
gca_tuning_job.supervised_tuning_spec = tuning_spec
elif isinstance(tuning_spec, gca_tuning_job_types.DistillationSpec):
gca_tuning_job.distillation_spec = tuning_spec
else:
raise RuntimeError(f"Unsupported tuning_spec kind: {tuning_spec}")
if aiplatform_initializer.global_config.encryption_spec_key_name:
gca_tuning_job.encryption_spec.kms_key_name = (
aiplatform_initializer.global_config.encryption_spec_key_name
)
gca_tuning_job.service_account = (
aiplatform_initializer.global_config.service_account
)
tuning_job: TuningJob = cls._construct_sdk_resource_from_gapic(
gapic_resource=gca_tuning_job,
project=project,
location=location,
credentials=credentials,
)
parent = aiplatform_initializer.global_config.common_location_path(
project=project, location=location
)
created_gca_tuning_job = tuning_job.api_client.create_tuning_job(
parent=parent,
tuning_job=gca_tuning_job,
)
tuning_job._gca_resource = created_gca_tuning_job
_LOGGER.log_create_complete(
cls=cls,
resource=created_gca_tuning_job,
variable_name="tuning_job",
module_name="sft",
)
_LOGGER.info(f"View Tuning Job:\n{tuning_job._dashboard_url()}")
if tuning_job._experiment_name:
_LOGGER.info(f"View experiment:\n{tuning_job._experiment.dashboard_url}")
return tuning_job
def cancel(self):
self.api_client.cancel_tuning_job(name=self.resource_name)
@classmethod
def list(cls, filter: Optional[str] = None) -> List["TuningJob"]:
"""Lists TuningJobs.
Args:
filter: The standard list filter.
Returns:
A list of TuningJob objects.
"""
return cls._list(filter=filter)
def _dashboard_url(self) -> str:
"""Returns the Google Cloud console URL where job can be viewed."""
fields = self._parse_resource_name(self.resource_name)
location = fields.pop("location")
project = fields.pop("project")
job = list(fields.values())[0]
url = f"https://console.cloud.google.com/vertex-ai/generative/language/locations/{location}/tuning/tuningJob/{job}?project={project}"
return url
def rebase_tuned_model(
tuned_model_ref: str,
*,
# TODO(b/372291558): Add support for overriding tuning job config
# tuning_job_config: Optional["TuningJob"] = None,
artifact_destination: Optional[str] = None,
deploy_to_same_endpoint: Optional[bool] = False,
):
"""Re-runs fine tuning on top of a new foundational model.
Takes a legacy Tuned GenAI model Reference and creates a TuningJob based
on a new model.
Args:
tuned_model_ref: Required. TunedModel reference to retrieve
the legacy model information.
tuning_job_config: The TuningJob to be updated. Users
can use this TuningJob field to overwrite tuning
configs.
artifact_destination: The Google Cloud Storage location to write the artifacts.
deploy_to_same_endpoint:
Optional. By default, bison to gemini
migration will always create new model/endpoint,
but for gemini-1.0 to gemini-1.5 migration, we
default deploy to the same endpoint. See details
in this Section.
Returns:
The new TuningJob.
"""
parent = aiplatform_initializer.global_config.common_location_path(
project=aiplatform_initializer.global_config.project,
location=aiplatform_initializer.global_config.location,
)
if "/tuningJobs/" in tuned_model_ref:
gapic_tuned_model_ref = gca_types.TunedModelRef(
tuning_job=tuned_model_ref,
)
elif "/pipelineJobs/" in tuned_model_ref:
gapic_tuned_model_ref = gca_types.TunedModelRef(
pipeline_job=tuned_model_ref,
)
elif "/models/" in tuned_model_ref:
gapic_tuned_model_ref = gca_types.TunedModelRef(
tuned_model=tuned_model_ref,
)
else:
raise ValueError(f"Unsupported tuned_model_ref: {tuned_model_ref}.")
# gapic_tuning_job_config = tuning_job._gca_resource if tuning_job else None
gapic_tuning_job_config = None
gapic_artifact_destination = (
gca_types.GcsDestination(output_uri_prefix=artifact_destination)
if artifact_destination
else None
)
api_client: gen_ai_tuning_service_v1beta1.GenAiTuningServiceClient = (
TuningJob._instantiate_client(
location=aiplatform_initializer.global_config.location,
credentials=aiplatform_initializer.global_config.credentials,
)
)
rebase_operation = api_client.rebase_tuned_model(
gca_types.RebaseTunedModelRequest(
parent=parent,
tuned_model_ref=gapic_tuned_model_ref,
tuning_job=gapic_tuning_job_config,
artifact_destination=gapic_artifact_destination,
deploy_to_same_endpoint=deploy_to_same_endpoint,
)
)
_LOGGER.log_create_with_lro(TuningJob, lro=rebase_operation)
gapic_rebase_tuning_job: gca_types.TuningJob = rebase_operation.result()
rebase_tuning_job = TuningJob._construct_sdk_resource_from_gapic(
gapic_resource=gapic_rebase_tuning_job,
)
return rebase_tuning_job
|