def test_deployment_numpy_autoencoder():
arr = np.random.rand(100, 10, 3).astype(np.float32)
encoded = sidekick.encode.NumpyEncoder().encode_json(arr)
responses.add(responses.POST,
'http://deployment_url',
json={'rows': [{
'numpy_out': encoded
}]})
deployment = Deployment(url='http://deployment_url',
token='deployment_token',
dtypes_in={'numpy_in': 'Numpy (100x10x3)'},
dtypes_out={'numpy_out': 'Numpy (100x10x3)'})
# Single numpy prediction
prediction = deployment.predict(numpy_in=arr)
np.testing.assert_array_equal(prediction['numpy_out'], arr)
# List of numpy predictions
predictions = deployment.predict_many({'numpy_in': arr} for _ in range(10))
for prediction in predictions:
np.testing.assert_array_equal(prediction['numpy_out'], arr)
# Generator of numpy predictions
predictions = deployment.predict_lazy({'numpy_in': arr} for _ in range(10))
for prediction in predictions:
np.testing.assert_array_equal(prediction['numpy_out'], arr)
# Send bad shape
with pytest.raises(ValueError):
deployment.predict(numpy_in=np.random.rand(100, 1, 1))
def test_deployment_numpy_autoencoder():
shape = (100, 10, 3)
features_in = [FeatureSpec('input', 'numeric', shape)]
features_out = [FeatureSpec('output', 'numeric', shape)]
arr = np.random.rand(*shape).astype(np.float32)
encoded = sidekick.encode.NumpyEncoder().encode_json(arr)
responses.add(
responses.POST,
'http://peltarion.com/deployment/forward',
json={'rows': [{
'output': encoded
}]},
)
responses.add(
responses.GET,
'http://peltarion.com/deployment/openapi.json',
json=mock_api_specs(features_in, features_out),
)
deployment = Deployment(
url='http://peltarion.com/deployment/forward',
token='deployment_token',
)
# Single numpy prediction
prediction = deployment.predict(input=arr)
np.testing.assert_array_equal(prediction['output'], arr)
# List of numpy predictions
predictions = deployment.predict_many({'input': arr} for _ in range(10))
for prediction in predictions:
np.testing.assert_array_equal(prediction['output'], arr)
# Generator of numpy predictions
predictions = deployment.predict_lazy({'input': arr} for _ in range(10))
for prediction in predictions:
np.testing.assert_array_equal(prediction['output'], arr)
# Send bad shape
with pytest.raises(ValueError):
deployment.predict(input=np.random.rand(100, 1, 1))
def test_deployment_numeric_multiple_input():
features_in = [
FeatureSpec('input_1', 'numeric', (1, )),
FeatureSpec('input_2', 'numeric', (1, )),
]
features_out = [FeatureSpec('output', 'numeric', (1, ))]
output = 1
responses.add(responses.POST,
'http://peltarion.com/deployment/forward',
json={'rows': [{
'output': output
}]})
responses.add(
responses.GET,
'http://peltarion.com/deployment/openapi.json',
json=mock_api_specs(features_in, features_out),
)
deployment = Deployment(
url='http://peltarion.com/deployment/forward',
token='deployment_token',
)
# Single numeric prediction
predictions = deployment.predict(input_1=1.0, input_2=2)
assert predictions == {'output': output}
inputs = [{'input_1': 1.0, 'input_2': 2} for _ in range(10)]
# List of numeric predictions
predictions = deployment.predict_many(inputs)
for prediction in predictions:
np.testing.assert_array_equal(prediction['output'], output)
# Generator of numeric predictions
predictions = deployment.predict_lazy(inputs)
for prediction in predictions:
np.testing.assert_array_equal(prediction['output'], output)
# Incorrect type
with pytest.raises(TypeError):
deployment.predict(input_1=1.0, input_2='foo')