Spaces:
Running
on
Zero
Running
on
Zero
File size: 10,035 Bytes
df6c67d |
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 |
import queue
from queue import Queue
from threading import Thread
from typing import Any, List, Optional
from inference.core import logger
from inference.core.active_learning.accounting import image_can_be_submitted_to_batch
from inference.core.active_learning.batching import generate_batch_name
from inference.core.active_learning.configuration import (
prepare_active_learning_configuration,
prepare_active_learning_configuration_inplace,
)
from inference.core.active_learning.core import (
execute_datapoint_registration,
execute_sampling,
)
from inference.core.active_learning.entities import (
ActiveLearningConfiguration,
Prediction,
PredictionType,
)
from inference.core.cache.base import BaseCache
from inference.core.utils.image_utils import load_image
MAX_REGISTRATION_QUEUE_SIZE = 512
class NullActiveLearningMiddleware:
def register_batch(
self,
inference_inputs: List[Any],
predictions: List[Prediction],
prediction_type: PredictionType,
disable_preproc_auto_orient: bool = False,
) -> None:
pass
def register(
self,
inference_input: Any,
prediction: dict,
prediction_type: PredictionType,
disable_preproc_auto_orient: bool = False,
) -> None:
pass
def start_registration_thread(self) -> None:
pass
def stop_registration_thread(self) -> None:
pass
def __enter__(self) -> "NullActiveLearningMiddleware":
return self
def __exit__(self, exc_type, exc_val, exc_tb) -> None:
pass
class ActiveLearningMiddleware:
@classmethod
def init(
cls, api_key: str, model_id: str, cache: BaseCache
) -> "ActiveLearningMiddleware":
configuration = prepare_active_learning_configuration(
api_key=api_key,
model_id=model_id,
cache=cache,
)
return cls(
api_key=api_key,
configuration=configuration,
cache=cache,
)
@classmethod
def init_from_config(
cls, api_key: str, model_id: str, cache: BaseCache, config: Optional[dict]
) -> "ActiveLearningMiddleware":
configuration = prepare_active_learning_configuration_inplace(
api_key=api_key,
model_id=model_id,
active_learning_configuration=config,
)
return cls(
api_key=api_key,
configuration=configuration,
cache=cache,
)
def __init__(
self,
api_key: str,
configuration: Optional[ActiveLearningConfiguration],
cache: BaseCache,
):
self._api_key = api_key
self._configuration = configuration
self._cache = cache
def register_batch(
self,
inference_inputs: List[Any],
predictions: List[Prediction],
prediction_type: PredictionType,
disable_preproc_auto_orient: bool = False,
) -> None:
for inference_input, prediction in zip(inference_inputs, predictions):
self.register(
inference_input=inference_input,
prediction=prediction,
prediction_type=prediction_type,
disable_preproc_auto_orient=disable_preproc_auto_orient,
)
def register(
self,
inference_input: Any,
prediction: dict,
prediction_type: PredictionType,
disable_preproc_auto_orient: bool = False,
) -> None:
self._execute_registration(
inference_input=inference_input,
prediction=prediction,
prediction_type=prediction_type,
disable_preproc_auto_orient=disable_preproc_auto_orient,
)
def _execute_registration(
self,
inference_input: Any,
prediction: dict,
prediction_type: PredictionType,
disable_preproc_auto_orient: bool = False,
) -> None:
if self._configuration is None:
return None
image, is_bgr = load_image(
value=inference_input,
disable_preproc_auto_orient=disable_preproc_auto_orient,
)
if not is_bgr:
image = image[:, :, ::-1]
matching_strategies = execute_sampling(
image=image,
prediction=prediction,
prediction_type=prediction_type,
sampling_methods=self._configuration.sampling_methods,
)
if len(matching_strategies) == 0:
return None
batch_name = generate_batch_name(configuration=self._configuration)
if not image_can_be_submitted_to_batch(
batch_name=batch_name,
workspace_id=self._configuration.workspace_id,
dataset_id=self._configuration.dataset_id,
max_batch_images=self._configuration.max_batch_images,
api_key=self._api_key,
):
logger.debug(f"Limit on Active Learning batch size reached.")
return None
execute_datapoint_registration(
cache=self._cache,
matching_strategies=matching_strategies,
image=image,
prediction=prediction,
prediction_type=prediction_type,
configuration=self._configuration,
api_key=self._api_key,
batch_name=batch_name,
)
class ThreadingActiveLearningMiddleware(ActiveLearningMiddleware):
@classmethod
def init(
cls,
api_key: str,
model_id: str,
cache: BaseCache,
max_queue_size: int = MAX_REGISTRATION_QUEUE_SIZE,
) -> "ThreadingActiveLearningMiddleware":
configuration = prepare_active_learning_configuration(
api_key=api_key,
model_id=model_id,
cache=cache,
)
task_queue = Queue(max_queue_size)
return cls(
api_key=api_key,
configuration=configuration,
cache=cache,
task_queue=task_queue,
)
@classmethod
def init_from_config(
cls,
api_key: str,
model_id: str,
cache: BaseCache,
config: Optional[dict],
max_queue_size: int = MAX_REGISTRATION_QUEUE_SIZE,
) -> "ThreadingActiveLearningMiddleware":
configuration = prepare_active_learning_configuration_inplace(
api_key=api_key,
model_id=model_id,
active_learning_configuration=config,
)
task_queue = Queue(max_queue_size)
return cls(
api_key=api_key,
configuration=configuration,
cache=cache,
task_queue=task_queue,
)
def __init__(
self,
api_key: str,
configuration: ActiveLearningConfiguration,
cache: BaseCache,
task_queue: Queue,
):
super().__init__(api_key=api_key, configuration=configuration, cache=cache)
self._task_queue = task_queue
self._registration_thread: Optional[Thread] = None
def register(
self,
inference_input: Any,
prediction: dict,
prediction_type: PredictionType,
disable_preproc_auto_orient: bool = False,
) -> None:
logger.debug(f"Putting registration task into queue")
try:
self._task_queue.put_nowait(
(
inference_input,
prediction,
prediction_type,
disable_preproc_auto_orient,
)
)
except queue.Full:
logger.warning(
f"Dropping datapoint registered in Active Learning due to insufficient processing "
f"capabilities."
)
def start_registration_thread(self) -> None:
if self._registration_thread is not None:
logger.warning(f"Registration thread already started.")
return None
logger.debug("Staring registration thread")
self._registration_thread = Thread(target=self._consume_queue)
self._registration_thread.start()
def stop_registration_thread(self) -> None:
if self._registration_thread is None:
logger.warning("Registration thread is already stopped.")
return None
logger.debug("Stopping registration thread")
self._task_queue.put(None)
self._registration_thread.join()
if self._registration_thread.is_alive():
logger.warning(f"Registration thread stopping was unsuccessful.")
self._registration_thread = None
def _consume_queue(self) -> None:
queue_closed = False
while not queue_closed:
queue_closed = self._consume_queue_task()
def _consume_queue_task(self) -> bool:
logger.debug("Consuming registration task")
task = self._task_queue.get()
logger.debug("Received registration task")
if task is None:
logger.debug("Terminating registration thread")
self._task_queue.task_done()
return True
inference_input, prediction, prediction_type, disable_preproc_auto_orient = task
try:
self._execute_registration(
inference_input=inference_input,
prediction=prediction,
prediction_type=prediction_type,
disable_preproc_auto_orient=disable_preproc_auto_orient,
)
except Exception as error:
# Error handling to be decided
logger.warning(
f"Error in datapoint registration for Active Learning. Details: {error}. "
f"Error is suppressed in favour of normal operations of registration thread."
)
self._task_queue.task_done()
return False
def __enter__(self) -> "ThreadingActiveLearningMiddleware":
self.start_registration_thread()
return self
def __exit__(self, exc_type, exc_val, exc_tb) -> None:
self.stop_registration_thread()
|