algenta-general-b
Domain kernels from ensembles to model compare. Compiled Mojo, loaded in-process.
100 modules · 894 functions · CPU
Get started
pip install kernels torch
from kernels import get_kernel
kernel = get_kernel(
"thyn-ai/algenta-general-b",
version=1,
backend="cpu",
trust_remote_code=["thyn-ai/algenta-general-b"],
)
kernel.epidemiology.herd_immunity_threshold(4) # -> 0.75 herd immunity threshold, 1 - 1/R0
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("esrnn.bias_correction", "esrnn_apply_bias", [...])
What's inside
| Module | Functions | What it does |
|---|---|---|
ensembles |
4 | Model ensembling: majority and weighted voting, prediction blending, ridge stacking |
entomology |
12 | Insect science: degree days, exponential growth, generation time, Simpson diversity, biocontrol |
epidemiology |
10 | Epidemic models: SIR and SEIR steps, R0, herd immunity threshold, final size, doubling time |
ergonomics |
12 | Ergonomics: NIOSH lifting equation, noise dose, decibel addition, Fitts' and Hick's laws, BMI |
esrnn.bias_correction |
2 | Forecast bias correction: mean of holdout residuals, added as a constant to every forecast step |
esrnn.global_residual |
7 | Forecast residual correction: Holt, damped or Theta base by in-sample MAE, scaled windows |
esrnn.per_frequency |
2 | Ridge-regression forecast correction fitted per data frequency over pooled residual windows |
etiology |
10 | Risk and diagnostics: relative risk, odds ratio, attributable risk, NNT, sensitivity, PPV, NPV |
evaluation.binning |
4 | Calibration binning: equal-width and equal-count bins, per-bin counts, assignments and edges |
evaluation.contract |
13 | Status codes, input validation, binning policy names and work budget for calibration metrics |
evaluation.gates |
3 | Release gates: metric regression with tolerance and direction, population stability index drift |
evaluation.resample |
3 | Seeded percentile bootstrap confidence intervals for Brier, log-loss, accuracy and ECE |
evaluation.scores |
6 | Brier score with Murphy decomposition, log-loss, ECE and MCE, reliability diagram data |
evaluation_feedback |
10 | Model monitoring: rolling accuracy, KS and PSI drift terms, prediction shift, streaming AUC |
evolutionary_search |
7 | Seeded genetic algorithm: elitism, tournament selection, uniform crossover, Gaussian mutation |
exactnum.eft |
24 | Error-free transformations and double-double arithmetic: two-sum, two-product, compensated sums |
exactnum.lstsq |
9 | Least squares by Householder QR in double-double precision, polynomial fits, float64 controls |
execution_planner |
10 | Serving heuristics: op fusion score, roofline device placement, KV eviction, batching, prefetch |
explainability |
4 | Model explanations: permutation importance, partial dependence, KernelSHAP coalition weights |
fault_tolerance |
10 | Training fault tolerance: optimal checkpoint interval, failure probability, gradient staleness |
feature_engineering |
10 | Feature transforms: min-max, z-score, robust scaling, Box-Cox, Yeo-Johnson, target encoding |
feature_selection |
9 | Feature selection: variance threshold, correlation filter, F-test, info gain, forward selection |
fermentation |
10 | Fermentation: Gay-Lussac ethanol yield, ABV from gravity, Monod growth, oxygen transfer rate |
filepath.pathlib |
8 | Path strings: join, basename, dirname, extension, stem, split, normalize; no filesystem access |
fire_safety |
12 | Fire safety: fire load density, heat release rate, Thomas flashover, smoke layer, egress time |
fisheries |
12 | Fish stocks: CPUE, Schaefer surplus production and MSY, mortality rates, von Bertalanffy growth |
fluid |
12 | Fluid flow: Reynolds and Mach numbers, Bernoulli, Darcy-Weisbach loss, drag, buoyancy, Stokes |
food_science |
12 | Food science: water activity, pH, Q10 shelf life, D and z value thermal death time, moisture |
forecast_metrics |
10 | Forecast scoring: Brier skill score, log loss, reliability bins, ECE, Murphy decomposition |
forecasting.backtest |
4 | Rolling-origin backtests over expanding or sliding windows; MAE, RMSE, MAPE, sMAPE and MASE |
forecasting.bagging |
3 | Bagged forecasts (Bergmeir-Hyndman): moving-block bootstrap of the remainder, refit and average |
forecasting.contract |
18 | Shared forecasting input checks: series, horizon, period, parameters, tolerances, quantiles |
forecasting.fitted |
7 | Fitted smoothing: Holt and damped trend with fitted initial state, seasonal ETS, Theta, DOTM |
forecasting.hierarchy |
5 | Hierarchical forecast reconciliation: bottom-up and proportional top-down, plus coherence audit |
forecasting.intermittent |
4 | Intermittent demand: Croston, SBA and SBJ rate forecasts, Syntetos-Boylan demand classification |
forecasting.origin_select |
3 | Method selection by paired rolling-origin errors, with a paired-t tournament that blends ties |
forecasting.point |
8 | Point forecasts: simple, Holt and seasonal exponential smoothing, Theta, Yule-Walker AR, naive |
forecasting.pool |
5 | Holdout-scored pool of SES, Holt, damped, Theta, ETS and naive forecasts: pick best or blend |
forecasting.select |
2 | Per-series choice among SES, damped trend and Theta by in-sample error, or inverse-MAE blend |
forensics |
12 | Forensics: Newton cooling time of death, Widmark blood alcohol, bullet drop, bloodstain angle |
formal_logic |
14 | Propositional logic: connectives, modus ponens and tollens, syllogisms, tautology checks |
fused_kernels |
10 | Fused transformer math: masked attention scores, softmax, LayerNorm, bias plus ReLU/GELU, RoPE |
fuzzy_logic |
11 | Fuzzy logic: triangular, trapezoidal and Gaussian membership, min/max operators, defuzzifiers |
game_physics |
12 | Game physics: Euler and Verlet motion, elastic collisions, restitution, projectiles, springs |
game_theory |
8 | Game theory: 2x2 mixed Nash equilibrium, minimax, dominated strategies, iterated PD, Shapley |
gaussian_process |
5 | Exact Gaussian process regression, RBF or Matern 3/2 kernel: Cholesky fit, mean and variance |
gbdt |
3 | Gradient-boosted regression trees with squared loss, depth-limited splits and row subsampling |
generation_loop |
10 | Token sampling for text generation: temperature, top-k, top-p, min-p, typical, eta, penalties |
genomics |
10 | Genome statistics: GC content, coverage depth, N50, SNP density, indel rate, Phred error rate |
glaciology |
10 | Glaciers: mass balance, Glen's flow law, basal shear stress, degree-day melt, sea-level rise |
greenhouse |
12 | Greenhouse climate: conduction heat loss, solar gain, ventilation, CO2 enrichment, light hours |
hematology |
10 | Blood indices: hematocrit, MCV, MCH, MCHC, RDW, INR, oxygen capacity, relative viscosity, ESR |
hmm |
4 | Hidden Markov models: log-space forward-backward, Viterbi decoding, Baum-Welch (EM) training |
horticulture |
10 | Crop physiology: light-response photosynthesis, growth and assimilation rates, VPD, chill hours |
hvac |
12 | HVAC loads: sensible and latent heat, cooling tons, air changes, duct velocity, degree days |
hybrid_reasoning |
10 | ML and simulation blending, feasibility gating, rule penalties, rollout values, uncertainty |
hydraulics |
12 | Fluid power: Pascal's law, cylinder force and velocity, pump displacement, accumulator sizing |
hydrology |
12 | Hydrology and hydraulics: Manning, Darcy-Weisbach, rational method, SCS curve number |
hydrometeorology |
10 | Rainfall-runoff: SCS curve number, Green-Ampt infiltration, Palmer drought index, Gumbel floods |
hydroponics |
12 | Hydroponic nutrients and lighting: EC/TDS conversion, ppm dosing, pH acid dose, DLI, VPD, ROI |
imbalance |
5 | Imbalanced-class resampling: SMOTE, seeded random over- and undersampling, balanced weights |
immunology |
10 | Immunology: antibody titer, binding equilibrium, T-cell growth, vaccine efficacy, herd immunity |
information_theory |
10 | Shannon entropy, joint and conditional entropy, mutual information, KL divergence, BSC capacity |
insurance |
12 | Actuarial pricing: pure and gross premium, loss and combined ratios, Buhlmann credibility |
integrate |
6 | Numerical integration: trapezoid, Simpson, Romberg, cumulative trapezoid, Gauss-Legendre |
interpolate |
6 | Interpolation: linear, nearest neighbor, natural cubic spline, Lagrange, linear extrapolation |
inventory |
11 | Inventory planning: EOQ, reorder point, safety stock, newsvendor, ABC classification, fill rate |
kalman |
11 | Kalman filter predict and update steps, a 1D smoother, and small dense matrix utilities |
laundry |
12 | Laundry science: water and detergent dosing, wash heating energy, spin g-force, drying time |
lidar |
12 | Airborne LiDAR geometry: time-of-flight range, swath width, footprint, density, canopy height |
lidar_mapping |
10 | LiDAR mapping: canopy height model, leaf area index, ground filtering, beam footprint, RMSE |
limnochemistry |
11 | Freshwater chemistry: alkalinity, ionic strength, Davies activity, calcite saturation, redox pe |
limnology |
10 | Lake science: retention time, flushing rate, Carlson trophic state, Secchi depth, DO saturation |
linear_models |
4 | Logistic regression by Newton (IRLS) with standard errors, elastic net by coordinate descent |
linear_programming |
8 | Two-variable linear programs by vertex enumeration, RHS sensitivity, northwest corner transport |
linguistics |
12 | Readability (Flesch, Gunning fog, ARI, Coleman-Liau), TTR, Yule's K, Heaps' and Zipf's laws |
logging.logger |
1 | Structured logging: one JSON line per event with debug, info, warning and error levels |
logic_circuit |
13 | Digital logic: basic gates, half, full and ripple-carry adders, 2:1 multiplexer, 2-to-4 decoder |
logistics |
12 | Supply-chain metrics: fill rate, perfect orders, inventory turns, days of supply, on-time rate |
lora_adapters |
10 | LoRA, QLoRA, DoRA, prefix and prompt tuning: forward terms, merged weights, gradients, scaling |
lottery |
12 | Lottery math: combinations, permutations, odds, expected value, house edge, annuity vs lump sum |
marine |
12 | Oceanography: seawater density, wave speed and period, Coriolis, Ekman depth, stratification |
materials |
12 | Materials: elastic moduli, stress, strain, thermal expansion, Basquin fatigue, beam deflection |
metabolomics |
10 | Metabolomics: log2 fold change, pathway enrichment, Tanimoto similarity, flux ratios, VIP |
metallurgy |
12 | Metallurgy: Hall-Petch, Johnson-Cook, carbon equivalent, Jominy, Avrami, Hollomon-Jaffe |
meteorology |
10 | Weather: Magnus dew point, wet-bulb estimate, heat index, wind chill, potential temperature |
metrics_agg |
12 | Metrics: counters, gauges, histograms with exact percentiles, rates, throughput, latency stats |
microbiology |
10 | Microbial growth and inactivation: doubling time, Monod kinetics, CFU, MIC, D-value, z-value |
mineralogy |
10 | Mineralogy: density, specific gravity, Bragg d-spacing, formula weight, saturation index, Mohs |
mining |
12 | Mining: ore grade, strip ratio, recovery, Bond work index, cut-off grade, net smelter return |
mixture_models |
5 | Gaussian mixture models: EM with k-means++ seeding, 1D and diagonal fits, responsibilities, BIC |
mixture_of_experts |
10 | Mixture-of-experts routing: gate scores, load-balance and z-loss, capacity, dispatch, Soft-MoE |
ml.classifiers |
7 | L1 and L2 distances, cosine similarity, sigmoid, softmax, k-nearest-neighbour voting, k-means |
ml.clustering |
4 | K-means clustering (Lloyd's algorithm): centroids, labels, inertia and silhouette score |
ml.knn |
10 | K-nearest neighbours: majority and weighted voting, regression, leave-one-out choice of k |
ml.metrics |
7 | Model evaluation: accuracy, confusion matrix, precision, recall, F1, MSE, RMSE, MAE, R² |
ml.preprocessing |
6 | Scaling (min-max, z-score, robust), one-hot encoding, polynomial features, train/test split |
ml.random_forest |
8 | Random forest regression over bootstrapped stumps: out-of-bag R², permutation importance |
model_abi |
10 | Model layout arithmetic: tensor strides, data-type byte widths, offsets, head dims, GQA, RoPE |
model_compare |
4 | Paired model comparison: bootstrap CIs (means, Sharpe), exact McNemar, Diebold-Mariano |
kernel.CONTRACT holds every signature, including the length rules for list arguments;
help(kernel.ensembles) documents each function.
Not included
gbdt.gbdt_predict— walks a tree encoding that must be well-formed; a malformed tree 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