Add the `state_vector`, `measurement` class and `simulate` method for the LightningGPU with the new device API #892

LuisAlfredoNu · 2024-09-05T20:00:45Z

Before submitting

Please complete the following checklist when submitting a PR:

All new features must include a unit test.
If you've fixed a bug or added code that should be tested, add a test to the
tests directory!
All new functions and code must be clearly commented and documented.
If you do make documentation changes, make sure that the docs build and
render correctly by running make docs.
Ensure that the test suite passes, by running make test.
Add a new entry to the .github/CHANGELOG.md file, summarizing the
change, and including a link back to the PR.
Ensure that code is properly formatted by running make format.

When all the above are checked, delete everything above the dashed
line and fill in the pull request template.

Context:
Migrate LightningGPU to the new device API

Description of the Change:
Create the state_vector, and measurement class for the new device API to achieve the simulate method

Benefits:
Integration of LGPU with the new device API

Possible Drawbacks:

Related GitHub Issues:

Freezzed PR ⚠️ ❄️

To make a smooth integration of LightningGPU with the new device API, we set the branch gpuNewAPI_backend as the base branch target for future developments related to this big task.

The branch gpuNewAPI_backend has the mock of all classes and methods necessary for the new API. Also, several tests were disabled with

if device_name == "lightning.gpu":
    pytest.skip("LGPU new API in WIP.  Skipping.",allow_module_level=True)

However, these tests will unblocked as the implementation progresses.

After all the developments for integrating LightningGPU with the new API have been completed then the PR will be open to merge to master

[sc-70932]

github-actions · 2024-09-05T20:26:16Z

Hello. You may have forgotten to update the changelog!
Please edit .github/CHANGELOG.md with:

A one-to-two sentence description of the change. You may include a small working example for new features.
A link back to this PR.
Your name (or GitHub username) in the contributors section.

codecov · 2024-09-11T14:32:34Z

Codecov Report

Attention: Patch coverage is 79.60000% with 51 lines in your changes missing coverage. Please review.

Project coverage is 93.55%. Comparing base (b375dd3) to head (96728cb).
Report is 2 commits behind head on gpuNewAPI_backend.

Files with missing lines	Patch %	Lines
pennylane_lightning/lightning_gpu/_mpi_handler.py	41.66%	28 Missing ⚠️
pennylane_lightning/lightning_gpu/_state_vector.py	88.88%	12 Missing ⚠️
pennylane_lightning/lightning_gpu/_measurements.py	86.66%	6 Missing ⚠️
pennylane_lightning/lightning_gpu/lightning_gpu.py	82.14%	5 Missing ⚠️

Additional details and impacted files

@@                  Coverage Diff                   @@
##           gpuNewAPI_backend     #892       +/-   ##
======================================================
+ Coverage              74.66%   93.55%   +18.89%     
======================================================
  Files                     27       76       +49     
  Lines                   1954     8801     +6847     
======================================================
+ Hits                    1459     8234     +6775     
- Misses                   495      567       +72

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

LuisAlfredoNu · 2024-09-11T14:34:24Z

pennylane_lightning/core/_measurements_base.py

-        diagonalizing_gates = measurementprocess.diagonalizing_gates()
-        if diagonalizing_gates:
-            self._qubit_state.apply_operations(diagonalizing_gates)
-        results = self._measurement_lightning.probs(measurementprocess.wires.tolist())
-        if diagonalizing_gates:
-            self._qubit_state.apply_operations(
-                [qml.adjoint(g, lazy=False) for g in reversed(diagonalizing_gates)]
-            )
-        return results


LGPU has different behaviour in probs from LQ and LK. LGPU needs to perform an extra conversion of results due to cuQuantum returns as col-major orderings, so performing transpose on data for bit-index shuffle is needed.

pennylane_lightning/core/src/simulators/lightning_gpu/measurements/MeasurementsGPU.hpp

pennylane_lightning/lightning_gpu/_measurements.py

LuisAlfredoNu · 2024-09-11T15:49:53Z

pennylane_lightning/lightning_gpu/_mpi_handler.py

+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""


A new file and class will be used to keep the device implementation similar between LQ, LK, and LGPU. Also, in case of future implementation of MPI resources in other devices.

multiphaseCFD

Thanks @LuisAlfredoNu . A few more questions.

pennylane_lightning/lightning_gpu/_mpi_handler.py

pennylane_lightning/lightning_gpu/_state_vector.py

multiphaseCFD · 2024-09-13T13:06:48Z

pennylane_lightning/lightning_gpu/_state_vector.py

+        num_wires,
+        dtype=np.complex128,
+        device_name="lightning.gpu",
+        mpi_handler=None,


Do we want to pass mpi_handler around?

Yes because I tried to keep the MPI parts inside the LGPU class but at some point I had circular import problems. That is why I created an independent class. Also, It will be necessary to pass mpi_handler to get the functionality of the MPI methods.

pennylane_lightning/lightning_gpu/_state_vector.py

AmintorDusko

Please let us know if these coverage issues are some fluke.

AmintorDusko · 2024-09-13T14:07:24Z

pennylane_lightning/lightning_gpu/_state_vector.py

+            # TODO
+            # state_data = allocate_aligned_array(state.size, np.dtype(self.dtype), True)
+            # state.getState(state_data)
+            # state = state_data


When do we plan to do that?

With luck, after completing the migration of LGPU. However, do you think that this will be necessary to develop? 🤔

pennylane_lightning/lightning_gpu/_state_vector.py

vincentmr

Looks good to me, thanks @LuisAlfredoNu . I left some comments, feel free to address them as you see fit. About the .s in docstrings, we could do a full sync across all modules in another PR.

pennylane_lightning/lightning_gpu/_mpi_handler.py

pennylane_lightning/lightning_gpu/_state_vector.py

pennylane_lightning/lightning_gpu/lightning_gpu.py

pennylane_lightning/lightning_kokkos/_measurements.py

vincentmr · 2024-09-13T14:19:28Z

pennylane_lightning/lightning_kokkos/_measurements.py

+        Returns:
+            Probabilities of the supplied observable or wires
+        """
+        diagonalizing_gates = measurementprocess.diagonalizing_gates()


Seems we're missing a test here.

I do not understand why CodeCov never catch the data from LKokkos 😕

Co-authored-by: Vincent Michaud-Rioux <[email protected]>

LuisAlfredoNu · 2024-09-13T15:36:21Z

@AmintorDusko

Please let us know if these coverage issues are some fluke.

The coverage issues are related to the MPI part which is under development and has not been tested on the current PR. Although, this does not affect the code for a single device. For LK, I do know why CodeCov never catch the coverage of this device. 😕

AmintorDusko

Nothing more to add. Thank you for that!

maliasadi

Thanks @LuisAlfredoNu! Happy to approve the PR after addressing the design-related concerns.

pennylane_lightning/core/_measurements_base.py

pennylane_lightning/lightning_gpu/_measurements.py

pennylane_lightning/lightning_gpu/_mpi_handler.py

pennylane_lightning/lightning_gpu/_state_vector.py

pennylane_lightning/lightning_gpu/lightning_gpu.py

maliasadi · 2024-09-13T17:22:43Z

tests/lightning_qubit/test_measurements_class.py

+    @pytest.mark.skipif(
+        device_name == "lightning.gpu",
+        reason="lightning.gpu does not support out of order prob.",
+    )


@multiphaseCFD Is there any 3rd-party blocker to add the out-of-order prob support?

cuquantum libs does not support out of order probs.

In this case, I'm wondering if we could: order the wires, calculate the probs, and transpose back to the unordered set of wires. @multiphaseCFD, did we try it already? Do you see any blockers for it? I'm just thinking about it, nothing to do now, of course.

Co-authored-by: Ali Asadi <[email protected]>

maliasadi

Thanks @LuisAlfredoNu! Just a few minor suggestions before merging the PR 🥳

pennylane_lightning/core/src/simulators/lightning_gpu/StateVectorCudaManaged.hpp

pennylane_lightning/lightning_gpu/_state_vector.py

pennylane_lightning/core/_measurements_base.py

pennylane_lightning/lightning_kokkos/_measurements.py

pennylane_lightning/lightning_qubit/_measurements.py

Co-authored-by: Ali Asadi <[email protected]>

multiphaseCFD

LGTM! Thanks @LuisAlfredoNu !

multiphaseCFD · 2024-09-16T14:53:29Z

pennylane_lightning/lightning_gpu/_measurements.py

+            num_local_wires = len(probs_results).bit_length() - 1 if len(probs_results) > 0 else 0
+            return probs_results.reshape([2] * num_local_wires).transpose().reshape(-1)
+
+        return probs_results


Is this line missed in the unit tests or a false positive?

Yes, you are right, this line is not covered. I propose to let it and try to catch the coverage in the following PR 😄

multiphaseCFD · 2024-09-16T14:54:47Z

tests/lightning_qubit/test_measurements_class.py

+    @pytest.mark.skipif(
+        device_name == "lightning.gpu",
+        reason="lightning.gpu does not support out of order prob.",
+    )


cuquantum libs does not support out of order probs.

include the new device base class

120bbed

LuisAlfredoNu added 4 commits September 9, 2024 09:19

add measurements

bdf4daa

State vector almos done

b05a6a4

tmp commit

5dac907

Solve prop issue

bfd0771

maliasadi added the urgent Mark a pull request as high priority label Sep 10, 2024

LuisAlfredoNu added 7 commits September 10, 2024 22:37

ready measurenment class for LGPU

a1ff6c6

print helps for measurements

a0cfb1d

Merge gpuNewAPI_simulate

f472471

grammar correction

c7ac82d

cleaning code

6627913

apply format

6290953

delete usuless variables

219262b

LuisAlfredoNu changed the title ~~include the new device base class~~ Add the state_vector, measurement and MCM class for the new device and simulate method Sep 11, 2024

LuisAlfredoNu changed the title ~~Add the state_vector, measurement and MCM class for the new device and simulate method~~ Add the state_vector, measurement class and simulate method for the LightningGPU with the new device API Sep 11, 2024

LuisAlfredoNu marked this pull request as ready for review September 11, 2024 14:28

LuisAlfredoNu commented Sep 11, 2024

View reviewed changes

pennylane_lightning/core/src/simulators/lightning_gpu/measurements/MeasurementsGPU.hpp Show resolved Hide resolved

LuisAlfredoNu commented Sep 11, 2024

View reviewed changes

pennylane_lightning/lightning_gpu/_measurements.py Outdated Show resolved Hide resolved

LuisAlfredoNu added 3 commits September 11, 2024 10:48

delete prints

bca1a74

Revert change in measurenment test

0f8f957

Merge branch 'gpuNewAPI_backend' into gpuNewAPI_simulate

a9ccf62

LuisAlfredoNu requested a review from a team September 11, 2024 15:17

Merge branch 'gpuNewAPI_backend' into gpuNewAPI_simulate

a99b6e8

LuisAlfredoNu marked this pull request as draft September 11, 2024 15:27

add simulate method

0399f18

LuisAlfredoNu marked this pull request as ready for review September 11, 2024 15:45

LuisAlfredoNu commented Sep 11, 2024

View reviewed changes

LuisAlfredoNu requested review from vincentmr and maliasadi September 12, 2024 16:05

multiphaseCFD reviewed Sep 13, 2024

View reviewed changes

AmintorDusko reviewed Sep 13, 2024

View reviewed changes

vincentmr approved these changes Sep 13, 2024

View reviewed changes

LuisAlfredoNu and others added 2 commits September 13, 2024 11:27

Apply suggestions from code review. Vincent's suggestion

5819efc

Co-authored-by: Vincent Michaud-Rioux <[email protected]>

review comments

2dbc7db

LuisAlfredoNu added 2 commits September 13, 2024 15:40

apply format

0630edf

apply format

3fa8409

LuisAlfredoNu requested review from multiphaseCFD and AmintorDusko September 13, 2024 16:18

AmintorDusko approved these changes Sep 13, 2024

View reviewed changes

maliasadi requested changes Sep 13, 2024

View reviewed changes

LuisAlfredoNu and others added 6 commits September 13, 2024 22:12

Ali suggestion 1

af16b8d

add reset

0cb050f

apply_basis_state as abstract in GPU

54afeb5

apply format

ac87663

Apply suggestions from code review Ali suggestion docs

35270fb

Co-authored-by: Ali Asadi <[email protected]>

propagate namming suggestion

f51cbb9

LuisAlfredoNu requested a review from maliasadi September 16, 2024 20:28

maliasadi approved these changes Sep 16, 2024

View reviewed changes

LuisAlfredoNu and others added 3 commits September 16, 2024 17:54

Apply suggestions from code review. Ali's suggestion 3

65e66e9

Co-authored-by: Ali Asadi <[email protected]>

solve errors with kokkos

0472fdd

apply format

96728cb

multiphaseCFD approved these changes Sep 16, 2024

View reviewed changes

LuisAlfredoNu merged commit da75534 into gpuNewAPI_backend Sep 16, 2024
76 of 78 checks passed

LuisAlfredoNu deleted the gpuNewAPI_simulate branch September 16, 2024 22:52

LuisAlfredoNu restored the gpuNewAPI_simulate branch September 16, 2024 22:52

LuisAlfredoNu deleted the gpuNewAPI_simulate branch October 3, 2024 13:12

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add the `state_vector`, `measurement` class and `simulate` method for the LightningGPU with the new device API #892

Add the `state_vector`, `measurement` class and `simulate` method for the LightningGPU with the new device API #892

LuisAlfredoNu commented Sep 5, 2024 •

edited

Loading

github-actions bot commented Sep 5, 2024

codecov bot commented Sep 11, 2024 •

edited

Loading

LuisAlfredoNu Sep 11, 2024

LuisAlfredoNu Sep 11, 2024

multiphaseCFD left a comment

multiphaseCFD Sep 13, 2024

LuisAlfredoNu Sep 13, 2024

AmintorDusko left a comment

AmintorDusko Sep 13, 2024

LuisAlfredoNu Sep 13, 2024

vincentmr left a comment

vincentmr Sep 13, 2024

LuisAlfredoNu Sep 13, 2024

LuisAlfredoNu commented Sep 13, 2024

AmintorDusko left a comment

maliasadi left a comment

maliasadi Sep 13, 2024

multiphaseCFD Sep 16, 2024

AmintorDusko Sep 17, 2024

maliasadi left a comment

multiphaseCFD left a comment

multiphaseCFD Sep 16, 2024

LuisAlfredoNu Sep 16, 2024

multiphaseCFD Sep 16, 2024

Add the state_vector, measurement class and simulate method for the LightningGPU with the new device API #892

Add the state_vector, measurement class and simulate method for the LightningGPU with the new device API #892

Conversation

LuisAlfredoNu commented Sep 5, 2024 • edited Loading

Before submitting

Freezzed PR ⚠️ ❄️

github-actions bot commented Sep 5, 2024

codecov bot commented Sep 11, 2024 • edited Loading

Codecov Report

Choose a reason for hiding this comment

Choose a reason for hiding this comment

multiphaseCFD left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

AmintorDusko left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vincentmr left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

LuisAlfredoNu commented Sep 13, 2024

AmintorDusko left a comment

Choose a reason for hiding this comment

maliasadi left a 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

maliasadi left a comment

Choose a reason for hiding this comment

multiphaseCFD left a 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

Add the `state_vector`, `measurement` class and `simulate` method for the LightningGPU with the new device API #892

Add the `state_vector`, `measurement` class and `simulate` method for the LightningGPU with the new device API #892

LuisAlfredoNu commented Sep 5, 2024 •

edited

Loading

codecov bot commented Sep 11, 2024 •

edited

Loading