File size: 26,842 Bytes
747451d | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 | # /*---------------------------------------------------------------------------------------------
# * Copyright (c) 2023 STMicroelectronics. All rights reserved.
# * This software is licensed under terms that can be found in the LICENSE
# * file in the root directory of this software component.
# * If no LICENSE file comes with this software, it is provided AS-IS.
# *--------------------------------------------------------------------------------------------*/
import os
import functools
import typing
from common.stm32ai_dc.backend.cloud.generate_nbg_service import GenerateNbgService
from common.stm32ai_dc.backend.cloud.benchmark_service import BenchmarkService
from common.stm32ai_dc.backend.cloud.file_service import FileService
from common.stm32ai_dc.backend.cloud.helpers import get_supported_versions
from common.stm32ai_dc.backend.cloud.login_service import LoginService
from common.stm32ai_dc.backend.cloud.user_service import UserService
from common.stm32ai_dc.backend.cloud.stm32ai_service import Stm32AiService
from common.stm32ai_dc.errors import GenerateNbgFailure, AnalyzeServerError, BenchmarkServerError, InvalidCrendetialsException
from common.stm32ai_dc.errors import FileFormatError, GenerateServerError
from common.stm32ai_dc.errors import InternalErrorThatShouldNotHappened
from common.stm32ai_dc.errors import ParameterError, ValidateServerError
from common.stm32ai_dc.errors import LoginFailureException
from common.stm32ai_dc.types import AnalyzeResult, BackendVersionType, BenchmarkResult, BoardData, MpuBenchmarkResult, MpuParameters
from common.stm32ai_dc.types import GenerateResult, Stm32AiBackend, CliParameters
from common.stm32ai_dc.types import ValidateResult, ValidateResultMetrics
class CloudBackend(Stm32AiBackend):
def __init__(
self,
username: str,
password: str,
version: typing.Union[str, None] = None,
platform: BackendVersionType = BackendVersionType.STM32,
silent = False,
) -> None:
self.username = username
self.password = password
self.version = version
self.supportedVersions = get_supported_versions()
self.silent = silent
self.login_service = LoginService()
try:
# user_version = next(x for x in self.supportedVersions if x.get('platform',{}).get(platform, x['version']) == version)
# self.version = user_version['version']
self.version = self.supportedVersions['version']
except Exception as e:
print(
f'[WARN] Version {self.version} for platform {platform} is not supported by Developer Cloud.')
for v in self.supportedVersions:
if (v.get('isLatest', False) == True):
latest = v
if (latest):
self.version = latest.get('version', None)
print(
f"[WARN] It will use the latest version by default ({self.version}, {platform} version: {latest.get('platform').get(platform, '---')})")
if username is None or password is None or not len(username) or not len(password):
# Try to use previous tokens saved in home directory
sso_resp = self.login_service.get_access_token()
if (sso_resp):
self.auth_token = sso_resp
else:
raise LoginFailureException(
username,
password,
details='Empty login or stored token invalid.')
else:
try:
self.auth_token = self.login_service.login(
username=username, password=password)
except InvalidCrendetialsException as e :
raise e
except Exception:
raise LoginFailureException(
username,
password,
details='Login failure, try again and please check your credentials and/or your network'
)
if self.auth_token is None:
raise LoginFailureException(
username,
password,
details='Please check your credentials')
self.user_service = UserService(self.auth_token)
self.stm32ai_service = Stm32AiService(self.auth_token, self.version, self.silent)
self.file_service = FileService(self.auth_token)
self.benchmark_service = BenchmarkService(self.auth_token, self.silent)
self.generate_nbg_service = GenerateNbgService(self.auth_token)
def get_user(self):
return self.user_service.get_user()
def analyze(self, options: CliParameters) -> AnalyzeResult:
rid = self.stm32ai_service.trigger_analyze(options)
result = self.stm32ai_service.wait_for_run(rid)
if result is None:
if 'message' in result:
raise AnalyzeServerError(f"Missing data in server \
response: {result['message']}")
raise AnalyzeServerError('Missing data in server response')
report = result.get('report', None)
graph = result.get('graph', None)
info = result.get('info', None)
cinfo_graph = info['graphs'][0] if info is not None else None
date_time = info['environment']['generated_model']['generated_time'] if info is not None else report['date_time']
cli_version_str = info['environment']['tools'][0]['version'] if info is not None else report['cli_version_str']
cli_parameters = info['environment']['tools'][0]['arguments'] if info is not None else report['cli_parameters']
memory_footprint = graph.get('memory_footprint', {}) if graph else info.get('memory_footprint', {})
macc = functools.reduce(lambda a, b: a+b, map(lambda a: a['macc'], cinfo_graph['nodes']), 0) if cinfo_graph \
else graph['macc'] if graph \
else 0
analyze_result = AnalyzeResult(
activations_size=memory_footprint.get('activations', 0),
weights=memory_footprint.get('weights', 0),
macc=macc,
rom_size=memory_footprint.get('weights', 0)
+ memory_footprint.get('kernel_flash', 0)
+ memory_footprint.get('toolchain_flash', 0),
ram_size=memory_footprint.get('activations', 0)
+ functools.reduce(
lambda a, b: a+b,
memory_footprint.get('io', []), 0)
+ memory_footprint.get('kernel_ram', 0)
+ memory_footprint.get('toolchain_ram',
memory_footprint.get('extra_ram', 0)),
total_ram_io_size=functools.reduce(
lambda a, b: a+b,
memory_footprint.get('io', []), 0),
report=report,
graph=graph,
info=info,
date_time=date_time,
cli_version_str=cli_version_str,
cli_parameters=cli_parameters,
estimated_library_flash_size=memory_footprint.get(
'kernel_flash', -1),
estimated_library_ram_size=memory_footprint.get(
'kernel_ram', -1)
)
return analyze_result
def generate(self, options: CliParameters) -> GenerateResult:
import zipfile # Local import to avoid global perf drawback
if not os.path.exists(options.output):
os.makedirs(options.output, exist_ok=True)
rid = self.stm32ai_service.trigger_generate(options)
result = self.stm32ai_service.wait_for_run(rid)
if result is None or 'url' not in result:
if 'message' in result:
raise GenerateServerError(f"Missing data in server \
response: {result.get('message')}")
raise GenerateServerError('Missing data in server response')
local_filepath = self.file_service.download_generated_file(
result.get('url', None),
options.output)
if local_filepath:
# TODO: Check is file is really zip file or JSON content
if zipfile.is_zipfile(local_filepath):
zfile = zipfile.ZipFile(local_filepath)
zfile.extractall(options.output)
zfile.close()
return GenerateResult(
server_url=result.get('url', None),
zipfile_path=local_filepath, output_path=options.output,
graph=result.get('graph', {}),
report=result.get('report', {}))
else:
raise FileFormatError("A zip file was expected \
as server reply")
raise InternalErrorThatShouldNotHappened("No local file received")
def validate(self, options: CliParameters) -> ValidateResult:
rid = self.stm32ai_service.trigger_validate(options)
result = self.stm32ai_service.wait_for_run(rid)
if result is None:
if 'message' in result:
raise ValidateServerError(f"Missing data in server \
response: {result.get('message')}")
raise ValidateServerError('Missing data in server response')
report = result.get('report', {})
graph = result.get('graph', {})
cinfo = result.get('info', None)
if cinfo != None:
validate_result_metrics = []
for valmetrics in cinfo['validation']['val_metrics']:
validate_result_metrics.append(ValidateResultMetrics(
accuracy=valmetrics['acc'],
description=valmetrics['desc'],
l2r=valmetrics['l2r'],
mae=valmetrics['mae'],
mean=valmetrics['mean'],
rmse=valmetrics['rmse'],
std=valmetrics['std'],
ts_name=valmetrics.get('ts_name', '')
))
cinfo_graph = cinfo['graphs'][0]
memory_footprint=cinfo.get('memory_footprint', {})
validate_result = ValidateResult(
rom_size=memory_footprint['weights'],
macc=functools.reduce(lambda a, b: a+b, map(lambda a: a['macc'], cinfo_graph['nodes']), 0),
ram_size=memory_footprint['activations'],
total_ram_io_size=functools.reduce(
lambda a, b: a+b,
memory_footprint['io'], 0),
validation_metrics=validate_result_metrics,
validation_error=cinfo['validation']['val_error'],
validation_error_description=cinfo['validation']['val_error_desc'],
# rom_compression_factor=report['rom_cfact'],
# report_version=report['report_version'],
date_time=cinfo['environment']['generated_model']['generated_time'],
cli_version_str=cinfo['environment']['tools'][0]['version'],
cli_parameters=cinfo['environment']['tools'][0]['arguments'],
report=report,
graph=graph,
info=cinfo,
estimated_library_flash_size=int(memory_footprint.get("kernel_flash", 0) or 0)
+ int(memory_footprint.get("extra_flash", 0) or 0)
+ int(memory_footprint.get("toolchain_flash", 0) or 0),
estimated_library_ram_size=int(memory_footprint.get("kernel_ram", 0) or 0)
+ int(memory_footprint.get("extra_ram", 0) or 0)
+ int(memory_footprint.get("toolchain_ram", 0) or 0),
)
return validate_result
else:
validate_result_metrics = []
for valmetrics in report['val_metrics']:
validate_result_metrics.append(ValidateResultMetrics(
accuracy=valmetrics['acc'],
description=valmetrics['desc'],
l2r=valmetrics['l2r'],
mae=valmetrics['mae'],
mean=valmetrics['mae'],
rmse=valmetrics['rmse'],
std=valmetrics['std'],
ts_name=valmetrics['ts_name']
))
validate_result = ValidateResult(
rom_size=report['rom_size'],
macc=report['rom_n_macc'],
ram_size=report['ram_size'][0],
total_ram_io_size=functools.reduce(
lambda a, b: a+b,
report.get('ram_io_size', 0), 0),
validation_metrics=validate_result_metrics,
validation_error=report['val_error'],
validation_error_description=report['val_error_desc'],
# rom_compression_factor=report['rom_cfact'],
# report_version=report['report_version'],
date_time=report['date_time'],
cli_version_str=report['cli_version_str'],
cli_parameters=report['cli_parameters'],
report=report,
graph=graph,
info=None,
estimated_library_flash_size=memory_footprint.get("kernel_flash", 0)
+ memory_footprint.get("toolchain_flash", 0)
+ memory_footprint.get("extra_flash", 0),
estimated_library_ram_size=memory_footprint.get("kernel_ram", 0)
+ memory_footprint.get("extra_ram", 0)
+ memory_footprint.get("toolchain_ram", 0),
)
return validate_result
def benchmark(self, options: typing.Union[CliParameters, MpuParameters], board_name: str, timeout: int):
if options.model not in list(map(lambda x: x['name'], self.file_service.list_models())):
raise ParameterError("options.model should be a file name that is \
already uploaded on the cloud")
bid = self.benchmark_service.trigger_benchmark(
options, board_name, self.version)
result = self.benchmark_service.wait_for_run(bid, timeout=timeout)
if (isinstance(options, MpuParameters)):
benchmark_report = result.get('benchmark', {})
exec_time = benchmark_report.get('exec_time', {})
tool_version = benchmark_report.get('tool_version', {})
benchmark_result = MpuBenchmarkResult(
device=benchmark_report.get('board', 'N/A'),
duration_ms=exec_time.get('duration_ms', -1),
npu_percent=exec_time.get('npu_percent', -1),
gpu_percent=exec_time.get('gpu_percent', -1),
cpu_percent=exec_time.get('cpu_percent', -1),
rom_size=benchmark_report.get('model_size', -1),
ram_size=benchmark_report.get('ram_peak', -1),
date_time=benchmark_report.get('date'),
info=benchmark_report,
cli_version_str=tool_version.get('name') + ' ' + tool_version.get('major') + '.' + tool_version.get('minor') + '.' + tool_version.get('micro')
)
return benchmark_result
elif (isinstance(options, CliParameters)):
if result.get('benchmark', {}).get('info', None) != None:
cinfo = result['benchmark']['info']
memory_mapping = result.get('memoryMapping', {}) if result.get('memoryMapping') is not None else {}
report = result['benchmark']['report']
validate_result_metrics = []
for valmetrics in cinfo.get('validation', {}).get('val_metrics', []):
validate_result_metrics.append(ValidateResultMetrics(
accuracy=valmetrics['acc'],
description=valmetrics['desc'],
l2r=valmetrics['l2r'],
mae=valmetrics['mae'],
mean=valmetrics['mean'],
rmse=valmetrics['rmse'],
std=valmetrics['std'],
ts_name=valmetrics.get('ts_name', '')
))
cinfo_graph = cinfo['graphs'][0]
graph = result['benchmark']['graph']
exec_time = cinfo_graph.get('exec_time') if cinfo_graph else graph.get('exec_time', {})
c_arrays = {}
for arr in [b for b in cinfo['buffers'] if b['is_param'] is True]:
c_arrays[arr['name']] = arr
memory_footprint = memory_mapping.get('memoryFootprint', cinfo.get("memory_footprint", {}))
memory_pools = [p for p in cinfo.get('memory_pools', []) if p.get('size_bytes', -1) >= 0 and not p.get('virtual', False) and len(p.get('subpools', [])) == 0 and p.get('rights', 'ACC_WRITE') != 'ACC_READ']
internal_memory_pools = [p for p in memory_pools if "EXTERNAL" not in p['name'] and 0x60000000 > int(p.get('address', 0), 16 if p.get('address', '').lower().startswith('0x') else 10)]
external_memory_pools = [p for p in memory_pools if "EXTERNAL" in p['name'] or 0x60000000 <= int(p.get('address', 0), 16 if p.get('address', '').lower().startswith('0x') else 10)]
layers_in_internal_flash = memory_mapping.get('layersInExternalFlash', [])
layers_in_external_flash = memory_mapping.get('layersInInternalFlash', [])
internal_flash_usage = functools.reduce(lambda a, b: a+b, map(lambda a: c_arrays.get(a + "_array", {}).get('size_bytes', 0), layers_in_internal_flash), 0)
external_flash_usage = functools.reduce(lambda a, b: a+b, map(lambda a: c_arrays.get(a + "_array", {}).get('size_bytes', 0), layers_in_external_flash), 0)
internal_ram_consumption=functools.reduce(lambda a, b: a + b.get("used_size_bytes", 0), internal_memory_pools, 0)
external_ram_consumption=functools.reduce(lambda a, b: a + b.get("used_size_bytes", 0), external_memory_pools, 0)
benchmark_result = BenchmarkResult(
activations_size=memory_footprint.get('activations', 0),
weights=memory_footprint.get('weights', 0),
rom_size=memory_footprint.get('weights', 0)
+ memory_footprint.get('kernel_flash', 0)
+ memory_footprint.get('toolchain_flash', 0),
ram_size=memory_footprint.get('activations', 0)
+ functools.reduce(
lambda a, b: a+b,
memory_footprint.get('io', []), 0)
+ memory_footprint.get('kernel_ram', 0)
+ memory_footprint.get('toolchain_ram',
memory_footprint.get('extra_ram', 0)),
macc=functools.reduce(lambda a, b: a+b, map(lambda a: a['macc'], cinfo_graph['nodes']), 0),
total_ram_io_size=functools.reduce(
lambda a, b: a+b,
memory_footprint['io'], 0),
validation_metrics=validate_result_metrics,
validation_error=cinfo.get('validation', {}).get('val_error', ''),
validation_error_description=cinfo.get('validation', {}).get('val_error_desc', ''),
# rom_compression_factor=report['rom_cfact'],
# report_version=report['report_version'],
date_time=cinfo['environment']['generated_model']['generated_time'],
cli_version_str=cinfo['environment']['tools'][0]['version'],
cli_parameters=cinfo['environment']['tools'][0]['arguments'],
report=report,
graph=graph,
info=cinfo,
cycles=exec_time.get('cycles', -1),
duration_ms=exec_time.get('duration_ms', -1),
device=cinfo['environment'].get('device', ""),
cycles_by_macc=exec_time.get('cycles_by_macc', -1),
estimated_library_flash_size=int(memory_footprint.get("kernel_flash", 0) or 0)
+ int(memory_footprint.get("extra_flash", 0) or 0)
+ int(memory_footprint.get("toolchain_flash", 0) or 0),
estimated_library_ram_size=int(memory_footprint.get("kernel_ram", 0) or 0)
+ int(memory_footprint.get("extra_ram", 0) or 0)
+ int(memory_footprint.get("toolchain_ram", 0) or 0),
use_external_ram=memory_mapping.get("useExternalRam", external_ram_consumption > 0),
use_external_flash=memory_mapping.get("useExternalFlash", False),
internal_ram_consumption=internal_ram_consumption,
external_ram_consumption=external_ram_consumption,
internal_flash_consumption=internal_flash_usage,
external_flash_consumption=external_flash_usage,
)
return benchmark_result
elif result:
benchmark = result.get('benchmark', {})
memory_mapping = result.get('memoryMapping', {}) if result.get('memoryMapping') is not None else {}
report = benchmark['report']
validate_result_metrics = []
for valmetrics in report['val_metrics']:
validate_result_metrics.append(ValidateResultMetrics(
accuracy=valmetrics['acc'],
description=valmetrics['desc'],
l2r=valmetrics['l2r'],
mae=valmetrics['mae'],
mean=valmetrics['mean'],
rmse=valmetrics['rmse'],
std=valmetrics['std'],
ts_name=valmetrics['ts_name']
))
graph = benchmark['graph']
exec_time = graph['exec_time']
c_arrays = {}
for arr in graph.get('c_arrays', []):
c_arrays[arr['name']] = arr
memory_footprint = memory_mapping.get('memoryFootprint', graph.get("memory_footprint", {}))
memory_pools = [p for p in graph.get('memory_pools', []) if p.get('size_bytes', -1) >= 0]
internal_memory_pools = [p for p in memory_pools if "EXTERNAL" not in p['name']]
external_memory_pools = [p for p in memory_pools if "EXTERNAL" in p['name']]
layers_in_internal_flash = memory_mapping.get('layersInExternalFlash', [])
layers_in_external_flash = memory_mapping.get('layersInInternalFlash', [])
internal_flash_usage = functools.reduce(lambda a, b: a+b, map(lambda a: c_arrays.get(a + "_array").get('c_size_in_byte'), layers_in_internal_flash), 0)
external_flash_usage = functools.reduce(lambda a, b: a+b, map(lambda a: c_arrays.get(a + "_array").get('c_size_in_byte'), layers_in_external_flash), 0)
benchmark_result = BenchmarkResult(
activations_size=memory_footprint.get('activations', 0),
weights=memory_footprint.get('weights', 0),
rom_size=report['rom_size'],
macc=report['rom_n_macc'],
ram_size=report['ram_size'][0],
total_ram_io_size=functools.reduce(
lambda a, b: a+b,
report.get('ram_io_size', 0), 0),
validation_metrics=validate_result_metrics,
validation_error=report['val_error'],
validation_error_description=report['val_error_desc'],
# rom_compression_factor=report['rom_cfact'],
# report_version=report['report_version'],
date_time=report['date_time'],
cli_version_str=report['cli_version_str'],
cli_parameters=report['cli_parameters'],
report=report,
graph=graph,
info=None,
cycles=exec_time.get('cycles', -1),
duration_ms=exec_time.get('duration_ms', -1),
device=exec_time.get('device', ""),
cycles_by_macc=exec_time.get('cycles_by_macc', -1),
estimated_library_flash_size=memory_footprint.get("kernel_flash", 0)
+ memory_footprint.get("extra_flash", 0)
+ memory_footprint.get("toolchain_flash", 0),
estimated_library_ram_size=memory_footprint.get("kernel_ram", 0)
+ memory_footprint.get("extra_ram", 0)
+ memory_footprint.get("toolchain_ram", 0),
use_external_ram=memory_mapping.get("useExternalRam", False),
use_external_flash=memory_mapping.get("useExternalFlash", False),
internal_ram_consumption=functools.reduce(lambda a, b: a + b.get("used_size", 0), internal_memory_pools, 0),
external_ram_consumption=functools.reduce(lambda a, b: a + b.get("used_size", 0), external_memory_pools, 0),
internal_flash_consumption=internal_flash_usage,
external_flash_consumption=external_flash_usage,
)
return benchmark_result
else:
raise BenchmarkServerError("Benchmark service return wrong format")
def generate_nbg(self, model_name, timeout = 300):
bid = self.generate_nbg_service.trigger_optimize(
model_name)
result = self.generate_nbg_service.wait_for_run(bid, timeout=timeout)
if result.get('blobName', None) != None:
return result.get('blobName')
else:
raise GenerateNbgFailure('NBG Generation failed. No file found')
def get_benchmark_boards(self):
out: typing.List[BoardData] = []
boards_data = self.benchmark_service.list_boards()
for k in boards_data.keys():
out.append(BoardData(
name=k,
boardCount=boards_data[k].get('boardCount', -1),
flashSize=boards_data[k].get('flashSize', ''),
deviceCpu=boards_data[k].get('deviceCpu', ''),
deviceId=boards_data[k].get('deviceId', '')))
return out
def list_models(self):
return self.file_service.list_models()
def upload_model(self, model_path: str):
return self.file_service.upload_model(model_path)
def download_model(self, model_name: str, path: str):
return self.file_service.download_model(model_name, path)
def delete_model(self, model_name: str):
return self.file_service.delete_model(model_name)
|