|
|
| from __future__ import annotations |
| from numbers import Integral |
| import torch |
|
|
| def _validate_refinement_steps(value: object, *, source: str) -> int: |
| if isinstance(value, bool) or not isinstance(value, Integral): |
| raise TypeError(f'{source} must be a positive integer') |
| value = int(value) |
| if value < 1: |
| raise ValueError(f'{source} must be a positive integer') |
| return value |
|
|
| def refinement_mask_fraction(ages: torch.Tensor, *, steps: int) -> torch.Tensor: |
| steps = _validate_refinement_steps(steps, source='steps') |
| if ages.dtype == torch.bool or ages.is_floating_point(): |
| raise TypeError('ages must use an integer dtype') |
| effective_age = ages.clamp(min=0, max=steps).to(dtype=torch.float32) |
| return (1.0 - effective_age / steps).mean(dim=-1) |
|
|
