Package com.yahoo.schema.processing
Class OnnxModelTypeResolver
java.lang.Object
com.yahoo.schema.processing.Processor
com.yahoo.schema.processing.OnnxModelTypeResolver
Processes every "onnx-model" element in the schema. Associates model type
information by retrieving from either the ONNX model file directly or from
preprocessed information in ZK. Adds missing input and output mappings
(assigning default names).
Must be processed before RankingExpressingTypeResolver.
- Author:
- lesters
-
Field Summary
Fields inherited from class com.yahoo.schema.processing.Processor
deployLogger, queryProfiles, rankProfileRegistry, schema
-
Constructor Summary
ConstructorDescriptionOnnxModelTypeResolver
(Schema schema, com.yahoo.config.application.api.DeployLogger deployLogger, RankProfileRegistry rankProfileRegistry, QueryProfiles queryProfiles) -
Method Summary
Modifier and TypeMethodDescriptionvoid
process
(boolean validate, boolean documentsOnly) Processes the input search definition by modifying the input search and its documents, and returns the input search definition.Methods inherited from class com.yahoo.schema.processing.Processor
addField, fail, formatError, info, info, matchingRankSettingsIterator, newProcessException, newProcessException, process, warn, warn
-
Constructor Details
-
OnnxModelTypeResolver
public OnnxModelTypeResolver(Schema schema, com.yahoo.config.application.api.DeployLogger deployLogger, RankProfileRegistry rankProfileRegistry, QueryProfiles queryProfiles)
-
-
Method Details
-
process
public void process(boolean validate, boolean documentsOnly) Description copied from class:Processor
Processes the input search definition by modifying the input search and its documents, and returns the input search definition.- Specified by:
process
in classProcessor
- Parameters:
validate
- true to throw exceptions on validation errors, false to make the best possible effort at completing processing without throwing an exception. If we are not validating, emitting warnings have no effect and can (but must not) be skipped.documentsOnly
- true to skip processing (including validation, regardless of the validate setting) of aspects not relating to document definitions (e.g rank profiles)
-