Integrate offset into grid (#7262)

Eliminate 1 op during training and inference.

models/yolo.py

@@ -64,11 +64,11 @@ class Detect(nn.Module):
 
                 y = x[i].sigmoid()
                 if self.inplace:
-                    y[..., 0:2] = (y[..., 0:2] * 2 - 0.5 + self.grid[i]) * self.stride[i]  # xy
+                    y[..., 0:2] = (y[..., 0:2] * 2 + self.grid[i]) * self.stride[i]  # xy
                     y[..., 2:4] = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i]  # wh
                 else:  # for YOLOv5 on AWS Inferentia https://github.com/ultralytics/yolov5/pull/2953
                     xy, wh, conf = y.split((2, 2, self.nc + 1), 4)  # y.tensor_split((2, 4, 5), 4)  # torch 1.8.0
-                    xy = (xy * 2 + (self.grid[i] - 0.5)) * self.stride[i]  # xy
+                    xy = (xy * 2 + self.grid[i]) * self.stride[i]  # xy
                     wh = (wh * 2) ** 2 * self.anchor_grid[i]  # wh
                     y = torch.cat((xy, wh, conf), 4)
                 z.append(y.view(bs, -1, self.no))
@@ -82,7 +82,7 @@ class Detect(nn.Module):
             yv, xv = torch.meshgrid(torch.arange(ny, device=d), torch.arange(nx, device=d), indexing='ij')
         else:
             yv, xv = torch.meshgrid(torch.arange(ny, device=d), torch.arange(nx, device=d))
-        grid = torch.stack((xv, yv), 2).expand(shape).float()
+        grid = torch.stack((xv, yv), 2).expand(shape).float() - 0.5  # add grid offset, i.e. y = 2.0 * x - 0.5
         anchor_grid = (self.anchors[i] * self.stride[i]).view((1, self.na, 1, 1, 2)).expand(shape).float()
         return grid, anchor_grid