[feature] Shots backend #130
Conversation
… Added `GeneralState.update_gates()` method.
There was a problem hiding this comment.
I've moved get_operator_expectation_value and get_circuit_overlap here because they did not match with the intended use of pytket Backend class: there's an intended way to support expectation values in there, but through a BackendResult, not this.
Additionally, these functions are based on the tensor_network_convert approach that I intend to eventually deprecate once GeneralState is full-featured, so it makes sense they are in the same file that will eventually be moved to a deprecated submodule.
|
|
||
| # TODO: this should be optionally parallelised with MPI | ||
| # (both wrt Pauli strings and contraction itself). | ||
| def get_operator_expectation_value( |
There was a problem hiding this comment.
Unfortunately, this git diff completely butchers what actually happened with get_operator_expectation_value: it was moved (verbatim, except for removals of self argument) to the tensor_network_convert.py file. See https://github.com/CQCL/pytket-cutensornet/pull/130/files#r1653061092 for details on why.
|
|
||
| def get_circuit_overlap( |
There was a problem hiding this comment.
Unfortunately, this git diff completely butchers what actually happened with get_circuit_overlap: it was moved (verbatim, except for removals of self argument) to the tensor_network_convert.py file. See https://github.com/CQCL/pytket-cutensornet/pull/130/files#r1653061092 for details on why.
There was a problem hiding this comment.
The diff in this file is unfortunately quite confusing. The changes where quite simple, but they are somehow intertwinned in the diff, so I suggest to compare the files to each other separately. Here's a brief of the changes:
- Changed the name of
CuTensorNetBackendtoCuTensorNetStateBackend, since it is meant to support state vector output only. - Added a new
CuTensorNetShotsBackendthat satisfies the requirements of Support multi-shot exact simulation backends usingSamplerfrom cuTensorNet #121. - Created a
_CuTensorNetBaseBackendthat the other two backends are based on, where all of the common code lives. - Some chore tasks: provided details of
BackendInfo, removed the gateset requirements and replaced them with a custom predicate that checks for commands with.get_unitary()implemented, removed the use ofauto_rebase_passsince it is no longer needed. - Moved some functions to another file, see https://github.com/CQCL/pytket-cutensornet/pull/130/files#r1653061092.
The approach of using a base class _CuTensorNetBaseBackend is inspired on AerBackend from pytket-qiskit, same with the creation of specialised backends for statevector output and shots output.
I've been using a tutorial for backend creation. It's a bit out of date in places, but mostly fine.
tests/test_cutensornet_backend.py
Outdated
There was a problem hiding this comment.
The tests added for CuTensorNetShotsBackend are very basic because these just need to check that the backend wrapper works. More involved tests for the shot extraction logic were already added in #129 in test_general_state.py.
Description
CuTensorNetBackendintoCuTensorNetStateBackendwhich now uses the newGeneralState.CuTensorNetShotsBackendfor shot-based simulation.Related issues
Completes #121.
Checklist