| ConstantTensorTransformer |
Transforms named references to constant tensors with the rank feature 'constant'.
|
| ExpressionTransforms |
The transformations done on ranking expressions done at config time before passing them on to the Vespa
engine for execution.
|
| FunctionInliner |
Inlines functions in ranking expressions
|
| FunctionShadower |
Transforms function nodes to reference nodes if a rank profile function shadows a built-in function.
|
| LightGBMFeatureConverter |
Replaces instances of the lightgbm(model-path) pseudofeature with the
native Vespa ranking expression implementing the same computation.
|
| OnnxFeatureConverter |
Replaces instances of the onnx(model-path, output)
pseudofeature with the native Vespa ranking expression implementing
the same computation.
|
| OnnxModelTransformer |
Transforms ONNX model features of the forms:
onnxModel(config_name)
onnxModel(config_name).output
onnxModel("path/to/model")
onnxModel("path/to/model").output
onnxModel("path/to/model", "path/to/output")
onnxModel("path/to/model", "unused", "path/to/output") // signature is unused
onnx(...) // same as with onnxModel, onnx is an alias of onnxModel
To the format expected by the backend:
onnxModel(config_name).output
|
| RankProfileTransformContext |
Extends the transform context with rank profile information
|
| TensorFlowFeatureConverter |
Replaces instances of the tensorflow(model-path, signature, output)
pseudofeature with the native Vespa ranking expression implementing
the same computation.
|
| TokenTransformer |
Convenience feature transforms for inputs to Transformer type models.
|
| XgboostFeatureConverter |
Replaces instances of the xgboost(model-path)
pseudofeature with the native Vespa ranking expression implementing
the same computation.
|