def set_dependencies(self, env: BentoServiceEnv):
# Note that keras module is not required, user can use tf.keras as an
# replacement for the keras module. Although tensorflow module is required to
# be used as the default Keras backend
pip_deps = ['tensorflow']
if self._keras_module_name == 'keras':
pip_deps.append('keras')
env.add_pip_packages(pip_deps)
def set_dependencies(self, env: BentoServiceEnv):
logger.warning(
"BentoML by default does not include spacy and torchvision package when "
"using FastaiModelArtifact. To make sure BentoML bundle those packages if "
"they are required for your model, either import those packages in "
"BentoService definition file or manually add them via "
"`@env(pip_packages=['torchvision'])` when defining a BentoService"
)
env.add_pip_packages(['torch', "fastai<2.0.0"])
def _config_environments(self):
self._env = self.__class__._env or BentoServiceEnv()
for api in self._inference_apis:
self._env.add_pip_packages(api.input_adapter.pip_dependencies)
self._env.add_pip_packages(api.output_adapter.pip_dependencies)
for artifact in self.artifacts.get_artifact_list():
artifact.set_dependencies(self.env)
def decorator(bento_service_cls):
bento_service_cls._env = BentoServiceEnv(
pip_packages=pip_packages or pip_dependencies,
pip_index_url=pip_index_url,
pip_trusted_host=pip_trusted_host,
pip_extra_index_url=pip_extra_index_url,
infer_pip_packages=infer_pip_packages or auto_pip_dependencies,
requirements_txt_file=requirements_txt_file,
conda_channels=conda_channels,
conda_dependencies=conda_dependencies,
conda_env_yml_file=conda_env_yml_file,
setup_sh=setup_sh,
docker_base_image=docker_base_image,
)
return bento_service_cls
def decorator(bento_service_cls):
artifact_names = set()
for artifact in artifacts:
if not isinstance(artifact, BentoServiceArtifact):
raise InvalidArgument(
"BentoService @artifacts decorator only accept list of "
"BentoServiceArtifact instances, instead got type: '%s'" %
type(artifact))
if artifact.name in artifact_names:
raise InvalidArgument(
"Duplicated artifact name `%s` detected. Each artifact within one"
"BentoService must have an unique name" % artifact.name)
artifact_names.add(artifact.name)
bento_service_cls._declared_artifacts = artifacts
bento_service_cls._env = BentoServiceEnv(infer_pip_packages=True)
return bento_service_cls
def set_dependencies(self, env: BentoServiceEnv):
if env._infer_pip_packages:
env.add_pip_packages(['xgboost'])
def set_dependencies(self, env: BentoServiceEnv):
if self.backend == 'onnxruntime':
env.add_pip_packages(['onnxruntime'])
def set_dependencies(self, env: BentoServiceEnv):
env.add_pip_packages(["fasttext"])
def set_dependencies(self, env: BentoServiceEnv):
env.add_pip_packages(["easyocr>=1.3.0"])
def set_dependencies(self, env: BentoServiceEnv):
env.add_pip_packages(['h2o'])
env.add_conda_dependencies(['openjdk'])
def set_dependencies(self, env: BentoServiceEnv):
if env._infer_pip_packages:
env.add_pip_packages(["mxnet"])
def set_dependencies(self, env: BentoServiceEnv):
env.add_pip_packages(["numpy"])
def set_dependencies(self, env: BentoServiceEnv):
env.add_pip_packages(['pytorch-lightning'])
def set_dependencies(self, env: BentoServiceEnv):
env.add_pip_packages(['tensorflow'])
def set_dependencies(self, env: BentoServiceEnv):
env.add_pip_packages(['torch', "detectron2"])
def set_dependencies(self, env: BentoServiceEnv):
if self.backend == "onnxruntime":
env.add_pip_packages(["onnxruntime"])
elif self.backend == "onnxruntime-gpu":
env.add_pip_packages(["onnxruntime-gpu"])
def set_dependencies(self, env: BentoServiceEnv):
env.add_pip_packages(['spacy'])
def set_dependencies(self, env: BentoServiceEnv):
env.add_pip_packages(['coremltools>=4.0b2'])
def set_dependencies(self, env: BentoServiceEnv):
if env._infer_pip_packages:
env.add_pip_packages(['scikit-learn'])
