fix pos_ids

modeling_siglip.py

@@ -323,7 +323,7 @@ class SiglipVisionEmbeddings(nn.Module):
             bucket_coords_h = torch.bucketize(fractional_coords_h, boundaries, right=True)
             bucket_coords_w = torch.bucketize(fractional_coords_w, boundaries, right=True)
 
-            pos_ids = (self.num_patches_per_side * bucket_coords_w[:, None] + bucket_coords_h[None, :]).flatten()
+            pos_ids = (bucket_coords_h[:, None] * self.num_patches_per_side + bucket_coords_w).flatten()
             position_ids[batch_idx][p_attn_mask.view(-1).cpu()] = pos_ids
 
         position_ids = position_ids.to(self.position_embedding.weight.device)