def predict(self, data):
# TODO: this evaluates a single example, i.e. mini-batch of one.
# generalize this to general case.
if len(data.shape) == 1:
data = data.reshape((1, data.shape[0]))
classification_response = self.skil.api.multipredict(
deployment_name=self.deployment.name,
model_name=self.model_name,
version_name="default",
body=skil_client.MultiPredictRequest(
id=str(uuid.uuid1()),
needs_pre_processing=False,
inputs=[
skil_client.INDArray(ordering='c',
shape=list(data.shape),
data=data.tolist()[0]),
skil_client.
INDArray( # This is the keep_prob placeholder data
ordering='c',
shape=[1],
data=[1.0])
]))
output = classification_response.outputs[0]
prediction = np.asarray(output.data)
shape = output.shape
return prediction.reshape(shape)
def _indarray(np_array):
"""Convert a numpy array to `skil_client.INDArray` instance.
# Arguments
np_array: `numpy.ndarray` instance.
# Returns
`skil_client.INDArray` instance.
"""
return skil_client.INDArray(ordering='c',
shape=list(np_array.shape),
data=np_array.reshape(-1).tolist())