GitLab

Improve Aer noise sim and fix XACC -> IBM noise model conversion

Fixed a bug in the QLM Accelerator plugin when doing noisy sim

The procedure is similar to state-vector but using the diagonal elements of the density matrix rather than computing the square norm of the state vector.
Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

The qubit indexing of IBM json should be relative to the gate operands.

Add a test with multiple qubit noise model to catch this bug.
Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

In order to run the exp-val calc (OBSERVABLE mode) w/ the noisy simulation, I need to set the `sim_method="deterministic-vectorized"` when constructing the Noisy QProc.
Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

Signed-off-by: Alex McCaskey <mccaskeyaj@ornl.gov>

Added noise model doc to XACC

Fixes Apple native build

Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

There are a couple of hard-coded paths which are not portable to native Apple build.

Checking with a plain CI runner on Mojave.
Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

The name is actually AppleClang, not just Clang.
Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

- Wrap the HetMap helper func. for std::any workaround in ifdef.

- Wno-deprecated-copy is not supported by Clang. Cppmicroservices has Werror flag, hence will cause error if it sees an unknown pragma.

- bind2nd is removed in libstdc++, just use the generic bind w/ placeholder.

- libstdc++ doesn't have std::beta in cmath/math.h. Detect this and use boost instead.

- Explicitly specialize complex<dual> multiplication for autodiff. libstdc++ has a check for floating type which fails the dual type.

- Clang doesn't allow lambda capture of structured binding.

Tested by: compiler.
Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

Signed-off-by: Alex McCaskey <mccaskeyaj@ornl.gov>

Signed-off-by: Alex McCaskey <mccaskeyaj@ornl.gov>

Signed-off-by: Alex McCaskey <mccaskeyaj@ornl.gov>

Minor fixes for pulse sim 

Calling operator()() only once before VQE loop over measured circuits

Signed-off-by: Daniel Claudino <6d3@ornl.gov>

Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

(handle the same way as u1, 1 param)

Fixed for new qiskit version and complex array

- Qiskit Aer updated its PulseSystemModel CTor. Need to pass qubit_freq_est to the PulseSimulator defaults instead.

- Changed order of type checking when passing data from Python to recognize complex-valued arrays.
Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

Signed-off-by: Alex McCaskey <mccaskeyaj@ornl.gov>

Potential fix for the qsim build on VirtualBox

Removed deprecated NoiseModel interfaces

Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

Switching to a universal noise channel model.
Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

Support IBM new gateset

They changed the convention of gate-qubit naming: no underscore for single-qubit gates.
Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

Adapt to return the final circuit so that we can see how much it grew

Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

First pass of generic JSON-based noise model implementation 

Currently, IBM mixed two gate sets: some backends (premium ones, I think) still use the old u3-based gate set, while the others (open/main provider) have been switched to the new gate-set.

This is the workaround that I'm using atm to handle this transition period...
Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

Signed-off-by: Daniel Claudino <6d3@ornl.gov>

Also, support retrieving CompositeInstruction from PyHetMap.
Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>

Signed-off-by: Thien Nguyen <nguyentm@ornl.gov>