#yolo

codes/models/arch_util.py

@@ -492,9 +492,14 @@ class RelativeQKBias(nn.Module):
         r = o.unsqueeze(0).repeat(max_positions,1)
         M = ((-(r-c).abs())+l).clamp(0,l)
         self.register_buffer('M', M, persistent=False)
+        self.initted = False
 
     def forward(self, n):
-        return self.emb[self.M[:n, :n]].view(1,n,n)
+        # Ideally, I'd return this:
+        # return self.emb[self.M[:n, :n]].view(1,n,n)
+        # However, indexing operations like this have horrible efficiency on GPUs: https://github.com/pytorch/pytorch/issues/15245
+        # So, enter this horrible, equivalent mess:
+        return torch.gather(self.emb.unsqueeze(-1).repeat(1,n), 0, self.M[:n,:n]).view(1,n,n)
 
 
 class AttentionBlock(nn.Module):