def test_encode(content_type):
mock_encoder = Mock()
with patch.dict(encoder._encoder_map, {content_type: mock_encoder},
clear=True):
encoder.encode(42, content_type)
mock_encoder.assert_called_once_with(42)
def default_output_fn(self, prediction, accept):
"""Return encoded prediction for the response.
Args:
prediction (obj): prediction returned by predict_fn .
accept (str): accept content-type expected by the client.
Returns:
encoded response for MMS to return to client
"""
accept_type = accept.lower()
try:
if accept_type == content_types.CSV or accept_type == 'csv':
if SAGEMAKER_BATCH:
return_data = "\n".join(map(str, prediction.tolist())) + '\n'
else:
# FIXME: this is invalid CSV and is only retained for backwards compatibility
return_data = ",".join(map(str, prediction.tolist()))
encoded_prediction = return_data.encode("utf-8")
elif accept_type == content_types.JSON or accept_type == 'json':
encoded_prediction = encoder.encode(prediction, accept_type)
else:
raise ValueError("{} is not an accepted Accept type. Please choose one of the following:"
" ['{}', '{}'].".format(accept, content_types.CSV, content_types.JSON))
except Exception as e:
raise UnsupportedMediaTypeInferenceError(
"Encoding to accept type {} failed with exception: {}".format(accept,
e))
return encoded_prediction
def output_fn(prediction, accept):
if type(prediction) == torch.Tensor:
prediction = prediction.detach().cpu().numpy().tolist()
encoded_prediction = encoder.encode(prediction, accept)
if accept == content_types.CSV:
encoded_prediction = encoded_prediction.encode("utf-8")
return encoded_prediction
def output_fn(self, prediction, accept):
"""A default output_fn for PyTorch. Serializes predictions from predict_fn to JSON, CSV or NPY format.
Args:
prediction: a prediction result from predict_fn
accept: type which the output data needs to be serialized
Returns: output data serialized
"""
return encoder.encode(prediction, accept)
def default_output_fn(self, prediction, accept):
"""Function responsible for serializing the prediction result to the desired accept type.
Args:
prediction (obj): prediction result returned by the predict_fn.
accept (str): accept header expected by the client.
Returns:
obj: prediction data.
"""
return encoder.encode(prediction, accept), accept
def default_output_fn(prediction, accept):
"""Function responsible to serialize the prediction for the response.
Args:
prediction (obj): prediction returned by predict_fn .
accept (str): accept content-type expected by the client.
Returns:
(worker.Response): a Flask response object with the following args:
* Args:
response: the serialized data to return
accept: the content-type that the data was transformed to.
"""
return encoder.encode(prediction, accept), accept
def default_output_fn(self, prediction, accept):
"""A default output_fn for PyTorch. Serializes predictions from predict_fn to JSON, CSV or NPY format.
Args:
prediction: a prediction result from predict_fn
accept: type which the output data needs to be serialized
Returns: output data serialized
"""
if type(prediction) == torch.Tensor:
prediction = prediction.detach().cpu().numpy().tolist()
encoded_prediction = encoder.encode(prediction, accept)
if accept == content_types.CSV:
encoded_prediction = encoded_prediction.encode("utf-8")
return encoded_prediction
def default_output_fn(self, prediction, accept): # pylint: disable=no-self-use
"""Function responsible for serializing the prediction result to the desired accept type.
Args:
prediction (obj): prediction result returned by the predict_fn.
accept (str): accept header expected by the client.
Returns:
obj: prediction data.
"""
for content_type in utils.parse_accept(accept):
if content_type in encoder.SUPPORTED_CONTENT_TYPES:
return encoder.encode(prediction, content_type), content_type
raise errors.UnsupportedFormatError(accept)
def default_output_fn(self, prediction, accept):
"""Serialize the prediction into a response.
Args:
prediction (mxnet.nd.array): an MXNet NDArray that is the result of a prediction
accept (str): the accept content type expected by the client
Returns:
obj: prediction data.
Raises:
sagemaker_inference.errors.UnsupportedFormatError: if an unsupported content type is used.
"""
if accept in self.VALID_CONTENT_TYPES:
return encoder.encode(prediction.asnumpy().tolist(), accept)
else:
raise errors.UnsupportedFormatError(accept)
def default_output_fn(self, prediction, accept):
"""A default output_fn for TensorFlow. Serializes predictions from predict_fn to JSON, CSV or NPY format.
Args:
prediction: a prediction result from predict_fn
accept: type which the output data needs to be serialized
Returns: output data serialized
"""
prediction = np.array(prediction).squeeze()
prediction[prediction > 0.5] = 1
prediction[prediction <= 0.5] = 0
# rescaling
prediction = (prediction * 255).astype(np.uint8)
return encoder.encode(prediction, accept)
def default_output_fn(self, prediction, accept):
"""
Gets the prediction output and format it to be returned to the user
Parameters
----------
prediction : pd.DataFrame
Predicted dataset
accept : string
Output type
Returns
-------
CSV : CSV file
"""
logging.info('Saving')
if accept != "text/csv":
raise Exception("Invalid accept: %s" % accept)
return encoder.encode(prediction, accept)
def default_output_fn(self, prediction, accept):
"""A default output_fn for PyTorch. Serializes predictions from predict_fn to JSON, CSV or NPY format.
Args:
prediction: a prediction result from predict_fn
accept: type which the output data needs to be serialized
Returns: output data serialized
"""
if type(prediction) == torch.Tensor:
prediction = prediction.detach().cpu().numpy().tolist()
for content_type in utils.parse_accept(accept):
if content_type in encoder.SUPPORTED_CONTENT_TYPES:
encoded_prediction = encoder.encode(prediction, content_type)
if content_type == content_types.CSV:
encoded_prediction = encoded_prediction.encode("utf-8")
return encoded_prediction
raise errors.UnsupportedFormatError(accept)
def default_output_fn(self, prediction, accept):
if accept != "text/csv":
raise Exception("Invalid accept: %s" % accept)
return encoder.encode(prediction, accept)
def output_fn(prediction_output, accept):
print('Serializing the generated output.')
if accept == CONTENT_TYPE_OUTPUT:
return encoder.encode(prediction_output, accept)
raise Exception('Requested unsupported ContentType in Accept: ' + accept)
def test_encode_error():
with pytest.raises(errors.UnsupportedFormatError):
encoder.encode(42, content_types.OCTET_STREAM)
def output_fn(prediction, content_type):
res = [{"probabilities": result["probabilities"], "top_n_grams": result["top_n_grams"]} for result in prediction]
return encoder.encode(res, content_type)
