Qiskit / qiskit-fermions / 34241535841
98%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 110
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 98.2%. Remained the same
34241535841

push

github

web-flow 
Wrap up fermionic trotterization (#333)

* docs(guides): move the flow-set Trotter step count into fermionic space

Both flow-set guides hand-rolled their own Trotter steps because, when they were
written, no fermionic product formula could see the operator's groups. There is
one now, so the step count moves to `FermionicSuzukiTrotter(order=1, reps=N)`,
which interleaves the flow sets inside a single `Evolution` gate. That is what
the guides' local synthesis classes were attempting and could not do, since each
receives one flow set at a time.

**Every pinned number in the 1D convergence table is unchanged, in both
columns.** Only the mechanism behind the flow-set column moves, which is worth
checking against the diff: the two columns now take their step count from
different levels, and they have to. Routing the Lie baseline through the
fermionic side as well would let it inherit the flow-set partition, and since
Trotterizing commuting terms is exact, its fidelities would collapse onto the
flow-set column's and the comparison would measure nothing. So the flow-set
column subdivides in fermionic space and is decomposed (that is how its
synthesis receives one set at a time) while the Lie column stays whole. The note
that used to caution about the undecomposed circuit now explains that asymmetry.

The 1D caution about `reps` is reworked rather than deleted: its A(t/2) A(t/2)
versus A(t/2) B(t/2) contrast is the clearest statement in either guide of why
the count belongs a level up, and that fact survives.

The 2D guide regains the multi-step check that #309 had to delete. It could not
be a real test then, comparing two syntheses that both nested identically, so it
passed while testing nothing. As a new step 10 it becomes a genuine convergence
test against the exact encoded reference built in step 9. It lands after step 9
because it needs that step's projector and stabilizers, and because step 8 asks a
different question: agreement between two syntheses, n... (continued)

11454 of 11664 relevant lines covered (98.2%)

565116.11 hits per line

Jobs
ID Job ID Ran Files Coverage
1 34241535841.1 110
98.2
GitHub Action Run
Source Files on build 34241535841