FIT mode now creates platform with no hardware memory. by lballerio · Pull Request #1624 · qiboteam/qibocal

lballerio · 2026-07-21T13:58:03Z

We noticed that when executing only FIT operation (e.g. though qq fit command) Qibocal was building the platform using the state of platform_folder (e.g. the ones in qibolab_platforms_qrc) and not the state saved in the experiment_data_folder.
We fix this issue in the PR, also avoiding to construct the Hardware sensitive fields of the Platform for operations that do not include data acquisition.

Also, Completed object now always requires a path:Path input an is no more optional, as I think it should be.

Note

Discuss about eventual tests to add.

linked to Qibolab #1535.

codecov · 2026-07-21T14:02:31Z

Codecov Report

❌ Patch coverage is 85.18519% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 91.41%. Comparing base (4d5df4d) to head (91ee78e).

Files with missing lines	Patch %	Lines
src/qibocal/auto/output.py	40.00%	3 Missing ⚠️
src/qibocal/auto/task.py	83.33%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1624   +/-   ##
=======================================
  Coverage   91.41%   91.41%           
=======================================
  Files         148      148           
  Lines       11841    11851   +10     
=======================================
+ Hits        10824    10834   +10     
  Misses       1017     1017

Flag	Coverage Δ
unittests	`91.41% <85.18%> (+<0.01%)`	⬆️

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

Files with missing lines	Coverage Δ
src/qibocal/auto/execute.py	`88.88% <100.00%> (+0.09%)`	⬆️
src/qibocal/auto/history.py	`92.53% <100.00%> (-0.22%)`	⬇️
src/qibocal/auto/runcard.py	`97.67% <ø> (ø)`
src/qibocal/calibration/platform.py	`95.55% <100.00%> (+1.26%)`	⬆️
src/qibocal/cli/update.py	`97.14% <100.00%> (ø)`
src/qibocal/auto/task.py	`96.31% <83.33%> (-0.64%)`	⬇️
src/qibocal/auto/output.py	`97.69% <40.00%> (-2.31%)`	⬇️

... and 4 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

RoyStegeman

Some comments. The main one being a bug the line below: while running qq fit, dummy_hardware will be True instead of None.

platform = create_dummy() if dummy_hardware is None else create_platform(platform_name)

RoyStegeman · 2026-07-23T11:43:31Z

+from qibolab._core.dummy.platform import create_dummy
+from qibolab._core.platform.platform import PARAMETERS


This is not part of the public API, meaning qibolab can change this without warning, so we should not use it in qibocal

Ok I'll find a solution.

in order to address this point, I opened Qibolab #1535 and then corrected the import in e47d647.

RoyStegeman · 2026-07-23T11:49:26Z

+        rather than the platform definition. If a ``platform_name`` is provided, the hardware
+        configuration is loaded from that platform; otherwise, a dummy hardware
+        configuration is used so that acquisition-related fields are still present


If a platform_name is provided, the hardware
configuration is loaded from that platform; otherwise, a dummy hardware
is used

According to this the boolean dummy_hardware is not needed, but depends on whether a platform_name is provided. Although platform_name can't be None, so the docstring is not very clear.

addressed in e47d647

RoyStegeman · 2026-07-23T11:51:49Z

+
+        The platform is rebuilt from the configuration saved in the experiment history,
+        using the ``parameters.json`` and ``calibration.json`` files stored in the data folder
+        rather than the platform definition. If a ``platform_name`` is provided, the hardware


I think by "the platform definition" you mean the QIBOLAB_PLATFORMS environment variable. If so, please write that. If you meant something else, please clarify.

addressed in e47d647

RoyStegeman · 2026-07-23T11:55:04Z

+            create_dummy() if dummy_hardware is None else create_platform(platform_name)
+        )
+        platform.parameters = parameters
+        platform.name = platform_name


Why do we need to overwrite the name?

That action is only effective when platform is a DummyPlatform, I've decided to overwrite to propagate the name even for dummy platforms, maybe you need later in the code and I didn't want to create confusion by printing dummy (or whatever name DummyPlatform has).

RoyStegeman · 2026-07-23T12:00:05Z



-def check_overlap_in_input_qubits(targets: np.typing.ArrayLike):
+def check_overlap_in_input_qubits(targets: list):


The first thing this function does is call targ = np.asarray(targets) so the previous type should have been correct.

it's true, but also we call this function after casting self.targets as Target types. which indeed is always a list.
I assumed a more strict typehint since still list is an ArrayLike.

addressed in e47d647

RoyStegeman · 2026-07-23T12:14:54Z

+        )
+
+        platform = (
+            create_dummy() if dummy_hardware is None else create_platform(platform_name)


Suggested change

create_dummy() if dummy_hardware is None else create_platform(platform_name)

create_dummy() if dummy_hardware is True else create_platform(platform_name)

Or maybe if platform_name is None; see comment.

yes, thanks!

addressed in e47d647

RoyStegeman · 2026-07-23T12:18:59Z

+                else CalibrationPlatform.from_datafolder(
+                    folder_path=output,
+                    platform_name=self.platform.name,
+                    dummy_hardware=False,


Suggested change

dummy_hardware=False,

dummy_hardware=True,

I think? Since in this case you're just fitting.

yes thanks for pointing out

addressed in e47d647

RoyStegeman · 2026-07-23T12:21:44Z

-        if self.path is not None:
-            self._data.save(self.path)
+        self._data.save(self.path)

    def dump_results(self):
        """Dumping results."""
-        if self.path is not None:
-            self._results.save(self.path)
+        self._results.save(self.path)


self._data and self._results can still be None, in which case calling None.save will raise an error.

yes you are right, thanks

addressed in e47d647

alecandido · 2026-08-04T15:41:56Z

The proposal is fine by me. Possibly, we just want to eventually exclude platform access from the Protocol.fit invocation, and have qq fit to just work without creating any platform, not even a dummy one. But, for the time being, this is a much faster workaround. And seemingly solid.

Also, Completed object now always requires a path:Path input an is no more optional, as I think it should be.

This is a more complicated situation. The original design was meant to avoid passing paths everywhere, since disk I/O can be quite annoying in a HPC-like infrastructure. Because of the non-locality of the storage, and also the recommended folders access.

However, we may even forget this part now, and decide to always work with paths. This happened accidentally, because the original design was violated (following the frequent quick & dirty approach), but we could even decide to systematically reconsider the initial assumptions.
In principle, if we need performance (no disk I/O) and using a temporary scratch-like folder, we could also deal with at high-level, with an in-memory filesystem, and copying the results folder out of scratch at the desired time - e.g. at the end of the workload, or with intermediate "backups".

Whatever it is done here it is fine and acceptable. But let's keep this in mind, and redesign together.

RoyStegeman · 2026-08-05T06:26:29Z

@@ -25,15 +25,21 @@ def update(path: pathlib.Path, skip_qubits: list[QubitId] | None):
    platform_name = json.loads((path / META).read_text())["platform"]

    platform_path = locate_platform(platform_name)


(qpu165) roy.stegeman@dalma:~/calibration$ qq update /home/users/roy.stegeman/calibration/results/tuna5/260804/17-19-51-cryoscope-0 Traceback (most recent call last): File "/nfs/users/roy.stegeman/.venvs/qpu165/bin/qq", line 8, in <module> sys.exit(command()) File "/nfs/users/roy.stegeman/.venvs/qpu165/lib/python3.10/site-packages/click/core.py", line 1514, in __call__ return self.main(*args, **kwargs) File "/nfs/users/roy.stegeman/.venvs/qpu165/lib/python3.10/site-packages/click/core.py", line 1435, in main rv = self.invoke(ctx) File "/nfs/users/roy.stegeman/.venvs/qpu165/lib/python3.10/site-packages/click/core.py", line 1902, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx)) File "/nfs/users/roy.stegeman/.venvs/qpu165/lib/python3.10/site-packages/click/core.py", line 1298, in invoke return ctx.invoke(self.callback, **ctx.params) File "/nfs/users/roy.stegeman/.venvs/qpu165/lib/python3.10/site-packages/click/core.py", line 853, in invoke return callback(*args, **kwargs) File "/nfs/users/roy.stegeman/github/qibocal/src/qibocal/cli/_base.py", line 145, in update updating(folder, skip_qubits) File "/nfs/users/roy.stegeman/github/qibocal/src/qibocal/cli/update.py", line 27, in update platform_path = locate_platform(platform_name) File "/nfs/users/roy.stegeman/github/qibolab/src/qibolab/_core/platform/load.py", line 81, in locate_platform return _search(name, paths) File "/nfs/users/roy.stegeman/github/qibolab/src/qibolab/_core/platform/load.py", line 56, in _search raise ValueError( ValueError: Platform tuna5 not found. Check $QIBOLAB_PLATFORMS environment variable.

Sine no path is passed to locate_platform it will _search in the default _platforms_paths(), which is still
PLATFORMS_PATH = "QIBOLAB_PLATFORMS"

Copilot

Pull request overview

Reconstructs platforms from saved experiment data for FIT-only workflows, avoiding live hardware state.

Changes:

Adds platform reconstruction from serialized parameters and calibration.
Makes task output paths mandatory.
Updates execution and fitting flows to use stored experiment state.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
`src/qibocal/calibration/platform.py`	Adds data-folder platform reconstruction.
`src/qibocal/auto/execute.py`	Selects live or saved platform by execution mode.
`src/qibocal/auto/output.py`	Uses saved platforms during offline fitting.
`src/qibocal/auto/task.py`	Requires task paths and validates fitting data.
`src/qibocal/auto/history.py`	Makes generated task paths mandatory.
`src/qibocal/auto/runcard.py`	Uses the executor’s configured output path.
`src/qibocal/cli/update.py`	Loads updated platforms from serialized output.
`tests/test_output.py`	Adapts output tests to the executor API.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

+    """This function builds a ``CalibrationPlatform`` object which is sentitive of the hardware,
+    so it needs information about the clusters and its connection. Has to be used for acquisition.
+    """


+            # but also we need the params saved in the datafolder, since in the
+            # platform folder might been changed.


+                self.platform
+                if ExecutionMode.ACQUIRE in mode
+                else CalibrationPlatform.from_datafolder(
+                    folder_path=output,


FIT mode now creates platform with no hardware memory.

91ee78e

lballerio requested a review from RoyStegeman July 21, 2026 13:58

lballerio marked this pull request as ready for review July 22, 2026 09:47

lballerio requested review from a team July 22, 2026 09:47

RoyStegeman reviewed Jul 23, 2026

View reviewed changes

lballerio mentioned this pull request Jul 31, 2026

expose create_dummy() to public API. qiboteam/qibolab#1535

Merged

4 tasks

applying suggestions from Roy Stegeman

e47d647

lballerio requested review from RoyStegeman and scarrazza July 31, 2026 10:00

RoyStegeman requested changes Aug 5, 2026

View reviewed changes

RoyStegeman linked an issue Aug 5, 2026 that may be closed by this pull request

qq fit uses the QIBOLAB_PLATFORM, while it should load the platform from the results folder #1413

Open

alecandido mentioned this pull request Aug 17, 2026

Simplify qubit spectroscopy data #1663

Open

RoyStegeman requested a balanced review from Copilot August 21, 2026 14:36

Copilot started reviewing on behalf of RoyStegeman August 21, 2026 14:37 View session

Copilot AI reviewed Aug 21, 2026

View reviewed changes

		from qibolab._core.dummy.platform import create_dummy
		from qibolab._core.platform.platform import PARAMETERS



		def check_overlap_in_input_qubits(targets: np.typing.ArrayLike):
		def check_overlap_in_input_qubits(targets: list):

	create_dummy() if dummy_hardware is None else create_platform(platform_name)
	create_dummy() if dummy_hardware is True else create_platform(platform_name)

		@@ -25,15 +25,21 @@ def update(path: pathlib.Path, skip_qubits: list[QubitId] \| None):
		platform_name = json.loads((path / META).read_text())["platform"]

		platform_path = locate_platform(platform_name)

		# but also we need the params saved in the datafolder, since in the
		# platform folder might been changed.

Uh oh!

Conversation

lballerio commented Jul 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov Bot commented Jul 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

RoyStegeman left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lballerio Jul 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

alecandido commented Aug 4, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

lballerio commented Jul 21, 2026 •

edited

Loading

codecov Bot commented Jul 21, 2026 •

edited

Loading

lballerio Jul 23, 2026 •

edited

Loading