Update Rigetti demos (#1042)

A few updates to the Rigetti demos:

- Pytorch Noise
- Update Aspen-M-2 to Aspen-M-3 for the hardware version (M-2 is
decommissioned)
- Add the info about connecting to Rigetti with Docker (included in the
other Rigetti demos)
- Ensemble Multi QPU
- Move the information about connecting to the Rigetti device to
_before_ the section where we try to connect to the Rigetti device
   - Change how we cast to torch during prediction phase

---------

Co-authored-by: Mudit Pandey <[email protected]>

demonstrations/ensemble_multi_qpu.py

@@ -46,6 +46,16 @@
 # /interfaces.html>`_, which can be installed from `here
 # <https://pytorch.org/get-started/locally/>`__.
 #
+# .. warning::
+#    Rigetti's QVM and Quil Compiler services must be running for this tutorial to execute. They
+#    can be installed by consulting the `Rigetti documentation
+#    <http://docs.rigetti.com/qcs/>`__ or, for users with Docker, by running:
+#
+#    .. code-block:: bash
+#
+#        docker run -d -p 5555:5555 rigetti/quilc -R -p 5555
+#        docker run -d -p 5000:5000 rigetti/qvm -S -p 5000
+#
 # Load data
 # ---------
 #
@@ -186,15 +196,6 @@ def plot_points(x_train, y_train, x_test, y_test):
 #    swap ``qiskit.aer`` for ``qiskit.ibmq`` and specify their chosen backend (see `here
 #    <https://docs.pennylane.ai/projects/qiskit/en/latest/devices/ibmq.html>`__).
 #
-# .. warning::
-#    Rigetti's QVM and Quil Compiler services must be running for this tutorial to execute. They
-#    can be installed by consulting the `Rigetti documentation
-#    <http://docs.rigetti.com/qcs/>`__ or, for users with Docker, by running:
-#
-#    .. code-block:: bash
-#
-#        docker run -d -p 5555:5555 rigetti/quilc -R -p 5555
-#        docker run -d -p 5000:5000 rigetti/qvm -S -p 5000
 #
 # The circuits for both QPUs are shown in the figure below:
 #
@@ -259,9 +260,8 @@ def decision(softmax):
 
 def predict_point(params, x_point=None, parallel=True):
     if parallel:
-        results = tuple(dask.delayed(q)(params, x=x_point) for q in qnodes)
-        results = dask.compute(*results, scheduler="threads")
-        results = torch.tensor(torch.vstack(results))
+        results = tuple(dask.delayed(q)(params, x=torch.from_numpy(x_point)) for q in qnodes)
+        results = torch.tensor(dask.compute(*results, scheduler="threads"))
     else:
         results = tuple(q(params, x=x_point) for q in qnodes)
         results = torch.tensor(results)

demonstrations/pytorch_noise.py

@@ -25,9 +25,16 @@
 To follow along with this tutorial on your own computer, you will require the
 following dependencies:
 
-* The `Rigetti SDK <https://qcs.rigetti.com/sdk-downloads>`_, which contains the quantum virtual
+* Rigetti's QVM and Quil Compiler services. One option for setting this up is the
+  `Rigetti SDK <https://qcs.rigetti.com/sdk-downloads>`_, which contains the quantum virtual
   machine (QVM) and quilc quantum compiler. Once installed, the QVM and quilc can be
   started by running the commands ``quilc -S`` and ``qvm -S`` in separate terminal windows.
+  Alternatively, for users with Docker, the QVM and Quil Compiler services can be run with commands:
+
+  .. code-block:: bash
+
+      docker run -d -p 5555:5555 rigetti/quilc -R -p 5555
+      docker run -d -p 5000:5000 rigetti/qvm -S -p 5000
 
 * `PennyLane-Rigetti plugin <https://docs.pennylane.ai/projects/rigetti/en/latest/>`_, in order
   to access the QVM as a PennyLane device. This can be installed via pip:
@@ -198,7 +205,7 @@ def cost(phi, theta, step):
 my_prefix = "Your-Folder-Name"  # the name of the folder in the bucket
 s3_folder = (my_bucket, my_prefix)
 
-device_arn = "arn:aws:braket:us-west-1::device/qpu/rigetti/Aspen-M-2"
+device_arn = "arn:aws:braket:us-west-1::device/qpu/rigetti/Aspen-M-3"
 
 qpu = qml.device(
     "braket.aws.qubit",
@@ -208,7 +215,7 @@ def cost(phi, theta, step):
 )
 
 # Note: swap dev to qpu here to use the QPU
-# Warning: check the pricing of Aspen-M-2 on Braket to make
+# Warning: check the pricing of Aspen-M-3 on Braket to make
 # sure you are aware of the costs associated with running the
 # optimization below.
 @qml.qnode(dev, interface="torch")