Upgrade cirq from 0.13.1 to ~= 1.0 and etc.

[jaeyoo]

• Added ./scripts/test_all.sh gpu for testing cuquantum ops together.

• For the recent qsim 0.16.1 compatibility with Cirq, TFQ also needs Cirq >= 1.0. In this case, there are some deprecated & changed features.

  • cirq_google.XMON was deprecated : [cirq-google] Proposal: Remove XmonDevice, Bristlecone, Foxtail and rename optimize_for_xmon. quantumlib/Cirq#4856
  • QuantumEngineSampler was deprecated : Deprecate QuantumEngineSampler quantumlib/Cirq#5371
    • So, we need ProcessorSampler() for testing
  • cirq.CNOT interface was changed.
    • https://quantumai.google/reference/python/cirq/CNOT
    • No more control, target argument.
  • cirq.SingleQubitGate was deprecated.
    • For testing, use cirq.testing.SingleQubitGate : https://github.com/quantumlib/Cirq/pull/5272/files
    • For implementation, use cirq.Gate.

• ParseContext ProtoParse() has bug.

  • 2023-05-02 04:52:39.342483: W tensorflow/core/framework/op_kernel.cc:1830] OP_REQUIRES failed at tfq_adj_grad_op_cuquantum.cu.cc:87 : INVALID_ARGUMENT: Unparseable proto: junk and segfault.
    • It should not segfault the program. For graceful termination, we need to allow OpKernel to run its destructor to return resources including GPU handlers.

• Test_simulate_state_large was too large.

  • 4x4=16 qubits ~ 160 GB.
  • → so, 3x3 = 9 qubits resolves.

• SampledExpectationCuquantum op bug

  • circuit_execution_ops_test:ExecutionOpsConsistentyTest.test_sampled_expectation_no_circuits4 failed
  • Because it has no empty tensor treatment, it fell into segfault.

• Lots of warnings fixed:

  • int vs unsigned, size type comparisons e.g. for (int … → for (size_t …
  • However, if the iterator decreases (e.g. i--), it should use int because it never goes under 0 in size_t, it runs forever.

• Adjoint gradient cuquantum op numerical error bug was fixed

  • There was no BulkSetAmpl() function for cuStatevec state space. so, I temporarily added a cuda kernel.
  • Please remove the temporary hack code after supporting the function in qsim.

• Added 4 Keras major layers (Expectation, SampledExpectation, Sample, State) use_cuquantum tests.

  • Also, if statements dealing with backend and use_cuquantum were unified.
  • To support older APIs, backend == 'noiseless' or backend is None will be the first condition to check use_cuquantum is required or not.
    • It also checks quantum_concurrent option together, to make it False whenever use_cuquantum is True.

• Fixed and verified ./scripts/test_all.sh gpu passes.