[BUG] Allow single_level_grid_priors to use cpu device.

Note: there is already an outstanding PR to resolve this issue: open-mmlab#2715

mmdeploy/codebase/mmdet/models/dense_heads/rtmdet_ins_head.py

@@ -163,7 +163,7 @@ def _mask_predict_by_feat_single(self, mask_feat, kernels, priors):
     batch_size = priors.shape[0]
     hw = mask_feat.size()[-2:]
     coord = self.prior_generator.single_level_grid_priors(
-        hw, level_idx=0).to(mask_feat.device)
+        hw, level_idx=0, device=mask_feat.device)
     coord = coord.unsqueeze(0).unsqueeze(0).repeat(batch_size, 1, 1, 1)
     priors = priors.unsqueeze(2)
     points = priors[..., :2]