ryanthompson591 commented on a change in pull request #16917: URL: https://github.com/apache/beam/pull/16917#discussion_r818729200
########## File path: sdks/python/apache_beam/ml/inference/api.py ########## @@ -0,0 +1,84 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +from dataclasses import dataclass +from enum import Enum +from typing import Tuple, TypeVar, Union + +import apache_beam as beam + + +class PyTorchDevice(Enum): + CPU = 1 + GPU = 2 + + +class SklearnSerializationType(Enum): + PICKLE = 1 + JOBLIB = 2 + + +@dataclass +class BaseModelSpec: + model_url: str + + +@dataclass +class PyTorchModelSpec(BaseModelSpec): + device: PyTorchDevice + + +@dataclass +class SklearnModelSpec(BaseModelSpec): + serialization_type: SklearnSerializationType + + +_K = TypeVar('_K') +_INPUT_TYPE = TypeVar('_INPUT_TYPE') +_OUTPUT_TYPE = TypeVar('_OUTPUT_TYPE') + + +@dataclass +class PredictionResult: + example: _INPUT_TYPE + inference: _OUTPUT_TYPE + + [email protected]_fn [email protected]_input_types(Union[_INPUT_TYPE, Tuple[_K, _INPUT_TYPE]]) [email protected]_output_types( + Union[PredictionResult, Tuple[_K, PredictionResult]]) +def RunInference( + examples: beam.pvalue.PCollection, + model: BaseModelSpec) -> beam.pvalue.PCollection: + """Run inference with a model. + + There one type of inference you can perform using this PTransform: + 1. In-process inference from a SavedModel instance. Review comment: I'm not sure this will be true once we start implementing this interface. Ideally one of the earlier things we should do is wrap the TFX implementation which will use a service like Vertex AI. ########## File path: sdks/python/apache_beam/ml/inference/api.py ########## @@ -0,0 +1,84 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +from dataclasses import dataclass +import apache_beam as beam +from typing import Tuple, TypeVar, Union +# TODO: implement RunInferenceImpl +# from apache_beam.ml.inference.base import RunInferenceImpl + + +@dataclass +class BaseModelSpec: + model_url: str + + +@dataclass +class PyTorchModelSpec(BaseModelSpec): Review comment: It's fine, though, in general I prefer shorter names where they are adequately precise. ########## File path: sdks/python/apache_beam/ml/inference/api.py ########## @@ -0,0 +1,84 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# + +from dataclasses import dataclass +from enum import Enum +from typing import Tuple, TypeVar, Union + +import apache_beam as beam + + +class PyTorchDevice(Enum): + CPU = 1 + GPU = 2 + + +class SklearnSerializationType(Enum): + PICKLE = 1 + JOBLIB = 2 + + +@dataclass +class BaseModelSpec: + model_url: str + + +@dataclass +class PyTorchModelSpec(BaseModelSpec): + device: PyTorchDevice + + +@dataclass +class SklearnModelSpec(BaseModelSpec): + serialization_type: SklearnSerializationType + + +_K = TypeVar('_K') +_INPUT_TYPE = TypeVar('_INPUT_TYPE') +_OUTPUT_TYPE = TypeVar('_OUTPUT_TYPE') + + +@dataclass +class PredictionResult: + example: _INPUT_TYPE + inference: _OUTPUT_TYPE + + [email protected]_fn [email protected]_input_types(Union[_INPUT_TYPE, Tuple[_K, _INPUT_TYPE]]) [email protected]_output_types( + Union[PredictionResult, Tuple[_K, PredictionResult]]) +def RunInference( + examples: beam.pvalue.PCollection, + model: BaseModelSpec) -> beam.pvalue.PCollection: + """Run inference with a model. Review comment: Getting this API doc bogged down in the details of what is implemented at any given time will probably have the effect of letting it get out of date and causing confusion. I suggest instead we just clearly define what this does without getting into any details about what is yet supported and what our plans are. Here's a suggestion: A transform that takes a pcollection of examples (or features) to be used on an ML model. It will then output inferences (or predictions) for those examples in a pcollection of PredictionResults, containing the input examples and output inferences. If examples are paired with keys, it will output a tuple (key, PredictionResult) for each (key, example) input. Models for supported frameworks can be loaded via a URI. Supported services can also be used. TODO: link to a help page that shows what is/isn't supported. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
