from __future__ import annotations
import logging
import os
import stat
import sys
from pathlib import Path
from typing import Any
from qualia_core.deployment.toolchain.CMake import CMake
from qualia_core.evaluation.target.STM32CubeAI import STM32CubeAI as STM32CubeAIEvaluator
from qualia_core.typing import TYPE_CHECKING
if TYPE_CHECKING:
if sys.version_info >= (3, 11):
from typing import Self
else:
from typing_extensions import Self
from qualia_core.postprocessing.Converter import Converter # noqa: TC001
from qualia_core.postprocessing.Keras2TFLite import Keras2TFLite # noqa: TC001
if sys.version_info >= (3, 12):
from typing import override
else:
from typing_extensions import override
logger = logging.getLogger(__name__)
[docs]
class STM32CubeAICMake(CMake):
evaluator = STM32CubeAIEvaluator # Suggested evaluator
def __init__(self,
projectdir: str | Path,
outdir: str | Path) -> None:
super().__init__(projectdir=projectdir, outdir=outdir)
# Built-in project made for 8.1.0
self.__stm32cubeai_bin = Path.home()/'STM32Cube'/'Repository'/'Packs'/'STMicroelectronics'/'X-CUBE-AI'/'8.1.0'/'Utilities'/'linux'/'stm32ai'
#self.__stm32cubeai_bin = next((Path.home()/'STM32Cube'/'Repository'/'Packs'/'STMicroelectronics'/'X-CUBE-AI').glob('*'))/'Utilities'/'linux'/'stm32ai'
def _create_modeloutdir(self, modelpath: Path) -> None:
modelpath.mkdir(parents=True, exist_ok=True)
def __write_model(self, model: Keras2TFLite, modelpath: Path) -> None:
with modelpath.open('wb') as f:
_ = f.write(model.data)
def __generate(self, modelpath: Path, modeloutdir: Path, compression: int) -> bool:
if compression != 1:
logger.error('FIXME: Evaluation not logging compression level != 1')
raise ValueError
if compression not in [1, 4, 8]:
logger.error('Compression factor %s is not supported, must be either 1, 4 or 8', compression)
raise ValueError
# Sometimes the executable may not have have exec permission, set it
if not os.access(self.__stm32cubeai_bin, os.X_OK):
self.__stm32cubeai_bin.chmod(self.__stm32cubeai_bin.stat().st_mode | stat.S_IXUSR)
return self._run(self.__stm32cubeai_bin,
'generate',
'--model', str(modelpath),
'--output', str(modeloutdir),
'--compression', str(compression),
)
@override
def _build(self,
modeldir: Path,
optimize: str,
outdir: Path) -> bool:
args = ('-D', f'NETWORK_DIR={modeldir.resolve()!s}')
return self._run_cmake(args=args, projectdir=self._projectdir, outdir=outdir)
[docs]
@override
def prepare(self,
tag: str,
model: Converter[Any],
optimize: str,
compression: int) -> Self | None:
# Keep here for isinstance() to avoid circual import
from qualia_core.postprocessing.Keras2TFLite import Keras2TFLite
if optimize != 'cmsis-nn':
logger.error('cmsis-nn optimize mandatory for %s', type(self).__name__)
raise ValueError
if not isinstance(model, Keras2TFLite):
logger.error('%s excepts the model to come from a Keras2TFLite Converter', type(self).__name__)
raise TypeError
outdir = self._outdir / tag
modeloutdir = Path('out')/'deploy'/'stm32cubeai'
modelpath = modeloutdir/f'{tag}.tflite'
self._create_modeloutdir(modeloutdir)
self._create_outdir(outdir)
self.__write_model(model=model, modelpath=modelpath)
if not self.__generate(modelpath=modelpath, modeloutdir=modeloutdir, compression=compression):
return None
if not self._build(modeldir=modeloutdir, optimize=optimize, outdir=outdir):
return None
return self