From 364acbfb7b229e8f2185573997f4426957d88517 Mon Sep 17 00:00:00 2001
From: simone bordoni <simone.bordoni@uniroma1.it>
Date: Mon, 16 Sep 2024 16:55:07 +0400
Subject: [PATCH] merge master

---
 doc/source/code-examples/test.py | 38 ++++++++++++++++++++++++++++++++
 src/qibo/backends/pytorch.py     |  6 +++--
 2 files changed, 42 insertions(+), 2 deletions(-)
 create mode 100644 doc/source/code-examples/test.py

diff --git a/doc/source/code-examples/test.py b/doc/source/code-examples/test.py
new file mode 100644
index 0000000000..090ae0d9fd
--- /dev/null
+++ b/doc/source/code-examples/test.py
@@ -0,0 +1,38 @@
+import qibo
+
+qibo.set_backend("pytorch")
+import torch
+
+from qibo import gates, models
+
+torch.set_anomaly_enabled(True)
+
+# Optimization parameters
+nepochs = 1
+optimizer = torch.optim.Adam
+target_state = torch.ones(4, dtype=torch.complex128) / 2.0
+
+# Define circuit ansatz
+params = torch.rand(2, dtype=torch.float64, requires_grad=True)
+print(params)
+optimizer = optimizer([params])
+c = models.Circuit(2)
+c.add(gates.RX(0, params[0]))
+c.add(gates.RY(1, params[1]))
+gate = gates.RY(0, params[1])
+
+print("Gate", gate.matrix())
+print(torch.norm(gate.matrix()).grad)
+
+# for _ in range(nepochs):
+#     optimizer.zero_grad()
+#     c.set_parameters(params)
+#     final_state = c().state()
+#     print("state", final_state)
+#     fidelity = torch.abs(torch.sum(torch.conj(target_state) * final_state))
+#     loss = 1 - fidelity
+#     loss.backward()
+#     optimizer.step()
+#     print("state", final_state)
+#     print("params", params)
+#     print("loss", loss.grad)
diff --git a/src/qibo/backends/pytorch.py b/src/qibo/backends/pytorch.py
index 392aeed405..e57977a1ee 100644
--- a/src/qibo/backends/pytorch.py
+++ b/src/qibo/backends/pytorch.py
@@ -1,5 +1,7 @@
 """PyTorch backend."""
 
+from typing import Optional
+
 import numpy as np
 
 from qibo import __version__
@@ -85,7 +87,7 @@ def cast(
         x,
         dtype=None,
         copy: bool = False,
-        requires_grad: bool = None,
+        requires_grad: Optional[bool] = None,
     ):
         """Casts input as a Torch tensor of the specified dtype.
 
@@ -117,7 +119,6 @@ def cast(
         # check if dtype is an integer to remove gradients
         if dtype in [self.np.int32, self.np.int64, self.np.int8, self.np.int16]:
             requires_grad = False
-
         if isinstance(x, self.np.Tensor):
             x = x.to(dtype)
         elif isinstance(x, list) and all(isinstance(row, self.np.Tensor) for row in x):
@@ -128,6 +129,7 @@ def cast(
         if copy:
             return x.clone()
 
+        print("Casting", x)
         return x
 
     def is_sparse(self, x):