Add `sparse_encoder` to `models.encodings` #1591

renatomello · 2025-02-17T06:48:21Z

This is based on Section IV of https://arxiv.org/abs/2405.20408

Depends on qiboteam/qibojit#220

Checklist:

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

codecov · 2025-02-20T09:23:53Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.11%. Comparing base (c818eba) to head (2505333).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1591   +/-   ##
=======================================
  Coverage   99.11%   99.11%           
=======================================
  Files          77       77           
  Lines       11475    11571   +96     
=======================================
+ Hits        11373    11469   +96     
  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 · 2025-04-09T10:24:21Z

Ah ok, so that one should be rebased to this one, I guess. I would prefer not to merge something that is failing in main, this will make all the other PR fail.

Can I merge that one into this one and make it just one PR then?

Making functions in `models.encodings` backend-aware

renatomello · 2025-04-10T10:47:10Z

@BrunoLiegiBastonLiegi @AlejandroSopena all tests in models.encodings are passing for GPUs now. I'm not going to fix all the other tests unrelated to that module in this PR to not make it even more bloated than it already is.

renatomello · 2025-04-10T12:14:03Z

@BrunoLiegiBastonLiegi @AlejandroSopena all tests in models.encodings are passing for GPUs now. I'm not going to fix all the other tests unrelated to that module in this PR to not make it even more bloated than it already is.

I lied. It's fixed now.

github-actions · 2025-04-10T12:25:30Z

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/14379519102

src/qibo/models/encodings.py

BrunoLiegiBastonLiegi · 2025-04-10T09:52:57Z

src/qibo/models/encodings.py

+    _data_test = data[0][1]
+    _data_test = (
+        _data_test.dtype if "array" in str(type(_data_test)) else type(_data_test)
+    )


Suggested change

_data_test = data[0][1]

_data_test = (

_data_test.dtype if "array" in str(type(_data_test)) else type(_data_test)

)

_data_test = (

_data_test.dtype if "array" in str(type(data[0][1])) else type(data[0][1])

)

I didn't want to do that to not call data[0][1] three times in the same line

BrunoLiegiBastonLiegi · 2025-04-11T14:29:21Z

src/qibo/models/encodings.py

+    # TODO: Fix this mess with qibo native dtypes
+    try:
+        type_test = bool("int" in str(data[0][0].dtype))
+    except AttributeError:
+        type_test = bool("int" in str(type(data[0][0])))


yeah indeed I think it's time we address this because it's extremely tedious.

Maybe this is a good issue for the unitaryHACK @alecandido @AlejandroSopena

BrunoLiegiBastonLiegi · 2025-04-14T08:06:44Z

src/qibo/models/encodings.py

+    thetas = _generate_rbs_angles(
+        _data_sorted, architecture="diagonal", backend=backend
+    )
+    phis = backend.np.zeros(len(thetas) + 1)


This may be failing with CupyBackend as backend.np is numpy and not cupy. Or better, it may not fail but it most likely won't use cupy arrays, and thus GPU.

This fix, in my opinion, is to be done outside of this PR. I have been thinking of refactoring this part of the backends as something like backend.engine or backend.platform (or even backend.math) and drop backend.np entirely from every backend.

Yeah yeah I agree, just pointing out here, if you want to simply patch it here you could just do a backend.cast even though it's not the cleanest solution...

BrunoLiegiBastonLiegi · 2025-04-14T08:10:01Z

src/qibo/models/encodings.py

    for qubit, bit in enumerate(bitstrings_sorted[0]):
        if bit == 1:
-            touched_qubits.append(qubit)
+            touched_qubits.append(int(qubit))


Do you actually need the explicit cast to int, enumerate should produce an int in any case I believe. In any case, if bitstring_sorted is a int8 np array you could probably just do

touched_qubits = (bitstrings_sorted[0] == 1).nonzero()[0]

Yes, the cast to int needs to be explicit because of the GPU backends.

BrunoLiegiBastonLiegi · 2025-04-14T08:19:52Z

src/qibo/models/encodings.py

+            list(backend.np.argsort(b_0)[-hw_0:]),
+            list(backend.np.argsort(b_1)[-hw_1:]),
+        )
+        ones, new_ones = {int(elem) for elem in ones}, {int(elem) for elem in new_ones}


Can't you just do

Suggested change

ones, new_ones = {int(elem) for elem in ones}, {int(elem) for elem in new_ones}

ones, new_ones = set(ones), set(new_ones)

?

This doesn't work again because of GPU backends. The elements are unhashable without the explicit casting to int.

src/qibo/models/encodings.py

BrunoLiegiBastonLiegi · 2025-04-14T08:33:46Z

src/qibo/models/encodings.py

        if qubit not in touched_qubits:
-            touched_qubits.append(qubit)
+            touched_qubits.append(int(qubit))


Suggested change

touched_qubits.append(int(qubit))

touched_qubits.append(qubit)

qubit should already be an int, as you just casted it

They need to be actual ints instead of cp.int8 otherwise some other errors are raised down the line.

Co-authored-by: BrunoLiegiBastonLiegi <[email protected]>

github-actions · 2025-04-14T11:23:07Z

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/14444165235

renatomello added 3 commits February 14, 2025 12:49

__init__

68a4ebb

function

639b704

bad fix

f3d00cf

renatomello added documentation Improvements or additions to documentation enhancement New feature or request labels Feb 17, 2025

renatomello added this to the Qibo 0.2.17 milestone Feb 17, 2025

renatomello self-assigned this Feb 17, 2025

renatomello linked an issue Feb 17, 2025 that may be closed by this pull request

apply method not working properly for gates.GeneralizedRBS #1592

Closed

renatomello mentioned this pull request Feb 17, 2025

apply method not working properly for gates.GeneralizedRBS #1592

Closed

renatomello added 2 commits February 20, 2025 10:53

fix gate apply

ca8eb71

undo change

dc71b05

renatomello removed a link to an issue Feb 26, 2025

apply method not working properly for gates.GeneralizedRBS #1592

Closed

renatomello mentioned this pull request Feb 26, 2025

Fix gates.GeneralizedRBS.apply method #1594

Merged

4 tasks

renatomello added 2 commits March 5, 2025 08:37

Merge branch 'master' into sparse_encoder

c9f76f9

Merge branch 'master' into sparse_encoder

6f7dd45

scarrazza modified the milestones: Qibo 0.2.17, Qibo 0.2.18 Mar 19, 2025

renatomello added 7 commits March 19, 2025 10:17

Merge branch 'master' into sparse_encoder

eafd205

Merge branch 'master' into sparse_encoder

05e2066

Merge branch 'master' into sparse_encoder

1e204ed

Merge branch 'master' into sparse_encoder

b3e6292

Merge branch 'master' into sparse_encoder

029d590

fix lint

38a1771

fix lint issues

75bb1e1

renatomello requested review from stavros11, AlejandroSopena and BrunoLiegiBastonLiegi March 24, 2025 09:20

renatomello added 2 commits March 26, 2025 10:54

sort by hamming weight

c830bba

complete encoder

a7fe2be

renatomello added 4 commits April 10, 2025 06:02

Merge pull request #1622 from qiboteam/encodings_backend

28c98e0

Making functions in `models.encodings` backend-aware

casting data

acc6ec6

GPU compatibility

aee0b8c

switch qibojit branch

eab8933

renatomello mentioned this pull request Apr 10, 2025

Fix CustomMatrices casting in the GPU backends qiboteam/qibojit#220

Merged

renatomello requested a review from AlejandroSopena April 10, 2025 09:55

remove print

235601d

renatomello added the run-on-sim label Apr 10, 2025

fix cuquantum

b217e67

renatomello added run-on-sim and removed run-on-sim labels Apr 10, 2025

update lock file

db5c529

renatomello added run-on-sim and removed run-on-sim labels Apr 10, 2025

renatomello linked an issue Apr 10, 2025 that may be closed by this pull request

Test failing on GPU for both cupy and cuquantum #1333

Open

BrunoLiegiBastonLiegi reviewed Apr 14, 2025

View reviewed changes

renatomello and others added 5 commits April 14, 2025 09:16

Update src/qibo/models/encodings.py

10fde9b

Co-authored-by: BrunoLiegiBastonLiegi <[email protected]>

move casting outside loop

bc6b711

updated lock file

91d1f31

cast list

dee4c8e

change order

27cf50d

renatomello added run-on-sim and removed run-on-sim labels Apr 14, 2025

coverage

2505333

renatomello requested a review from BrunoLiegiBastonLiegi April 14, 2025 12:01

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add `sparse_encoder` to `models.encodings` #1591

Add `sparse_encoder` to `models.encodings` #1591

renatomello commented Feb 17, 2025 •

edited

Loading

codecov bot commented Feb 20, 2025 •

edited

Loading

renatomello commented Apr 9, 2025

renatomello commented Apr 10, 2025 •

edited

Loading

renatomello commented Apr 10, 2025

github-actions bot commented Apr 10, 2025

BrunoLiegiBastonLiegi Apr 10, 2025

renatomello Apr 14, 2025

BrunoLiegiBastonLiegi Apr 11, 2025

renatomello Apr 14, 2025 •

edited

Loading

BrunoLiegiBastonLiegi Apr 14, 2025

renatomello Apr 14, 2025

BrunoLiegiBastonLiegi Apr 14, 2025

BrunoLiegiBastonLiegi Apr 14, 2025

renatomello Apr 14, 2025 •

edited

Loading

BrunoLiegiBastonLiegi Apr 14, 2025

renatomello Apr 14, 2025 •

edited

Loading

BrunoLiegiBastonLiegi Apr 14, 2025

renatomello Apr 14, 2025

github-actions bot commented Apr 14, 2025

	ones, new_ones = {int(elem) for elem in ones}, {int(elem) for elem in new_ones}
	ones, new_ones = set(ones), set(new_ones)

	touched_qubits.append(int(qubit))
	touched_qubits.append(qubit)

Add sparse_encoder to models.encodings #1591

Are you sure you want to change the base?

Add sparse_encoder to models.encodings #1591

Conversation

renatomello commented Feb 17, 2025 • edited Loading

codecov bot commented Feb 20, 2025 • edited Loading

Codecov Report

renatomello commented Apr 9, 2025

renatomello commented Apr 10, 2025 • edited Loading

renatomello commented Apr 10, 2025

github-actions bot commented Apr 10, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

renatomello Apr 14, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

renatomello Apr 14, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

renatomello Apr 14, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

github-actions bot commented Apr 14, 2025

Add `sparse_encoder` to `models.encodings` #1591

Add `sparse_encoder` to `models.encodings` #1591

renatomello commented Feb 17, 2025 •

edited

Loading

codecov bot commented Feb 20, 2025 •

edited

Loading

renatomello commented Apr 10, 2025 •

edited

Loading

renatomello Apr 14, 2025 •

edited

Loading

renatomello Apr 14, 2025 •

edited

Loading

renatomello Apr 14, 2025 •

edited

Loading