Merge pull request #1376 from JamesG9802/FixTensorBackwardCall

Fixing Tensor.backward's function signature

Author: NiklasGustafsson

RELEASENOTES.md

@@ -4,6 +4,9 @@ Releases, starting with 9/2/2021, are listed with the most recent release at the
 
 # NuGet Version 0.103.1
 
+__Breaking Changes__:
+#1376 `torch.Tensor.backward`'s function signature has been updated to match PyTorch's implementation. Previously, passing `create_graph` or `retain_graph` by position would work like PyTorch's `torch.Tensor.backward`, but not if passing by name (`create_graph`'s value was swapped with `retain_graph`). This has been corrected; however, this means any code that passes `create_graph` or `retain_graph` by name needs to be updated to reflect the intended functionality.<br/>
+
 __Bug Fixes__:
 
 #1383 `torch.linalg.vector_norm`: Make `ord`-argument optional, as specified in docs<br/>

src/TorchSharp/Tensor/Tensor.cs

@@ -697,8 +697,8 @@ public bool is_sparse {
                 }
             }
 
-            public void backward(IList<Tensor>? grad_tensors = null, bool create_graph = false, bool retain_graph = false, IList<Tensor>? inputs = null) =>
-                torch.autograd.backward(new[] { this }, grad_tensors, create_graph, retain_graph, inputs);
+            public void backward(IList<Tensor>? grad_tensors = null, bool retain_graph = false, bool create_graph = false, IList<Tensor>? inputs = null) =>
+                torch.autograd.backward(new[] { this }, grad_tensors, retain_graph, create_graph, inputs);
 
             /// <summary>
             /// Creates a tensor by loading it from a file.