ml_switcheroo.testing.fuzzer.heuristics¶
Name-Based Heuristics for Input Generation.
This module provides fallback logic for generating inputs when explicit type hints are missing, generally based on argument naming conventions.
Functions¶
|
Guesses the logical type of an argument based on its name. |
|
Generates a value based on the argument name when no type hint is provided. |
Module Contents¶
- ml_switcheroo.testing.fuzzer.heuristics.guess_dtype_by_name(name: str) str¶
Guesses the logical type of an argument based on its name.
- Parameters:
name (str) – The argument name.
- Returns:
‘bool’, ‘int’, or ‘float’.
- Return type:
str
- ml_switcheroo.testing.fuzzer.heuristics.generate_by_heuristic(name: str, base_shape: Tuple[int, Ellipsis], constraints: Dict[str, Any] = None) Any¶
Generates a value based on the argument name when no type hint is provided.
Respects provided constraints if any.
- Parameters:
name (str) – Argument name (e.g. ‘axis’, ‘x’).
base_shape (Tuple) – Default shape for array generation.
constraints (Dict) – Optional constraints (min/max/type).
- Returns:
Generated value.
- Return type:
Any