def test_pytorch_lightning_model_artifact():
svc = PytorchLightningService()
model = TorchLightningModel()
svc.pack('model', model)
with export_service_bundle(svc) as saved_path:
svc = bentoml.load(saved_path)
result = svc.predict(pd.DataFrame([[5, 4, 3, 2]]))
assert result.tolist() == [[6, 5, 4, 3]]
def test_keras_artifact_loaded(svc):
with export_service_bundle(svc) as saved_path:
loaded = bentoml.load(saved_path)
assert (
loaded.predict([test_data]) == 15.0
), 'Inference on saved and loaded Keras artifact does not match expected'
assert (
loaded.predict2([test_data]) == 15.0
), 'Inference on saved and loaded Keras artifact does not match expected'
def test_tensorflow_2_artifact_loaded(svc):
with export_service_bundle(svc) as saved_path:
svc_loaded = bentoml.load(saved_path)
assert (
svc_loaded.predict1(test_tensor) == 15.0
), 'Inference on saved and loaded TF2 artifact does not match expected'
assert (
svc_loaded.predict2(test_tensor) == 15.0
), 'Inference on saved and loaded TF2 artifact does not match expected'
assert (
(svc_loaded.predict3(ragged_data) == 15.0).numpy().all()
), 'Inference on saved and loaded TF2 artifact does not match expected'
def image(svc, clean_context):
with export_service_bundle(svc) as saved_path:
yield clean_context.enter_context(build_api_server_docker_image(saved_path))
def test_svc_bundle(clean_context, test_svc):
return clean_context.enter_context(export_service_bundle(test_svc))
def image(test_svc, clean_context):
with export_service_bundle(test_svc) as bundle_dir:
yield clean_context.enter_context(
build_api_server_docker_image(bundle_dir, "example_service"))
