EvSr"SS\5r"SS\5r"SS\5r"SS\5r"S S \5rg ) zFInterfaces and other classes for providing custom code for prediction.c.\rSrSrSrSr\S5rSrg)ModelaA Model performs predictions on a given list of instances. The input instances are the raw values sent by the user. It is the responsibility of a Model to translate these instances into actual predictions. The input instances and the output use python data types. The input instances have been decoded prior to being passed to the predict method. The output, which should use python data types is encoded after being returned from the predict method. c [5e)aReturns predictions for the provided instances. Instances are the decoded values from the request. Clients need not worry about decoding json nor base64 decoding. Args: instances: A list of instances, as described in the API. **kwargs: Additional keyword arguments, will be passed into the client's predict method. Returns: A list of outputs containing the prediction results. Raises: PredictionError: If an error occurs during prediction. NotImplementedErrorself instanceskwargss 9lib/third_party/ml_sdk/cloud/ml/prediction/_interfaces.pypredict Model.predicts "  c[5e)abCreates a model using the given model path. Path is useful, e.g., to load files from the exported directory containing the model. Args: model_path: The local directory that contains the exported model file along with any additional files uploaded when creating the version resource. Returns: An instance implementing this Model class. rcls model_paths r from_pathModel.from_path1s   rN) __name__ __module__ __qualname____firstlineno____doc__r classmethodr__static_attributes__rrr rrs   &  rrc*\rSrSrSrSrSrSrSrg)PredictionClientCadA client for Prediction. No assumptions are made about whether the prediction happens in process, across processes, or even over the network. The inputs, unlike Model.predict, have already been "columnarized", i.e., a dict mapping input names to values for a whole batch, much like Session.run's feed_dict parameter. The return value is the same format. cg)Nr)r argsr s r __init__PredictionClient.__init__Nsrc [5ezProduces predictions for the given inputs. Args: inputs: A dict mapping input names to values. **kwargs: Additional keyword arguments for prediction Returns: A dict mapping output names to output values, similar to the input dict. rr inputsr s r r PredictionClient.predictQ   rc [5er&rr's r explainPredictionClient.explain^r*rrN) rrrrrr#r r,rrrr rrCs     rrc(\rSrSrSr\S5rSrg) Processorlz/Interface for constructing instance processors.c[5e)zCreates a processor using the given model path. Args: model_path: The path to the stored model. Returns: An instance implementing this Processor class. rrs r from_model_pathProcessor.from_model_patho   rrN)rrrrrrr2rrrr r/r/ls7    rr/c\rSrSrSrSrSrg) Preprocessor|z?Interface for processing a list of instances before prediction.c [5e)zThe preprocessing function. Args: instances: A list of instances, as provided to the predict() method. **kwargs: Additional keyword arguments for preprocessing. Returns: The processed instance to use in the predict() method. rrs r preprocessPreprocessor.preprocessr4rrN)rrrrrr9rrrr r6r6|s G  rr6c\rSrSrSrSrSrg) Postprocessorz>Interface for processing a list of instances after prediction.c [5e)zThe postprocessing function. Args: instances: A list of instances, as provided to the predict() method. **kwargs: Additional keyword arguments for postprocessing. Returns: The processed instance to return as the final prediction output. rrs r postprocessPostprocessor.postprocessr4rrN)rrrrrr?rrrr r<r<s F  rr<N)robjectrrr/r6r<rrr rBsIM/ F/ d& v& R     6   F  r