Josh Izaac

Nice! So I guess the solution is stick with the existing Takagi decomposition? We could do something similar to what happens in SciPy/Matlab etc. when you ask for the matrix...

(Note that this really only makes sense for measurements that return integer results, such as `MeasureFock()` or `MeasureThreshold()`).

> Might make sense to make both results.samples and results.counts lazy-loading? Definitely would be nice! Especially to reduce network load.

That's weird - I don't remember this happening in the past. In this something we can trace to a specific PR or change in either the codebase or the tests?

> but it appears that if you reduce the NUM_REPEATS parameter in the test significantly (currently it is 50) then this the tests run and pass without ram issues. Oh,...

> Memory leak specific to pytest while it executes test_normalized_conditional_states with the Tensorflow backend only. I agree @zzzeid, my suspicion is that there is a backend initialization fixture which isn't...

Yep, this is on the to-do list. At the moment, I think this is limited by the FockState API, so we'll start looking at expanding the state API to make...

That makes sense! I can't quite recall my original thinking, maybe I chose a numeric parity flag just in case (at some point in the future) OpenFermion decided to implement...

@kevinsung what is the status of this issue? I can make a quick pull request if it is not already waiting for a merge

Hi @charmerDark, having a look at the Qiskit device's supported operations, https://github.com/PennyLaneAI/pennylane-qiskit/blob/57934622ee0531212e032c6dd24271c5fd1b4d6d/pennylane_qiskit/qiskit_device.py#L36 it appears that `SX` is not yet listed. Further, `ex.PhaseGate` and `ex.U3Gate` currently in used by the plugin...