File size: 6,402 Bytes
4198d45 e395748 4198d45 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 | {
"schema_version": 1,
"release_version": "v1",
"validated_at": "2026-07-29",
"status": "pass",
"upstream": {
"repository": "https://github.com/NVIDIA/Ising-Decoding.git",
"commit": "33acb152e403bc189f2effdb07f1a87b34c745f1",
"patch_sha256": "15f8e7d2f18158a769a5ba06d183239090c88c606519c8468f9a814998c6a986",
"git_apply_check": "pass",
"clean_clone_gpu_smoke": "pass"
},
"artifacts": {
"QAdapt": {
"relative_path": "Qadapt-r9-v1.safetensors",
"safetensors_sha256": "65f979c9f23f1b13b76876c6e5df204d1bd28b0e3322483d966db0efb840600d",
"source_checkpoint_sha256": "59d55a948a1a99458f7ccb0bba83335a779ef90eb8b36125407b65d45d771f9f",
"tensor_equality": "exact",
"tensor_count": 94,
"parameter_count": 650374
},
"Ising-Fast-T0-e100": {
"relative_path": "baselines/ising-fast-t0-e100/model.safetensors",
"safetensors_sha256": "af5e13e389358d4aa19afa92f6d1ff368e8f64e8497f23c51ddeed9d367f724a",
"source_checkpoint_sha256": "c23c9b1507cf4d213ce1b6526f87ee9e2f9db2bd8dc336c16f915b77724a0661",
"tensor_equality": "exact",
"tensor_count": 8,
"parameter_count": 912772
}
},
"environment": {
"python": "3.12.3",
"torch": "2.7.0a0+79aa17489c.nv25.04",
"numpy": "1.26.4",
"stim": "1.16.0",
"pymatching": "2.4.0",
"safetensors": "0.5.3",
"cuda": "12.9",
"gpu": "NVIDIA A100-SXM4-80GB"
},
"t0_t4": {
"status": "pass",
"protocol": {
"distance": 9,
"rounds": 9,
"shots_per_basis_per_task": 262144,
"bases": ["X", "Z"],
"seed": 12345
},
"actual_mean_ler_by_task": {
"T0": {
"pymatching": 0.045360565185546875,
"ising-fast": 0.04070091247558594,
"qadapt": 0.036182403564453125
},
"T1": {
"pymatching": 0.05428314208984375,
"ising-fast": 0.05145454406738281,
"qadapt": 0.046100616455078125
},
"T2": {
"pymatching": 0.15405654907226562,
"ising-fast": 0.14002418518066406,
"qadapt": 0.1307964324951172
},
"T3": {
"pymatching": 0.0500640869140625,
"ising-fast": 0.04494285583496094,
"qadapt": 0.040164947509765625
},
"T4": {
"pymatching": 0.098663330078125,
"ising-fast": 0.09106636047363281,
"qadapt": 0.0831298828125
}
},
"actual_mean_ler_across_tasks": {
"pymatching": 0.08048553466796875,
"ising-fast": 0.07363777160644532,
"qadapt": 0.06727485656738282
},
"qadapt_reduction_vs_ising_fast_percent": 8.64083051435735,
"max_absolute_ler_delta_vs_release_reference": 0.000681,
"interpretation": "All five tasks reproduce within ordinary finite-shot and software-version variation. The paper architecture table used different T0-only epochs and is not evidence for these final e100 artifacts."
},
"synthetic_ood": {
"status": "pass",
"protocol": {
"axis_combinations": 11,
"multipliers": [1.2, 1.5, 2.0, 2.5, 3.0],
"distances": [7, 9],
"jobs": 110,
"shots_per_basis_per_job": 262144,
"bases": ["X", "Z"]
},
"actual": {
"d7": {
"pymatching_mean_ler": 0.2453592127019709,
"ising_fast_mean_ler": 0.23453282442959872,
"qadapt_mean_ler": 0.2268583124334162,
"qadapt_reduction_vs_ising_fast_percent": 3.2722549668036893,
"qadapt_wins": 55,
"comparisons": 55,
"ising_fast_backend_latency_us_per_round": 2.446139354042408,
"qadapt_backend_latency_us_per_round": 2.300856614889897
},
"d9": {
"pymatching_mean_ler": 0.2519276358864524,
"ising_fast_mean_ler": 0.24441146850585938,
"qadapt_mean_ler": 0.23654133189808238,
"qadapt_reduction_vs_ising_fast_percent": 3.2200357274103615,
"qadapt_wins": 55,
"comparisons": 55,
"ising_fast_backend_latency_us_per_round": 5.014288594494714,
"qadapt_backend_latency_us_per_round": 4.730247553071063
}
},
"absolute_ler_delta_vs_paper": {
"d7_ising_fast": 0.00006020285866476,
"d7_qadapt": -0.00014780217950994,
"d9_ising_fast": -0.00002417130903765,
"d9_qadapt": 0.00000766407359728
},
"interpretation": "QAdapt wins all 110 retained comparisons. Backend timing is machine-dependent and excludes neural inference and transfers."
},
"willow": {
"status": "pass",
"protocol": {
"rounds": 10,
"bases": ["X", "Z"],
"fine_tuning": false,
"d5_shots": 400000,
"d7_shots": 100000
},
"actual": {
"d5": {
"ising_fast_logical_errors": 39852,
"ising_fast_ler": 0.09963,
"qadapt_logical_errors": 37543,
"qadapt_ler": 0.0938575,
"qadapt_reduction_vs_ising_fast_percent": 5.79393756900532,
"ising_fast_backend_latency_us_per_round": 0.7209028746001422,
"qadapt_backend_latency_us_per_round": 0.6945591699331999
},
"d7": {
"ising_fast_logical_errors": 8412,
"ising_fast_ler": 0.08412,
"qadapt_logical_errors": 8201,
"qadapt_ler": 0.08201,
"qadapt_reduction_vs_ising_fast_percent": 2.50832144555397,
"ising_fast_backend_latency_us_per_round": 1.344656078144908,
"qadapt_backend_latency_us_per_round": 1.3134985323995352
}
},
"ler_and_logical_error_match_to_paper": "exact",
"interpretation": "The third-party archive is not redistributed. Timing is machine-dependent; the LER and logical-error counts exactly reproduce the report."
},
"tests": {
"public_config": {"tests": 36, "status": "pass"},
"safetensors_export_and_load": {"tests": 9, "status": "pass"},
"compileall": "pass",
"full_upstream_suite": {
"tests": 456,
"successful": 424,
"skipped": 15,
"errors": 16,
"failures": 1,
"status": "environment_incomplete",
"non_release_blockers": {
"missing_optional_cuquantum": 11,
"missing_upstream_nvidia_model_files": 4,
"missing_optional_onnx_graphsurgeon": 2
}
},
"python_compat_script": {
"status": "environment_constraint_conflict",
"detail": "The script pinned the container's NVIDIA pre-release torch build while the public requirements requested the normal torch package, so pip stopped before code checks."
}
}
}
|