yobx.xoptim.patterns_api

class yobx.xoptim.patterns_api.OnnxEasyPatternOptimization(match_model: ModelProto | FunctionProto, apply_model: ModelProto | FunctionProto, verbose: int = 0)

Implementations pattern matching with onnx models.

Parameters:

• match_model – model expressing the pattern to match

• apply_model – model expression the replacement pattern

yobx.xoptim.patterns_api.make_pattern_from_onnx(name: str, match_model: ModelProto | FunctionProto, apply_model: ModelProto | FunctionProto, verbose: int = 0) → OnnxEasyPatternOptimization

Dynamically create a new class inheriting from EasyPatternOptimization.

Parameters:

• name – class name

• match_model – model expressing the pattern to match

• apply_model – model expression the replacement pattern

• verbose – verbosity

Returns:

instance of a new class

yobx.xoptim.patterns_api.pattern_table_doc(pattern_list: List[PatternOptimization], as_rst: bool = False) → str | List[Dict[str, Any]]

Builds a table for with some information about patterns. See yobx.xoptim.get_pattern_list() for an example.

yobx.xoptim.patterns_api.string_to_elem_type(name: str) → int

Converts a string into an element type. INT64 -> TensorProto.INT64