update LT docs

.github/CHANGELOG.md

@@ -134,6 +134,9 @@
 
 ### Bug fixes
 
+* Fix `qml.state()` support for `lightning.tensor`. The state returned by C++ backend is not normalized anymore to ensure the return state is aligned with `default.qubit`.
+  [(#971)](https://github.com/PennyLaneAI/pennylane-lightning/pull/971)
+
 *  Fix `liblightning_kokkos_catalyst.so` not copied to correct build path for editable installation.
   [(#968)](https://github.com/PennyLaneAI/pennylane-lightning/pull/968)
 

pennylane_lightning/core/src/simulators/lightning_tensor/tncuda/TNCudaBase.hpp

@@ -413,7 +413,7 @@ class TNCudaBase : public TensornetBase<PrecisionT, Derived> {
 
         DataBuffer<CFP_t, int> d_output_tensor(length, getDevTag(), true);
 
-        get_accessor_(d_output_tensor.getData(), length, projected_modes,
+        get_accessor_(d_output_tensor.getData(), projected_modes,
                       projected_mode_values, numHyperSamples);
 
         d_output_tensor.CopyGpuDataToHost(host_data, length);
@@ -423,33 +423,30 @@ class TNCudaBase : public TensornetBase<PrecisionT, Derived> {
      * @brief Get a slice of the full state tensor.
      *
      * @param tensor_data Pointer to the device memory for state tensor data.
-     * @param tensor_data_size Size of the state tensor data.
      * @param projected_modes Projected modes to get the state tensor for.
      * @param projected_mode_values Values of the projected modes.
      * @param numHyperSamples Number of hyper samples to use in the calculation
      * and is set to 1 by default.
      */
     void get_state_tensor(CFP_t *tensor_data,
-                          const std::size_t tensor_data_size,
                           const std::vector<int32_t> &projected_modes,
                           const std::vector<int64_t> &projected_mode_values,
                           const int32_t numHyperSamples = 1) const {
-        get_accessor_(tensor_data, tensor_data_size, projected_modes,
-                      projected_mode_values, numHyperSamples);
+        get_accessor_(tensor_data, projected_modes, projected_mode_values,
+                      numHyperSamples);
     }
 
   private:
     /**
      * @brief Get accessor of a state tensor
      *
      * @param tensor_data Pointer to the device memory for state tensor data.
-     * @param tensor_data_size Size of the tensor data.
      * @param projected_modes Projected modes to get the state tensor for.
      * @param projected_mode_values Values of the projected modes.
      * @param numHyperSamples Number of hyper samples to use in the calculation
      * and is set to 1 by default.
      */
-    void get_accessor_(CFP_t *tensor_data, const std::size_t tensor_data_size,
+    void get_accessor_(CFP_t *tensor_data,
                        const std::vector<int32_t> &projected_modes,
                        const std::vector<int64_t> &projected_mode_values,
                        const int32_t numHyperSamples = 1) const {
@@ -519,14 +516,6 @@ class TNCudaBase : public TensornetBase<PrecisionT, Derived> {
 
         PL_CUDA_IS_SUCCESS(cudaStreamSynchronize(getDevTag().getStreamID()));
 
-        const ComplexT scale_scalar = ComplexT{1.0, 0.0} / stateNorm2;
-
-        CFP_t scale_scalar_cu{scale_scalar.real(), scale_scalar.imag()};
-
-        scaleC_CUDA<CFP_t, CFP_t>(scale_scalar_cu, tensor_data,
-                                  tensor_data_size, getDevTag().getDeviceID(),
-                                  getDevTag().getStreamID(), getCublasCaller());
-
         PL_CUTENSORNET_IS_SUCCESS(
             cutensornetDestroyWorkspaceDescriptor(workDesc));
         PL_CUTENSORNET_IS_SUCCESS(cutensornetDestroyAccessor(accessor));

pennylane_lightning/core/src/simulators/lightning_tensor/tncuda/measurements/MeasurementsTNCuda.hpp

@@ -130,9 +130,8 @@ template <class TensorNetT> class MeasurementsTNCuda {
         std::vector<int64_t> projectedModeValues(projected_modes.size(), 0);
 
         if (projected_modes.size() == 0) {
-            tensor_network_.get_state_tensor(d_output_tensor.getData(),
-                                             d_output_tensor.getLength(), {},
-                                             {}, numHyperSamples);
+            tensor_network_.get_state_tensor(d_output_tensor.getData(), {}, {},
+                                             numHyperSamples);
             getProbs_CUDA(d_output_tensor.getData(), d_output_probs.getData(),
                           length, static_cast<int>(thread_per_block),
                           tensor_network_.getDevTag().getStreamID());
@@ -154,7 +153,7 @@ template <class TensorNetT> class MeasurementsTNCuda {
                 }
 
                 tensor_network_.get_state_tensor(
-                    d_output_tensor.getData(), length, projected_modes,
+                    d_output_tensor.getData(), projected_modes,
                     projectedModeValues, numHyperSamples);
 
                 getProbs_CUDA(d_output_tensor.getData(), tmp_probs.getData(),

tests/test_gates.py

@@ -102,11 +102,6 @@ def test_gate_unitary_correct(op, op_name):
     if wires == 1 and device_name == "lightning.tensor":
         pytest.skip("Skipping single wire device on lightning.tensor.")
 
-    if op_name == "QubitUnitary" and device_name == "lightning.tensor":
-        pytest.skip(
-            "Skipping QubitUnitary on lightning.tensor. It can't be decomposed into 1-wire or 2-wire gates"
-        )
-
     dev = qml.device(device_name, wires=wires)
 
     @qml.qnode(dev)