[Feat] Transpile.reverse for QuantumCircuit and AbstractBlock #45
Conversation
Add constructors
Digital decomposition, operations and Braket backend
add utils.py file
* Add README. * Remove useless heading.
Co-authored-by: Kaonan Micadei <[email protected]> Co-authored-by: Vytautas Abramavicius <[email protected]>
* state preparation * logging * register * execution helpers * qubit support * serialization * typing * wavefunction overlaps Co-authored-by: Aleksander Wennersteen <[email protected]> Co-authored-by: Mario Dagrada <[email protected]> Co-authored-by: Vincent Elfving <[email protected]> Co-authored-by: Dominik Seitz <[email protected]> Co-authored-by: Joao Moutinho <[email protected]> Co-authored-by: Vytautas Abramavicius <[email protected]> Co-authored-by: Niklas Heim <[email protected]> Co-authored-by: Roland Guichard <[email protected]>
* state preparation * logging * register * execution helpers * qubit support * serialization * typing * wavefunction overlaps Co-authored-by: Aleksander Wennersteen <[email protected]> Co-authored-by: Mario Dagrada <[email protected]> Co-authored-by: Vincent Elfving <[email protected]> Co-authored-by: Dominik Seitz <[email protected]> Co-authored-by: Joao Moutinho <[email protected]> Co-authored-by: Vytautas Abramavicius <[email protected]> Co-authored-by: Niklas Heim <[email protected]> Co-authored-by: Roland Guichard <[email protected]>
Co-authored-by: Aleksander Wennersteen <[email protected]> Co-authored-by: Mario Dagrada <[email protected]> Co-authored-by: Vincent Elfving <[email protected]> Co-authored-by: Gert-Jan Both <[email protected]>
* Update pre-commits. * Gather examples and notebooks tests on schedule. * Remove obsolete files. * Test against Python 3.9 and 3.10.
Co-authored-by: Mario Dagrada <[email protected]> Co-authored-by: Niklas Heim <[email protected]> Co-authored-by: Dominik Seitz <[email protected]> Co-authored-by: Roland Guichard <[email protected]> Co-authored-by: Joao Moutinho <[email protected]> Co-authored-by: Vytautas Abramavicius <[email protected]>
* Docs suggestions. * Fix typo. * Latest changes. * Latest changes. * Trigger docs tests on CI. * More files to docs. * fix circuit drawings * quadance to qadance * More docs improvements. * fix docs build * More suggestions and run test-docs again. * update pulser examples * remove mkdocs-jupyter * update pulser example * More improvements. --------- Co-authored-by: jpmoutinho <[email protected]> Co-authored-by: João P. Moutinho <[email protected]> Co-authored-by: Vytautas Abramavicius <[email protected]>
* update readme * remove admonition * lint --------- Co-authored-by: Roland Guichard <[email protected]>
* Improvements to the Pulser backend documents + change defaults * Shorter names to the workflows * Linting * Improvements to the docs * Added AnalogRZ operation to the enumeration * Go back to idealized device --------- Co-authored-by: Roland Guichard <[email protected]>
|
Let’s add a few tests (including nested chains), then I’m good with this:) |
dominikandreasseitz
changed the title
There was a problem hiding this comment.
Thanks @dominikandreasseitz. I am not sure what you intend to do here. Also, I am a bit worried that your reverse operation might not be allowed if non-commutative ops are in play.
|
|
||
| @singledispatch | ||
| def reverse(x: QuantumCircuit | AbstractBlock) -> QuantumCircuit | AbstractBlock: | ||
| """Reverses a QuantumCircuit or AbstractBlock.""" |
There was a problem hiding this comment.
Not sure what you mean by "reverse" ?
There was a problem hiding this comment.
what this is supposed to accomplish is just reversing a circuit in the same way that you would reverse a list. we need it (at some point) for the adjoint method
There was a problem hiding this comment.
Ah, so basically this is taking the adjoint of a circuit ?
There was a problem hiding this comment.
If so, I'd rename it adjoint or reuse the current dagger methods.
There was a problem hiding this comment.
Precisely, the dagger method should accomplish the same thing.
There was a problem hiding this comment.
no it doesnt exactly, it reverses the circuit, yes, but it also renames all the parameters to "-{parameter_name}" which its not what we want. we need the same parameter names to compute the gradients in the backward pass of the adjoint method
There was a problem hiding this comment.
lets move this to #78
|
@dominikandreasseitz please add a small description to the MR. If there is not corresponding issue, it is better to always add it as a best practice. |
|
moving to #78 |
In order to perform adjoint differentiation, we need to reverse the circuit in the backward pass.