Add samples and variance support #46
Conversation
Codecov Report
@@ Coverage Diff @@
## master #46 +/- ##
==========================================
- Coverage 96.96% 90.48% -6.48%
==========================================
Files 8 9 +1
Lines 264 347 +83
==========================================
+ Hits 256 314 +58
- Misses 8 33 +25
Continue to review full report at Codecov.
|
| if self.devices is None: | ||
| return | ||
| self.logTestName() | ||
| # def test_basicaer_chop_threshold(self): |
There was a problem hiding this comment.
Is this test supposed to be commented out?
There was a problem hiding this comment.
A bit embarrassing, but I spent a good couple of hours attempting to debug this test, but (a) couldn't work out what it was supposed to do, and (b) couldn't work out why any of my changes broke it. I figured with your test refactor PR happening at the same time, we could replace it and then see what happened 😆
Codecov Report
@@ Coverage Diff @@
## master #46 +/- ##
==========================================
- Coverage 96.96% 90.48% -6.48%
==========================================
Files 8 9 +1
Lines 264 347 +83
==========================================
+ Hits 256 314 +58
- Misses 8 33 +25
Continue to review full report at Codecov.
|
The last PR to get the Qiskit plugin up to feature parity with the latest PennyLane. This PR adds the following:
Support for sampling. If a hardware device, the keyword argument
memory=Trueis passed to the executor to force Qiskit to return raw sampling data. If a state simulator, sampling data is generated from the statevector probabilities by doing a spectral decomposition of the observable.Support for variance. On hardware devices, this is done by simply taking the variance of the samples generated via probability distributions. On state simulators, this is done by doing either (a) doing an inner product directly, or (b) if shots > 0, taking the variance of the samples.
Speed up expectation value calculations on state simulators. Previously, all state simulators were treated as hardware devices. Now, the expectation value is calculated using the same approach as variance: (a) doing an inner product directly, or (b) if shots > 0, taking the mean of the samples.
This PR addresses #42