algenta-general-d
Domain kernels from rlhf dpo to wind energy. Compiled Mojo, loaded in-process.
100 modules · 1049 functions · CPU
Get started
pip install kernels torch
from kernels import get_kernel
kernel = get_kernel(
"thyn-ai/algenta-general-d",
version=1,
backend="cpu",
trust_remote_code=["thyn-ai/algenta-general-d"],
)
kernel.set_ops.jaccard_index([1, 2, 3, 4], [3, 4, 5, 6]) # -> 0.3333 Jaccard similarity
backend="cpu" selects the CPU build. On a Mac the loader otherwise looks for a Metal build, which
this family does not ship. trust_remote_code names the repositories you allow; Hugging Face's
trusted publishers load without it.
Plain Python in, plain Python out. Lists, tuples, buffers and tensors are accepted wherever the
contract expects a list; structured results are dictionaries. Some functions take a list and the
number of elements to use from it, which may not exceed the list's length; multi-dimensional data is
passed flattened, row-major, with its dimensions. Functions that update an argument do so in place,
as help() says. Every call is checked against the published contract before it reaches native
code. An invalid call raises KernelError with a stable code, never a crash. Engine kernels
report shape and finiteness problems as a status; the wrapper raises KernelError named after it.
Any function can also be called by name, with args as a list or a dict of parameter names:
kernel.execute("sheet.api", "sheet_eval_cell", [...])
What's inside
| Module | Functions | What it does |
|---|---|---|
rlhf_dpo |
10 | RLHF losses: DPO, IPO, ORPO, PPO clipping, Bradley-Terry reward, GAE and GRPO advantages |
robotics |
12 | Two-link arm kinematics, one Jacobian entry, PID control, dead reckoning, odometry, diff drive |
robustness |
8 | Sensitivity on built-in test functions: one-at-a-time, Morris, Sobol first-order, tornado data |
root_finding |
9 | Scalar roots of built-in test functions: bisection, Newton, secant, regula falsi, bracketing |
rubber |
12 | Elastomer tests: Shore A to modulus, tensile and compression set, rebound, crosslink density |
scheduler_ops |
10 | Queue scheduling: fair priority, WFQ credit, token bucket, backpressure, backoff with jitter |
scheduling_algo |
9 | CPU scheduling: FCFS, SJF, priority, time-quantum cycling, plus wait and throughput metrics |
scipy_optimize_ops |
10 | Optimizer steps: BFGS update term, Fletcher-Reeves and Polak-Ribiere CG, Armijo and Wolfe tests |
scoring |
11 | Credit scoring: logit, weight of evidence, information value, Gini from AUC, KS, score bands |
security_governance |
10 | Governance heuristics: prompt and output safety, PII sensitivity, Gaussian DP noise scale, RBAC |
sedimentology |
10 | Sediment transport: Stokes settling, Shields stress, Meyer-Peter Muller bedload, Rouse profile |
seismology |
12 | Earthquakes: Richter and moment magnitude, energy, P and S travel times, PGA, Gutenberg-Richter |
semiconductors |
12 | Device physics: carrier density, drift, resistivity, depletion width, diode and MOSFET currents |
seq_nn |
9 | LSTM and GRU cells with exact forward and backward passes, parameter counts, seeded Xavier init |
series |
10 | Forecasts: moving average, OLS trend, exponential smoothing, seasonal naive, drift, MAE, RMSE |
set_ops |
12 | Sorted-integer sets: union, intersection, difference, subset tests, Jaccard index, membership |
sheet.api |
2 | Recalculate an extracted Excel workbook and compare every formula cell against its cached value |
sheet.engine |
1 | Demand-driven topological recalculation of spreadsheet formulas with named circular references |
sheet.formula |
3 | Tokenizer and recursive-descent parser for Excel formula text with Excel operator precedence |
sheet.grid |
1 | Loads extracted workbook text into sheets, defined names, constant cells and shared formulas |
sheet.refs |
7 | A1 cell and range references: column letters, packed coordinates, absolute markers, parsing |
sheet.values |
19 | Typed cell values: Excel errors, 15-digit number formatting, wildcard and case-insensitive text |
silviculture |
10 | Forest stand metrics: basal area, site index, Scribner volume, Reineke SDI, increment, thinning |
simulation |
8 | Discrete-event simulation: schedule, pop and cancel events, run to a deadline, M/M/1 queue |
sketch |
9 | Probabilistic sketches: Count-Min Sketch frequency and HyperLogLog cardinality estimates |
soil_mechanics |
10 | Geotechnics: Terzaghi bearing capacity, consolidation, Mohr-Coulomb shear, Rankine Ka, Hazen k |
soil_science |
12 | Soil physics: porosity, water content, field capacity, wilting point, CEC, Horton infiltration |
solar_energy |
12 | Solar PV: irradiance, panel output, NOCT cell temperature, sun position, capacity factor, LCOE |
sorting |
10 | Sorting: quicksort, mergesort, heapsort, insertion, selection, bubble and counting sort |
sparse |
6 | Sparse COO matrices: from dense, matrix-vector product, addition, transpose, density, trace |
sparse_numerics.csr |
34 | CSR and CSC sparse layouts: build from triplets or dense, validate, merge duplicates, transpose |
sparse_numerics.diagnostics |
21 | Sparse matrix health: density, duplicates, NaN/Inf entries, symmetry defect, condition estimate |
sparse_numerics.factor |
11 | Direct sparse solves: pivoted LU, Cholesky for SPD systems, triangular solves, least squares |
sparse_numerics.iterative |
18 | Krylov solvers on CSR: CG, Jacobi or Gauss-Seidel preconditioned CG, CGLS, restarted GMRES |
sparse_numerics.spmv |
16 | CSR kernels: transpose and Kahan-compensated matvec, Gustavson SpGEMM, add, Frobenius norm |
sparse_vector |
10 | Sparse vectors as index/value lists: dot product, merge-add, scale, L2 norm, cosine similarity |
special_functions |
12 | Gamma, beta, digamma, incomplete gamma, erf, erfc, Bessel J0, factorials, binomial coefficients |
speculative_decoding |
10 | Speculative decoding: acceptance, expected tokens, speedup, Medusa, grammar mask, watermarking |
speleology |
10 | Cave science: calcite saturation, stalactite growth, karst spring discharge, U-Th age, radon |
sports_analytics |
12 | Sports stats: OBP, SLG, OPS, ERA, WHIP, NFL passer rating, Pythagorean wins, Elo, TS%, PER |
stochastic |
10 | Simulated paths: random walks, Brownian motion, GBM, Poisson, Ornstein-Uhlenbeck, birth-death |
strategy.contract |
42 | Strategy comparison vocabulary: status names, tolerances, shape and budget checks, RNG streams |
strategy.crn |
17 | Common random numbers: shared scenario tapes, paired policy evaluation, pairing audits |
strategy.equilibrium |
22 | Normal-form games: pure and 2x2 mixed Nash, regret matching, fictitious play, exploitability |
strategy.policy |
8 | Policies: constant, threshold, tabular, mixed and linear rules, greedy from Q, MDP evaluation |
strategy.tournament |
8 | Strategy ranking: paired t-tests, percentile bootstrap, Bonferroni, Holm or BH corrections |
streaming.contract |
21 | Sketch building blocks: keyed Philox hashing, HyperLogLog, Count-Min and KLL error bounds |
streaming.distinct |
11 | HyperLogLog distinct counting: ingest, estimate, register-wise merge, flat snapshot and restore |
streaming.drift |
8 | Stream change detection: resumable two-sided Page-Hinkley and Drift Detection Method (DDM) |
streaming.frequency |
16 | Count-Min sketch: weighted ingest, count queries with bounds, heavy hitters, merge, snapshot |
streaming.quantiles |
15 | Streaming quantiles with a KLL sketch: add, merge, rank, quantile lookup, snapshots, min/max |
streaming.windows |
10 | Rolling windows: mean, standard deviation, min, max, range, EWMA, EW variance, window summary |
structs.containers |
2 | Histogram bin counts over a value range and sorted unique values |
structural |
12 | Structural analysis: bending, shear, axial stress, beam deflection, Euler buckling, von Mises |
structured_output |
10 | Structured output scoring: JSON token validity, SQL clause order, XML depth, field recall |
supply_chain |
12 | Supply chain: bullwhip ratio, MAE, MAPE, bias, tracking signal, EPQ, fill rate, moving average |
survival |
8 | Survival analysis: Kaplan-Meier, Nelson-Aalen, log-rank test, smoothed hazard, RMST, life table |
survival_analysis.base |
9 | Right-censored survival data: validation, time ordering, risk tables, Greenwood terms, z values |
survival_analysis.competing |
3 | Competing risks: Aalen-Johansen cumulative incidence per cause, with step-function lookup |
survival_analysis.cox |
3 | Cox proportional hazards: Newton-Raphson fit with Efron ties, relative risk, Breslow baseline |
survival_analysis.hazard |
2 | Nelson-Aalen cumulative hazard with Aalen variance, log-scale band and Epanechnikov smoothing |
survival_analysis.km |
3 | Kaplan-Meier with Greenwood standard errors, log-log confidence band, median, RMST, step lookup |
survival_analysis.parametric |
2 | Weibull and lognormal maximum likelihood under right censoring: profile likelihood, EM |
sustainability |
12 | ESG metrics: transport carbon, GHG scopes 1 to 3, energy and water intensity, waste diversion |
svm_kernel |
7 | SVM kernels: linear, RBF, polynomial, sigmoid, Gram matrix, centering, alignment, PSD heuristic |
svm_solver |
5 | Support vector machine training by simplified SMO (Platt), linear and RBF kernels, prediction |
taphonomy |
10 | Fossil preservation: soft-tissue decay, bone density loss, completeness, hydraulic sorting, C:N |
telecommunications |
12 | Telecom links: Shannon capacity, Erlang-B blocking, BPSK bit error rate, noise, link margin |
thermodynamics |
11 | Thermodynamics: ideal gas law, Carnot efficiency, entropy, adiabatic and isothermal processes |
timber |
12 | Wood science: board feet, moisture content, allowable stress, moment capacity, fire charring |
toxicology |
12 | Toxicology risk: LD50 dosing, therapeutic index, hazard quotient, cancer risk, Hill response |
transform |
12 | Feature transforms: log, Box-Cox, z-score, min-max, robust and quantile scaling, rank, clamp |
transportation |
12 | Traffic analysis: density, capacity, Webster signal delay, stopping distance, headway, crashes |
tribochemistry |
10 | Sliding contacts: Archard wear, Hertz contact, flash temperature, EHD film, lambda ratio |
tribology |
12 | Friction and wear: Coulomb friction, Hertz pressure, Archard wear, Sommerfeld number, Stribeck |
ultrasonics |
10 | Ultrasound physics: impedance, reflection, attenuation, Doppler shift, PRF, resolution, TI |
units |
8 | Unit conversion for length, mass, temperature, time, area, volume, speed and data sizes |
urban_planning |
12 | Urban planning ratios: density, floor area ratio, lot coverage, walk decay, V/C, affordability |
vector_search |
10 | Approximate nearest-neighbour math: HNSW levels, IVF and PQ distances, LSH buckets, recall@k |
vector_similarity |
10 | Distance and similarity from precomputed sums: cosine, Euclidean, Minkowski, Hamming, Jaccard |
veterinary |
12 | Veterinary dosing and fluids: mg/kg dose, fluid deficit and rate, RER calories, MAC, gestation |
video_accessibility |
9 | WCAG AA text checks for video frames and subtitle bands: worst-case contrast, colour, backplate |
video_color |
9 | Clip colour features: frame and clip mean RGB, dominant colour, saturation, warm ratio, drift |
video_features |
5 | Scene-change scores from frame differences, cut count and rate, peak-motion frame, keyframes |
video_gate |
6 | Routing for video requests: cache reuse, quality and flicker checks, static test, keyframes |
video_ops |
8 | Frame-buffer basics: frame count, mean intensity, frame differences, static ratio, sampling |
video_quality |
6 | Clip usability score from sampled per-frame quality and brightness flicker, plus stability |
video_regions |
7 | Motion regions between frames: difference masks, connected-component boxes, peak and union |
video_similarity |
7 | Clip-to-clip duplicate detection: aligned frame scores, dHash Hamming distance, temporal match |
video_storyboard |
6 | Scene spans from cut detection: start, end and length per scene, midpoint frames, storyboard |
virology |
10 | Virology: R0 and herd immunity threshold, doubling time, PFU titre, burst size, NT50, IC50 |
viticulture |
12 | Grape growing and winemaking: Brix to alcohol, chaptalization, GDD, Ravaz index, molecular SO2 |
volcanology |
12 | Volcanology: VEI energy, lava and lahar speed, eruption column height, tephra, viscosity |
voting |
11 | Voting methods: plurality, Borda, Condorcet pairwise, approval, Droop and Hare quotas, Banzhaf |
wastewater |
12 | Wastewater treatment: BOD removal, HRT and SRT, F/M ratio, SVI, oxygen demand, CT disinfection |
watchmaking |
12 | Mechanical watch maths: gear ratios, beats per hour, balance period, mainspring torque, reserve |
water_chemistry |
10 | Water quality: pH and hydrogen ion, hardness and alkalinity as CaCO3, Langelier index, TDS |
weight_loader |
10 | Checkpoint maths: file offsets, safetensors header size, GGUF block bytes, LoRA merge, hash |
welding |
12 | Welding: heat input, IIW carbon equivalent, cooling rate, dilution, fillet throat, crack risk |
wind_energy |
12 | Wind energy: available power, Betz limit, Cp, tip-speed ratio, capacity factor, Jensen wake |
kernel.CONTRACT holds every signature, including the length rules for list arguments;
help(kernel.rlhf_dpo) documents each function.
Not included
svm_kernel.is_positive_semidefinite— reads its matrix argument as square; a non-square nested list is not expressible as a length rule.svm_kernel.kernel_center— reads its matrix argument as square; a non-square nested list is not expressible as a length rule.
Requirements
- Apple silicon: macOS 15 or later for the CPU build.
- Linux arm64 and x86-64, glibc 2.35 or later.
kernels0.17 or later and PyTorch 2.5 to 2.14. PyTorch has to be installed: the loader picks the build for your PyTorch version. The kernel itself never imports it.
Windows is not supported.
Notes
Calls into one kernel instance run one at a time; use processes for parallelism. Runtime state
does not survive fork(); start worker processes with spawn.
License
Algenta Community License 1.1 (LICENSE). Free for personal, research and open-source use, and
for internal use at organizations with fewer than 50 employees and under $5M in annual revenue.
Beyond that, a commercial license is required: https://algenta.ai/pricing.
Enforced in the compiled library, not just in this text: one concurrent native worker per device (ABI §9). A second process, family or thread waits its turn rather than running in parallel. That is the Community licence's worker floor made real; parallel execution comes with a commercial license.
Support
Generally Available on the platforms listed under Requirements. Within v1, functions are only added;
removals or signature changes ship as v2. Platforms, accelerators and PyTorch releases not listed are not
supported. Documentation: https://docs.algenta.ai (the kernels guide:
https://docs.algenta.ai/guides/kernels-on-hugging-face). Community: https://discord.gg/w8NDsph9an or this
repository's Community tab. Commercial licences and support: https://algenta.ai/pricing.
- Downloads last month
- 2
algenta
mojo
cpu
other
- Torch
- 2.14
- OS
- macoslinux
- Arch
- x86_64aarch64