File size: 2,060 Bytes
8ec5bf2 cbea6c7 8ec5bf2 | 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 | {
"schema": "szl.kernel-operational/v1",
"hub": "SZLHOLDINGS/szl-nemo",
"github_source": "szl-holdings/szl-nemo",
"github_sha": "6e2f91181788679622613eed6b53871301c536e6",
"computed_at": "2026-08-29T16:00:43Z",
"tests": "PASS",
"eval": "PASS",
"get_kernel": "UNAVAILABLE",
"hub_kernel_repo": false,
"hub_model_repo": true,
"formula_tax": {
"kind": "SOFTWARE",
"registry_count": 21,
"locked_8": [
"F1",
"F4",
"F7",
"F11",
"F12",
"F18",
"F19",
"F22"
],
"lambda_aggregate": 0.316227766016838,
"lambda_label": "ADVISORY",
"lambda_uniqueness": "CONJECTURE_1",
"khipu_merkle_sha256": "a76be71282d5ae50ab2db239a0b36badd47b038b23d20ada03e0167ca9c22013",
"honesty": "Λ roll-up is advisory geomean over binary checks. Uniqueness is Conjecture 1, never a theorem. locked-8 = F1 F4 F7 F11 F12 F18 F19 F22."
},
"invariants_i1_i8": [
"I1 receipt-chain-continuity",
"I2 ledger-failure-shape",
"I3 served-run-has-model",
"I4 signed-columns-atomic",
"I5 loop-steps-positive",
"I6 receipt-ed25519-verify",
"I7 receipt-columns-consistent",
"I8 flywheel-lineage"
],
"load_path": "from kernels import get_kernel\nk = get_kernel(\"SZLHOLDINGS/szl-nemo\", revision=\"main\", trust_remote_code=True)",
"honesty": [
"Unsloth / LoRA training is the wrong tool for these kernels.",
"Λ uniqueness is Conjecture 1, never a theorem.",
"locked-8 formulas: F1 F4 F7 F11 F12 F18 F19 F22.",
"Do not pickle/joblib.load Hub kernels. GitHub/torch-ext source is the approved path.",
"torch.compile fullgraph failures on Windows Blackwell are MEASURED, not hidden.",
"Receipted kernels, not silent CUDA. No tokens/s or joules claimed here.",
"Do not convert a model-type card into a kernel-type repo by deleting nonexistent weights.",
"szl-nemo is a sklearn recipe-conformance surrogate, not a Kernel Hub CUDA kernel. Do not invent a kernel-type conversion."
]
}
|