Add noise model API for Stim circuit compilation#160
Add noise model API for Stim circuit compilation#160
Conversation
Enhance noise_model.py with detailed NumPy-style docstrings including: - Module-level documentation with usage examples - NoiseKind enum descriptions for each event type - NoiseOp and NoiseEvent parameter documentation with examples - NoiseModel abstract class with comprehensive examples and notes Add Sphinx documentation: - Create docs/source/noise_model.rst for API reference - Add noise_model to module reference in references.rst Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
…odel] - Replace string-based noise API with typed NoiseOp dataclasses (PauliChannel1, PauliChannel2, HeraldedPauliChannel1, HeraldedErase, RawStimOp) - Add typed event classes (PrepareEvent, EntangleEvent, MeasureEvent, IdleEvent) with NodeInfo and Coordinate for position-dependent noise models - Change NoiseModel method return type from Iterable[NoiseOp] to Sequence[NoiseOp] for covariance (allows subclasses to return Sequence[PauliChannel1] etc.) - Update stim_compile to accept Sequence[NoiseModel] instead of single NoiseModel for composing multiple independent noise sources - Add NoisePlacement enum (AUTO, BEFORE, AFTER) for controlling noise insertion - Add comprehensive tests for noise_model module Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Update module docstrings to use the standard "This module provides:" format with a list of public exports. Replace all double backticks with single backticks for inline code references to match the project's documentation style conventions. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Use double backticks for non-Python references (Stim instruction names, parameter names, example values) and proper Sphinx roles (:data:) for module constants. Single backticks remain only for Python object cross- references. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add utility function that creates a probability dict with all 15 Pauli pairs having equal probability p/15. Update the module docstring example to use this utility for proper 2-qubit depolarizing noise instead of the simplified ZZ-only version. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add utility function that creates px, py, pz probabilities each set to p/3 for single-qubit depolarizing channel. Update the module docstring example to use both depolarize1_probs and depolarize2_probs. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replace old parameter-based noise API with the new NoiseModel classes: - Use DepolarizingNoiseModel instead of p_depol_after_clifford - Use MeasurementFlipNoiseModel instead of p_before_meas_flip Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## master #160 +/- ##
==========================================
+ Coverage 83.65% 85.20% +1.54%
==========================================
Files 21 22 +1
Lines 2313 2650 +337
Branches 416 456 +40
==========================================
+ Hits 1935 2258 +323
- Misses 287 293 +6
- Partials 91 99 +8 🚀 New features to boost your workflow:
|
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a571d0fef9
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
graphqomb/stim_compiler.py
Outdated
| if self._noise_models: | ||
| return self._noise_models[0].default_placement(event) |
There was a problem hiding this comment.
The compiler flattens ops from all noise_models, but AUTO placement is resolved with only self._noise_models[0].default_placement(event). This means any later model that overrides default_placement has its AUTO ops inserted on the wrong side of the event (for example, AFTER-intended measurement noise can be emitted BEFORE), producing incorrect circuits whenever multiple models are combined.
Useful? React with 👍 / 👎.
graphqomb/stim_compiler.py
Outdated
| if isinstance(op, MeasurementFlip) and op.target == node: | ||
| meas_flip_p = max(meas_flip_p, op.p) |
There was a problem hiding this comment.
When more than one model emits MeasurementFlip for the same measurement, this logic keeps only the maximum probability and drops the rest. Because stim_compile explicitly supports multiple noise models, this underestimates measurement noise in combined-model runs (e.g., two independent flip sources do not both contribute), so simulation error rates are systematically too low.
Useful? React with 👍 / 👎.
1 participant
Summary
NoiseModelAPI for injecting custom noise into Stim circuits during pattern compilationNoiseOpdataclasses (PauliChannel1,PauliChannel2,HeraldedPauliChannel1,HeraldedErase,RawStimOp,MeasurementFlip)DepolarizingNoiseModelandMeasurementFlipNoiseModelimplementationsdepolarize1_probs()anddepolarize2_probs()utility functionsstim_compile()to acceptnoise_modelsparameter withSequence[NoiseModel]supportTest plan
🤖 Generated with Claude Code