rose-rong-liu commented on a change in pull request #11075: URL: https://github.com/apache/beam/pull/11075#discussion_r426830154
########## File path: website/www/site/content/en/documentation/patterns/ai-platform.md ########## @@ -0,0 +1,79 @@ +--- +title: "AI Platform integration patterns" +--- +<!-- +Licensed 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. +--> + +# AI Platform integration patterns + +This page describes common patterns in pipelines with Google Cloud AI Platform transforms. + +{{< language-switcher java py >}} + +## Getting predictions + +This section shows how to use [Google Cloud AI Platform Prediction](https://cloud.google.com/ai-platform/prediction/docs/overview) to make predictions about new data from a cloud-hosted machine learning model. + +[tfx_bsl](https://github.com/tensorflow/tfx-bsl) is a library with a Beam PTransform called `RunInference`. `RunInference` is able to perform an inference that can use an external service endpoint for receiving data. When using a service endpoint, the transform takes a PCollection of type `tf.train.Example` and, for every batch of elements, sends a request to AI Platform Prediction. The size of a batch may vary. For more details on how Beam finds the best batch size, refer to a docstring for [BatchElements](https://beam.apache.org/releases/pydoc/current/apache_beam.transforms.util.html?highlight=batchelements#apache_beam.transforms.util.BatchElements). + + The transform produces a PCollection of type `PredictionLog`, which contains predictions. + +Before getting started, deploy a TensorFlow model to AI Platform Prediction. The cloud service manages the infrastructure needed to handle prediction requests in both efficient and scalable way. Do note that only TensorFlow models are supported by the transform. For more information, see [Exporting a SavedModel for prediction](https://cloud.google.com/ai-platform/prediction/docs/exporting-savedmodel-for-prediction). Review comment: From the implementation side, it does not limit to tensorflow model, right? As cloud servers other model formats with the same predict API. Or is this for branding purpose? ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
