def test_anchor_text():
skmodel = SKLearnServer(MOVIE_MODEL_URI)
skmodel.load()
movies = fetch_movie_sentiment()
anchor_text = AnchorText(skmodel.predict, None)
np.random.seed(0)
explanation = anchor_text.explain(movies.data[4:5])
exp_json = json.loads(explanation.to_json())
assert exp_json["meta"]["name"] == "AnchorText"
def test_anchor_text():
os.environ.clear()
skmodel = SKLearnModel("adult", MOVIE_MODEL_URI)
skmodel.load()
predictor = Predictor(skmodel)
anchor_text = AnchorText(predictor.predict_fn, None)
movies = fetch_movie_sentiment()
np.random.seed(0)
explanation = anchor_text.explain(movies.data[4:5])
exp_json = json.loads(explanation.to_json())
print(exp_json["data"]["anchor"])
def test_anchor_text():
os.environ.clear()
skmodel = SKLearnServer(MOVIE_MODEL_URI)
skmodel.load()
movies = fetch_movie_sentiment()
anchor_text = AnchorText(skmodel.predict, None)
np.random.seed(0)
explanation = anchor_text.explain(movies.data[4:5])
exp_json = json.loads(explanation.to_json())
print(exp_json["data"]["anchor"])
def __init__(self,
name: str,
predictor_host: str,
method: ExplainerMethod,
config: Mapping,
explainer: object = None,
protocol: Protocol = Protocol.seldon_grpc,
tf_data_type: str = None,
keras_model: keras.Model = None):
super().__init__(name)
self.predictor_host = predictor_host
logging.info("Predict URL set to %s", self.predictor_host)
self.method = method
self.protocol = protocol
self.tf_data_type = tf_data_type
logging.info("Protocol is %s", str(self.protocol))
# Add type for first value to help pass mypy type checks
if self.method is ExplainerMethod.anchor_tabular:
self.wrapper: ExplainerWrapper = AnchorTabular(
self._predict_fn, explainer, **config)
elif self.method is ExplainerMethod.anchor_images:
self.wrapper = AnchorImages(self._predict_fn, explainer, **config)
elif self.method is ExplainerMethod.anchor_text:
self.wrapper = AnchorText(self._predict_fn, explainer, **config)
elif self.method is ExplainerMethod.kernel_shap:
self.wrapper = KernelShap(self._predict_fn, explainer, **config)
elif self.method is ExplainerMethod.integrated_gradients:
self.wrapper = IntegratedGradients(keras_model, **config)
elif self.method is ExplainerMethod.tree_shap:
self.wrapper = TreeShap(explainer, **config)
else:
raise NotImplementedError
class AlibiExplainer(kfserving.KFModel):
def __init__(self,
name: str,
predictor_host: str,
protocol: Protocol,
method: ExplainerMethod,
config: Mapping,
explainer: object = None):
super().__init__(name)
self.protocol = protocol
if self.protocol == Protocol.tensorflow_http:
self.predict_url = PREDICTOR_URL_FORMAT.format(predictor_host, name)
else:
self.predict_url = SELDON_PREDICTOR_URL_FORMAT.format(predictor_host)
logging.info("Predict URL set to %s", self.predict_url)
self.method = method
if self.method is ExplainerMethod.anchor_tabular:
self.wrapper = AnchorTabular(self._predict_fn, explainer, **config)
elif self.method is ExplainerMethod.anchor_images:
self.wrapper = AnchorImages(self._predict_fn, explainer, **config)
elif self.method is ExplainerMethod.anchor_text:
self.wrapper = AnchorText(self._predict_fn, explainer, **config)
else:
raise NotImplementedError
def load(self):
pass
def _predict_fn(self, arr: Union[np.ndarray, List]) -> np.ndarray:
if self.protocol == Protocol.seldon_http:
resp = SeldonRequestHandler.predict(arr, self.predict_url)
return np.array(resp)
elif self.protocol == Protocol.tensorflow_http:
inputs = []
for req_data in arr:
if isinstance(req_data, np.ndarray):
inputs.append(req_data.tolist())
else:
inputs.append(str(req_data))
resp = TensorflowRequestHandler.predict(inputs, self.predict_url)
return np.array(resp)
else:
raise NotImplementedError
def explain(self, inputs: List) -> Any:
if self.method is ExplainerMethod.anchor_tabular or self.method is ExplainerMethod.anchor_images or self.method is ExplainerMethod.anchor_text:
explanation = self.wrapper.explain(inputs)
logging.info("Explanation: %s", explanation)
return json.loads(json.dumps(explanation, cls=NumpyEncoder))
else:
raise NotImplementedError
def __init__( # pylint:disable=too-many-arguments
self,
name: str,
predictor_host: str,
method: ExplainerMethod,
config: Mapping,
explainer: object = None,
):
super().__init__(name)
self.predictor_host = predictor_host
logging.info("Predict URL set to %s", self.predictor_host)
self.method = method
if self.method is ExplainerMethod.anchor_tabular:
self.wrapper: ExplainerWrapper = AnchorTabular(
self._predict_fn, explainer, **config)
elif self.method is ExplainerMethod.anchor_images:
self.wrapper = AnchorImages(self._predict_fn, explainer, **config)
elif self.method is ExplainerMethod.anchor_text:
self.wrapper = AnchorText(self._predict_fn, explainer, **config)
else:
raise NotImplementedError
class AlibiExplainer(kserve.Model):
def __init__( # pylint:disable=too-many-arguments
self,
name: str,
predictor_host: str,
method: ExplainerMethod,
config: Mapping,
explainer: object = None,
):
super().__init__(name)
self.predictor_host = predictor_host
logging.info("Predict URL set to %s", self.predictor_host)
self.method = method
if self.method is ExplainerMethod.anchor_tabular:
self.wrapper: ExplainerWrapper = AnchorTabular(
self._predict_fn, explainer, **config
)
elif self.method is ExplainerMethod.anchor_images:
self.wrapper = AnchorImages(self._predict_fn, explainer, **config)
elif self.method is ExplainerMethod.anchor_text:
self.wrapper = AnchorText(self._predict_fn, explainer, **config)
else:
raise NotImplementedError
def load(self) -> bool:
pass
def _predict_fn(self, arr: Union[np.ndarray, List]) -> np.ndarray:
instances = []
for req_data in arr:
if isinstance(req_data, np.ndarray):
instances.append(req_data.tolist())
else:
instances.append(req_data)
loop = asyncio.get_running_loop() # type: ignore
resp = loop.run_until_complete(self.predict({"instances": instances}))
return np.array(resp["predictions"])
def explain(self, request: Dict) -> Any:
if (
self.method is ExplainerMethod.anchor_tabular
or self.method is ExplainerMethod.anchor_images
or self.method is ExplainerMethod.anchor_text
):
explanation = self.wrapper.explain(request["instances"])
explanationAsJsonStr = explanation.to_json()
logging.info("Explanation: %s", explanationAsJsonStr)
return json.loads(explanationAsJsonStr)
raise NotImplementedError
def __init__(self,
name: str,
predictor_host: str,
protocol: Protocol,
method: ExplainerMethod,
config: Mapping,
explainer: object = None):
super().__init__(name)
self.protocol = protocol
if self.protocol == Protocol.tensorflow_http:
self.predict_url = PREDICTOR_URL_FORMAT.format(predictor_host, name)
else:
self.predict_url = SELDON_PREDICTOR_URL_FORMAT.format(predictor_host)
logging.info("Predict URL set to %s", self.predict_url)
self.method = method
if self.method is ExplainerMethod.anchor_tabular:
self.wrapper = AnchorTabular(self._predict_fn, explainer, **config)
elif self.method is ExplainerMethod.anchor_images:
self.wrapper = AnchorImages(self._predict_fn, explainer, **config)
elif self.method is ExplainerMethod.anchor_text:
self.wrapper = AnchorText(self._predict_fn, explainer, **config)
else:
raise NotImplementedError