qliff
A Clifford stabilizer simulator with support for noisy and mid-circuit measurement-based simulation.
- Clifford stab simulation via the Aaronson–Gottesman tableau
- Noisy sim via stabilizer-channel
decomposition
E = Σ_μ q_μ S_μand stratified importance sampling (arXiv:2512.07304), "nearly as cheap as Pauli noise." - Decoder-ready QEC primitives: detectors, observables, detection-event sampling, a detector error model by Pauli-frame propagation, and exporters (parity-check matrix + priors, matching weights, syndrome/label tensors). These drop directly into MWPM (pymatching), BP, or ML decoders.
- Code factories for surface, toric, color/triangular/kagome, subsystem
(Bacon-Shor, gauge), and qLDPC (hypergraph-product, bivariate-bicycle) families, plus
from_stabilisers/from_gaugesfor arbitrary Pauli codes.
Install
pip install qliff
Docs at plutoniumm.github.io/qliff.
Clifford simulation
from qliff import Simulator
s = Simulator(2).H(0).CX(0, 1)
s.canon() # ['+XX', '+ZZ']
s.peek("ZZ") # +1
s.measure("XX") # (+1, False)
m0, m1 = Simulator(2).H(0).CX(0, 1).M(0, 1) # m0 == m1
Mid-circuit measurement and classical feedback are just Python — the simulator is stateful, so conditionals (teleportation, syndrome correction) need no special API:
s = Simulator(3, seed=0)
s.H(0)
s.H(1).CX(1, 2)
s.CX(0, 1).H(0)
if s.M(1) == 1:
s.X(2)
if s.M(0) == 1:
s.Z(2)
s.peek("__X") # +1
Noise
Build Circuit with gate/noise methods, then sample or estimate.
Circuit.estimate picks the right sampler by default — plain Monte-Carlo when every channel is Pauli, otherwise
stratified importance sampling
from qliff import Circuit
c = Circuit(1)
c.H(0).DEPOLARIZE1(0, 0.1).M(0)
c.sample(1000)
c = Circuit(1)
c.H(0).RZ(0, 0.3)
c.estimate("X", 20000) # ≈ cos(0.3)
c = Circuit(1)
c.X(0).AMPLITUDE_DAMP(0, 0.3)
c.estimate("Z", 60000) # ≈ 2p - 1
Force the variance strategy with c.estimate(obs, shots, stratify=False) (flat) or stratify=True
(stratified), or drive the sampler directly: from qliff.noise import Sampler, then
Sampler(c).expect(obs, shots, stratify=True). Add a custom channel by subclassing
qliff.noise.Channel and dropping it in with c.noise(ch, q).
For deep-tail logical error rates where plain Monte-Carlo needs prohibitively many shots,
from qliff.noise import SplittingEstimator runs a multilevel-splitting (subset-sampling)
estimator that reaches the far tail with orders of magnitude fewer decodes.
Quantum error correction
qliff.qec ships code-circuit generators, so you can go straight to a logical-error-rate
curve. Alongside rotated_surface_code there are unrotated_surface_code, toric_code,
repetition_code, the color family (color_code, hex_color_code, triangular_code,
kagome_code), subsystem codes (bacon_shor_code, from_gauges), and qLDPC codes
(hypergraph_product_code, bivariate_bicycle_code); from_stabilisers builds a full
memory circuit from an arbitrary Pauli stabiliser set. Every family whose code has an
X/Z dual takes memory="Z" or memory="X" to run either basis. Any circuit's detectors
and observables are declared with c.detector(...) / c.observable(...), then turned
into decoder inputs:
from qliff.qec import rotated_surface_code, logical_fidelity
from pymatching import Matching
c = rotated_surface_code(5, 5, 0.01) # distance, rounds, p (square d x d patch)
dem = c.dem()
H, priors, obs_matrix = dem.check_matrix()
m = Matching.from_check_matrix(
H,
weights=dem.weights(),
faults_matrix=obs_matrix
)
dets, flips = c.detector_sampler().sample(20000)
fidelity = logical_fidelity(m.decode_batch(dets), flips)
A memory is blind to noise that commutes with its own basis -- a Z-memory cannot see
Z_ERROR or RZ, and reports a logical error rate of exactly zero at any p. qliff
warns when you build one, and again if the detector error model comes out empty; run the
dual basis instead:
c = rotated_surface_code(3, 3, 0.01, channel="Z_ERROR", memory="X") # sees the noise
Extending
Everything you'd customize lives in Python. Add a noise channel by subclassing
qliff.noise.Channel and returning its stabilizer-channel branches; plug in a custom
sampler or observable; export the DEM to whatever decoder you like. The Rust core stays a thin,
fast tableau engine.
Develop
./do develop # build rust core
./do test
./do lint
./do bench
./do build # all wheels (macOS/Linux/Windows, x86_64 + arm) + sdist, no upload
./do deploy # publish what `build` made to PyPI (no rebuild)
cd docs && npm run build && npm run deploy # docs site -> GitHub Pages
License
MIT
If you are a company using this, please get a grad student to help you with issues. If you are a grad student, please feel free to email me :)
Metadata
Release files for qliff 0.5.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| qliff-0.5.0.tar.gz | 76.2 kB | Without provenance | Details for qliff-0.5.0.tar.gz |
Built distributions (wheels)
| File | Actions Reset | |||
|---|---|---|---|---|
| qliff-0.5.0-cp311-abi3-win_arm64.whl | CPython 3.11 | abi3 | Windows ARM64 | Details for qliff-0.5.0-cp311-abi3-win_arm64.whl |
| qliff-0.5.0-cp311-abi3-win_amd64.whl | CPython 3.11 | abi3 | Windows x86-64 | Details for qliff-0.5.0-cp311-abi3-win_amd64.whl |
| qliff-0.5.0-cp311-abi3-manylinux_2_28_x86_64.whl | CPython 3.11 | abi3 | Linux glibc 2.28+ x86-64 | Details for qliff-0.5.0-cp311-abi3-manylinux_2_28_x86_64.whl |
| qliff-0.5.0-cp311-abi3-manylinux_2_28_aarch64.whl | CPython 3.11 | abi3 | Linux glibc 2.28+ ARM64 | Details for qliff-0.5.0-cp311-abi3-manylinux_2_28_aarch64.whl |
| qliff-0.5.0-cp311-abi3-macosx_11_0_x86_64.whl | CPython 3.11 | abi3 | macOS 11.0+ x86-64 | Details for qliff-0.5.0-cp311-abi3-macosx_11_0_x86_64.whl |
| qliff-0.5.0-cp311-abi3-macosx_11_0_arm64.whl | CPython 3.11 | abi3 | macOS 11.0+ ARM64 | Details for qliff-0.5.0-cp311-abi3-macosx_11_0_arm64.whl |
Total release size: 2.4 MB
Release files / qliff-0.5.0.tar.gz
| Download URL | qliff-0.5.0.tar.gz |
|---|---|
| Size | 76.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
f2e6251d9d923da7f5c261b35f12e08a8d8aed2cb66fb9ab806fa4f8b7a5d981
|
|
BLAKE2b-256 checksum How to use checksums |
65f4f9b32833bde1b58b5a5c4f2ba0da52e60c1468de491b36af92cd6294f731
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.3
|
Release files / qliff-0.5.0-cp311-abi3-win_arm64.whl
| Download URL | qliff-0.5.0-cp311-abi3-win_arm64.whl |
|---|---|
| Size | 308.5 kB |
| Tags | CPython 3.11 Windows ARM64 abi3 |
|
SHA-256 checksum How to use checksums |
0ab2cb8b27ba28a158d2c9462bb1ad90c1bfe95336128021a9f477a204f4e5c8
|
|
BLAKE2b-256 checksum How to use checksums |
d4f5c0f1770e165817c1841eb1e71fbd26d6dd1de5a53e4fa8f79ac3a193ee19
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.3
|
Release files / qliff-0.5.0-cp311-abi3-win_amd64.whl
| Download URL | qliff-0.5.0-cp311-abi3-win_amd64.whl |
|---|---|
| Size | 326.0 kB |
| Tags | CPython 3.11 Windows x86-64 abi3 |
|
SHA-256 checksum How to use checksums |
2c74a76852f3d3daa3c9ccb1e2470b9d52b7d3ac25890374a88012da55b35be6
|
|
BLAKE2b-256 checksum How to use checksums |
eaeea4925d0169023bd02df1ef07df710a498d893731ca957fb0a922ff244dbb
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.3
|
Release files / qliff-0.5.0-cp311-abi3-manylinux_2_28_x86_64.whl
| Download URL | qliff-0.5.0-cp311-abi3-manylinux_2_28_x86_64.whl |
|---|---|
| Size | 453.2 kB |
| Tags | CPython 3.11 Linux glibc 2.28+ x86-64 abi3 |
|
SHA-256 checksum How to use checksums |
9d04b57c53dec984df6075e628d8a49feadb8d6ef35eed0c79612edccf22abef
|
|
BLAKE2b-256 checksum How to use checksums |
02fd94e17275a06433f3297b516bd761d5354a922a7e36e6a97c49f0173a6501
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.3
|
Release files / qliff-0.5.0-cp311-abi3-manylinux_2_28_aarch64.whl
| Download URL | qliff-0.5.0-cp311-abi3-manylinux_2_28_aarch64.whl |
|---|---|
| Size | 446.0 kB |
| Tags | CPython 3.11 Linux glibc 2.28+ ARM64 abi3 |
|
SHA-256 checksum How to use checksums |
92041b0ff4b08c94cc376920b0a58d0b8ffc67ad541133a9bf0497a423472e77
|
|
BLAKE2b-256 checksum How to use checksums |
c009d1d7ad8a6c634d230b800f089aea58dc541aa97e68de25bab9859a0a43d8
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.3
|
Release files / qliff-0.5.0-cp311-abi3-macosx_11_0_x86_64.whl
| Download URL | qliff-0.5.0-cp311-abi3-macosx_11_0_x86_64.whl |
|---|---|
| Size | 422.4 kB |
| Tags | CPython 3.11 abi3 macOS 11.0+ x86-64 |
|
SHA-256 checksum How to use checksums |
48a2cbcc30ea2f08c0b55528a9a500f48fe049f294441c7501e4d4cd00b5d8f0
|
|
BLAKE2b-256 checksum How to use checksums |
3ae88defca4504afcd0276e550a2d4f706b1da611d301ec33b5a545dc2c2d39a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.3
|
Release files / qliff-0.5.0-cp311-abi3-macosx_11_0_arm64.whl
| Download URL | qliff-0.5.0-cp311-abi3-macosx_11_0_arm64.whl |
|---|---|
| Size | 401.6 kB |
| Tags | CPython 3.11 abi3 macOS 11.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
0f8e7a08626e2a87a603b2ffacbc0a60b4af5ba6a4f7237525c2baa93b0f3a8e
|
|
BLAKE2b-256 checksum How to use checksums |
62c311a9cc11d482bf0652163e6d69f03e5b2cfbfa141141875c8821a2339488
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.3
|
Security for qliff 0.5.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.