fx

Author: kroening

src/nuterm/pytorch_tests.cpp

@@ -258,7 +258,7 @@ void pytorch_test6()
 
   auto my_loss =
     [](const torch::Tensor &curr, const torch::Tensor &next) -> torch::Tensor {
-      return torch::relu(next - curr + 1.0);
+    return torch::relu(next - curr + 1.0);
   };
 
   torch::optim::SGD optimizer(net->parameters(), /*lr=*/0.1);
@@ -318,7 +318,7 @@ void pytorch_test7()
 
   auto my_loss =
     [](const torch::Tensor &curr, const torch::Tensor &next) -> torch::Tensor {
-      return torch::relu(next - curr + 1.0);
+    return torch::relu(next - curr + 1.0);
   };
 
   torch::optim::SGD optimizer(net->parameters(), /*lr=*/0.1);

src/nuterm/training.cpp

@@ -24,8 +24,7 @@ void ranking_function_training(
     assert(batch.size(0) == 2);
 
   auto ranking_function_loss =
-    [](const torch::Tensor &curr, const torch::Tensor &next) -> torch::Tensor
-  {
+    [](const torch::Tensor &curr, const torch::Tensor &next) -> torch::Tensor {
     assert(curr.dim() == 1 && next.dim() == 1);
     // The ranking needs to decrease from 'curr' to 'next'
     // by at least 'delta'. Anything less than that is loss.