Update GST to take in gates.Unitary

[mho291]

Checklist:

• Reviewers confirm new code works as expected.
• Tests are passing.
• Coverage does not decrease.
• Documentation is updated.

[renatomello]

Suggested change

	if isinstance(params, list):
	init_args = signature(gate).parameters
	valid_angles = [arg for arg in init_args if arg in angles]
	angle_values = dict(zip(valid_angles, params))
	else:
	init_args = signature(gate).parameters
	valid_angles = [arg for arg in init_args if arg in matrix]
	angle_values = dict(zip(valid_angles, [params]))
	if not isinstance(params, list):
	params = [params]
	init_args = signature(gate).parameters
	valid_angles = [arg for arg in init_args if arg in angles]
	angle_values = dict(zip(valid_angles, params))

[mho291]

Thanks @renatomello for spotting this. I have made the changes. But still trying to get it to work. It's currently complaining

FAILED tests/test_tomography_gate_set_tomography.py::test_GST[numpy-False-target_gates0] - ValueError: axes don't match array
FAILED tests/test_tomography_gate_set_tomography.py::test_GST[numpy-True-target_gates0] - ValueError: axes don't match array
FAILED tests/test_tomography_gate_set_tomography.py::test_GST[qibojit-numba-False-target_gates0] - ValueError: axes don't match array
FAILED tests/test_tomography_gate_set_tomography.py::test_GST[qibojit-numba-True-target_gates0] - ValueError: axes don't match array

Will keep trying to debug.

[renatomello]

This docstring is generating a warning when the documentation is compiled because of identation. I fixed it in #1584. Please look at it to also fix it here.

[mho291]

Hi @MatteoRobbiati, I need some advice with this. When I run the pytests locally, they pass. But here, they always run into dimension error:

FAILED tests/test_tomography_gate_set_tomography.py::test_GST[qiboml-pytorch-False-target_gates0] - RuntimeError: permute(sparse_coo): number of dimensions in the tensor input does not match the length of the desired ordering of dimensions i.e. input.dim() = 1 is not equal to len(dims) = 2

Could there be an issue with qiboml-pytorch?

Thanks so much!

[MatteoRobbiati]

Could there be an issue with qiboml-pytorch?

Ehi @mho291! Thanks for asking :)
I tested locally your branch and in my case I get the same failure, but on all the backends (numpy, qibojit, pytorch, tensorflow), so ig the source of the problem is elsewhere.
It seems to me the problem is appearing when calling this line in gates.py.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.11%. Comparing base (3a13d61) to head (091519a).
Report is 11 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1587   +/-   ##
=======================================
  Coverage   99.11%   99.11%           
=======================================
  Files          77       77           
  Lines       11480    11518   +38     
=======================================
+ Hits        11378    11416   +38     
  Misses        102      102           

Flag	Coverage Δ
unittests	99.11% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[mho291]

Could there be an issue with qiboml-pytorch?

Ehi @mho291! Thanks for asking :) I tested locally your branch and in my case I get the same failure, but on all the backends (numpy, qibojit, pytorch, tensorflow), so ig the source of the problem is elsewhere. It seems to me the problem is appearing when calling this line in gates.py.

Hi @MatteoRobbiati ! Sorry for the radio silence for the past two weeks as I was away. I revisited this PR yesterday and the tests no longer fail on GitHub (but not my local machine), so I tidied it up anyway, and it is ready for review now. Greatly appreciate your help!

[mho291]

I have added an option for ancilla=True or ancilla=False. When ancilla=True, it will simulate a resetting to |0> state by disabling the line circ.add(_prepare_state(k, nqubits)). This is crucial to do GST of certain basis operations in probabilistic error cancellation that require resetting the quantum state to |0> to re-prepare state |0> , |+>, or |y+> (reference paper: Appendix D, Table II).

The ancilla flag, together with the ability to input gates.Unitary gates in GST, will most likely be the final version of GST that can be used for probabilistic error cancellation, which I am working on locally (single qubit & two qubit PEC), and will do a PR soon.

[mho291]

Sorry, please hold the review, I will need to modify the current changes to increase the functionality!

[mho291]

Sorry, please hold the review, I will need to modify the current changes to increase the functionality!

The code is now ready for review! :)