[unitaryHACK] Strawberry Fields: Gaussian Merge

[federico0112]

Adds hybrid Gaussian Merge compiler that merges gaussian operations into Gaussian Tranform operations in a program with non-Gaussian and Gaussian operations.

Before submitting

Please complete the following checklist when submitting a PR:

• [x] All new features must include a unit test. If you've fixed a bug or added code that should be tested, add a test to the test directory!

• [x] All new functions and code must be clearly commented and documented. If you do make documentation changes, make sure that the docs build and render correctly by running make docs.

• [x] Ensure that the test suite passes, by running make test.

• [x] Ensure that code is properly formatted, by running make format or black -l 100 strawberryfields. You will need to have the Black code format installed: pip install black.

• [x] Add a new entry to the .github/CHANGELOG.md file, summarizing the change, and including a link back to the PR.

• [x] The Strawberry Fields source code conforms to PEP8 standards. We check all of our code against Pylint. To lint modified files, simply pip install pylint, and then run pylint strawberryfields/path/to/file.py.

When all the above are checked, delete everything above the dashed line and fill in the pull request template.

---

Context: https://github.com/XanaduAI/strawberryfields/issues/574

Description of the Change: Adds hybrid Gaussian Merge compiler that merges gaussian operations into Gaussian Tranform operations in a program with non-Gaussian and Gaussian operations.

Benefits: Can potantially reduce calculation overhead by reducing the number of gaussian operations in a circuit

Possible Drawbacks: Circuits with displacement and squeezing gates require precise cutoff values, determining the correct cutoff and displacement/squeezing parameters has to be done manually.

Related GitHub Issues: https://github.com/XanaduAI/strawberryfields/issues/574

[nquesada]

Hi @federico0112 : as a first suggestion, I'd like to ask you to start looking into the CodeFactor issues reported by the first bot.

[nquesada]

@federico0112 : it looks like you are almost there in terms of linting. You only need two three more docstrings and a whitespace issue.

[codecov[bot]]

Codecov Report

Merging #591 (6c2f3a6) into master (07c6f95) will increase coverage by 0.02%. The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #591      +/-   ##
==========================================
+ Coverage   98.42%   98.44%   +0.02%     
==========================================
  Files          75       76       +1     
  Lines        8608     8752     +144     
==========================================
+ Hits         8472     8616     +144     
  Misses        136      136              

Impacted Files	Coverage Δ
strawberryfields/compilers/__init__.py	100.00% <100.00%> (ø)
strawberryfields/compilers/gaussian_merge.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 07c6f95...6c2f3a6. Read the comment docs.

[nquesada]

@federico0112 : can you pass black -l 100 in these two files:

strawberryfields/compilers/__init__.py strawberryfields/compilers/gaussian_merge.py

[nquesada]

@federico0112 : for what I can see you have not added new tests. A trivial tests this compiler should have is that it should pass all the tests the Gaussian compiler passes. Then you should create a new test file in tests/frontend/compilers that tests the new functionality.

[federico0112]

Can this test be the one you gave me? Would that be enough or do I need more test cases

[nquesada]

That is a potential test, but more are needed. For those I suggest you for example try to compile the program showed in the issue [#574]. Keep in mind that you will want to use small parameters for all the gates that generate photons so that you don't need gigantic photon number cutoff. Moreover, you will want to modify tests for gaussian_unitary so that your now compiler also passes. For example this test you should be able to repurpose for your compiler by something like

@pytest.mark.parametrize("depth", [1, 3, 6])
@pytest.mark.parametrize("width", [5, 10, 15])
@pytest.mark.parametrize("compiler", ["gaussian_unitary", "gaussian_merge"])
def test_gaussian_program(depth, width, compiler):
    """Tests that a circuit and its compiled version produce the same Gaussian state"""
    eng = sf.LocalEngine(backend="gaussian")
    eng1 = sf.LocalEngine(backend="gaussian")
    circuit = sf.Program(width)
    with circuit.context as q:
        for _ in range(depth):
            U, s, V, alphas = random_params(width, 2.0 / depth, 1.0)
            ops.Interferometer(U) | q
            for i in range(width):
                ops.Sgate(s[i]) | q[i]
            ops.Interferometer(V) | q
            for i in range(width):
                ops.Dgate(np.abs(alphas[i]), np.angle(alphas[i])) | q[i]
    compiled_circuit = circuit.compile(compiler=compiler)
    cv = eng.run(circuit).state.cov()
    mean = eng.run(circuit).state.means()

    cv1 = eng1.run(compiled_circuit).state.cov()
    mean1 = eng1.run(compiled_circuit).state.means()
    assert np.allclose(cv, cv1)
    assert np.allclose(mean, mean1)

[federico0112]

Ah I see, thats interesting I've never done test coverage in Python. Ill work on getting the test coverage up. Is this required for the bounty? I am going to be busy all tonight and I don't know if I'll be able to finish it by midnight tonight. It's okay if it is required, just curious.

[nquesada]

HI @federico0112 : Great progress so far. There are still a few lines of gaussian_merge that are not being covered by the tests: https://app.codecov.io/gh/XanaduAI/strawberryfields/compare/591/diff#diff-c3RyYXdiZXJyeWZpZWxkcy9jb21waWxlcnMvZ2F1c3NpYW5fbWVyZ2UucHk=

It seems like most of them are related to displacements.

[federico0112]

I made a change to the unit test. It should increase the code coverage

[nquesada]

Hi @federico0112 : there still some part of the compiler not covered by the tests (https://app.codecov.io/gh/XanaduAI/strawberryfields/compare/591/diff#diff-c3RyYXdiZXJyeWZpZWxkcy9jb21waWxlcnMvZ2F1c3NpYW5fbWVyZ2UucHk=). You might need to design some specific tests to cover those edge cases.

[federico0112]

I just realized that I left out a function that was not being used anymore (figured out a better way to achieve the same thing). Thats why it wasn't covered by the unit tests.

[josh146]

Great work so far @federico0112! I just had a look at the coverage (link here: https://app.codecov.io/gh/XanaduAI/strawberryfields/compare/591/tree/strawberryfields/compilers/gaussian_merge.py) and it looks like there are a couple of lines still not being tested:

1. 

2. 

3. 

4. 

Do you think you could modify existing tests where it makes sense to ensure that these parts of the logic are being tested?

[josh146]

Looks like all the checks are now passing, awesome work @federico0112. However, don't forget to also update the .github/CHANGELOG.md! You will need to add three things:

• A small summary of the addition (as well as a short example to illustrate how it is used for interested SF users 🙂 )
• A link back to this PR,
• Your name to the contributors section!

[federico0112]

Would I add it under the development release section?

[nquesada]

@federico0112 : yes

[federico0112]

Finished updating CHANGLOG.md let me know im missing anything

[crazy4pi314]

Hey @federico0112 and the PennyLane team! I wanted to let folks know I am extending the unitaryHACK deadline till the end of this week for PRs that were already in progress to give some time to iterate with maintainers. Let me know if you have questions, or if I can be of help in getting this issue closed!

[josh146]

Thanks @crazy4pi314!

[nquesada]

This is a fantastic addition to Strawberry Fields, thanks for all the hard work @federico0112 .

[nquesada]

Dear @crazy4pi314 --- The fantastic addition by @federico0112 is now merged into SF. It was a lot of fun being part of the unitaryhack this year. Looking forward to the next one!