``grid_anchors`` would be deprecated soon. Please use ``grid

Error message

``grid_anchors`` would be deprecated soon. Please use ``grid_priors``

What it means

Deprecation warning from AnchorGenerator.grid_anchors: the anchors-named API is being replaced by the priors API. grid_anchors(featmap_sizes, device) still works (delegating to single_level_grid_anchors) but grid_priors should be used instead.

Source

Thrown at mmdet/models/task_modules/prior_generators/anchor_generator.py:356

    def grid_anchors(self,
                     featmap_sizes: List[Tuple],
                     device: DeviceType = 'cuda') -> List[Tensor]:
        """Generate grid anchors in multiple feature levels.

        Args:
            featmap_sizes (list[tuple]): List of feature map sizes in
                multiple feature levels.
            device (str | torch.device): Device where the anchors will be
                put on.

        Return:
            list[torch.Tensor]: Anchors in multiple feature levels. \
                The sizes of each tensor should be [N, 4], where \
                N = width * height * num_base_anchors, width and height \
                are the sizes of the corresponding feature level, \
                num_base_anchors is the number of anchors for that level.
        """
        warnings.warn('``grid_anchors`` would be deprecated soon. '
                      'Please use ``grid_priors`` ')

        assert self.num_levels == len(featmap_sizes)
        multi_level_anchors = []
        for i in range(self.num_levels):
            anchors = self.single_level_grid_anchors(
                self.base_anchors[i].to(device),
                featmap_sizes[i],
                self.strides[i],
                device=device)
            multi_level_anchors.append(anchors)
        return multi_level_anchors

    def single_level_grid_anchors(self,
                                  base_anchors: Tensor,
                                  featmap_size: Tuple[int, int],
                                  stride: Tuple[int, int] = (16, 16),
                                  device: DeviceType = 'cuda') -> Tensor:

View on GitHub (pinned to cfd5d3a985)

Solutions

Replace grid_anchors(...) with grid_priors(...) (same signature semantics)
Rename any cached attribute usage (e.g. prior caching) accordingly
Update tests referencing grid_anchors

Example fix

# before
anchors = self.anchor_generator.grid_anchors(featmap_sizes, device=device)
# after
priors = self.prior_generator.grid_priors(featmap_sizes, device=device)

Defensive patterns

Strategy: type-guard

Validate before calling

assert hasattr(gen, 'grid_priors'), 'generator lacks grid_priors; upgrade mmdet'

Type guard

def has_priors_api(gen) -> bool:
    return hasattr(gen, 'grid_priors') and callable(gen.grid_priors)

Prevention

Feature-detect grid_priors before calling
Write heads against the priors API, not anchors

When it happens

Trigger: Calling anchor_generator.grid_anchors(featmap_sizes, device=...); many old unit tests (test_sparse_prior, test_strides, test_retina_anchor, SSD/YOLO anchor tests) and legacy dense heads call it.

Common situations: Custom dense heads or tests written against the pre-rename API; mmdet refactored 'anchors' to generic 'priors' to cover non-anchor detectors.

Related errors

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/786bde9a424d0c50. Report an issue: GitHub.