Allowing circuit execution using float32 and float64 dtypes

[renatomello]

Closes #1515 and #1626 . This has implications for qiboml and qibochem since it allows memory reduction for circuits that only contain gates with real-valued matrix representations.

Also fixes some tests on GPU.

Needs qiboteam/qibojit#219

Checklist:

• Reviewers confirm new code works as expected.
• Tests are passing.
• Coverage does not decrease.
• Documentation is updated.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.11%. Comparing base (c818eba) to head (80f7f45).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1625   +/-   ##
=======================================
  Coverage   99.11%   99.11%           
=======================================
  Files          77       77           
  Lines       11475    11486   +11     
=======================================
+ Hits        11373    11384   +11     
  Misses        102      102           

Flag	Coverage Δ
unittests	99.11% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[renatomello]

@scarrazza do we have any intention of supporting data types like complex256, float128, or float16?

https://numpy.org/doc/stable/user/basics.types.html

[scarrazza]

I would say not, but did you try somenof those with the code in this PR?

[github-actions]

Run on QPU sim completed!

You can download the coverage report as an artifact, from the workflow summary page:
https://github.com/qiboteam/qibo/actions/runs/14264634691

[stavros11]

Thanks @renatomello. This looks good to me and it's great to be able to simulate circuits using real dtypes whenever this is possible.

I just noticed the following issue: the following

import qibo
from qibo import Circuit, gates, matrices

c = Circuit(1)
c.add(gates.X(0))

qibo.set_backend("numpy")
qibo.set_dtype("float32")

state = c().state()
print(state.dtype)

gives float32 as expected, however if I switch the setting of backend and precision to

qibo.set_dtype("float32")
qibo.set_backend("numpy")

it goes back to the default complex128. Is this the behavior we want?

[renatomello]

Thanks @renatomello. This looks good to me and it's great to be able to simulate circuits using real dtypes whenever this is possible.

I just noticed the following issue: the following

import qibo
from qibo import Circuit, gates, matrices

c = Circuit(1)
c.add(gates.X(0))

qibo.set_backend("numpy")
qibo.set_dtype("float32")

state = c().state()
print(state.dtype)

gives float32 as expected, however if I switch the setting of backend and precision to

qibo.set_dtype("float32")
qibo.set_backend("numpy")

it goes back to the default complex128. Is this the behavior we want?

I would say so, yes. set_backend completely resets the backend. So, if nothing is specified regarding the data type, then it will default to the default data type.