Docs.rs

OxiCUDA: Compressed Sensing, Sparse Recovery, and Low-Rank Matrix Completion

Fully-customizable automation runtime using Lua

Continual and lifelong learning primitives for OxiCUDA: EWC/SI/MAS regularization, PackNet/Piggyback/Progressive architectures, Experience Replay/GEM/A-GEM, DER++, forgetting metrics — pure Rust, zero CUDA SDK dependency.

OxiCUDA BLAS - GPU-accelerated BLAS operations (cuBLAS equivalent)

Causal-inference primitives for OxiCUDA — NOTEARS, PC/GES, IPW, DML, DragonNet, causal forest, do-calculus, counterfactuals

Bayesian deep learning primitives for OxiCUDA: variational inference, Bayesian layers, MC Dropout, Deep Ensembles, SWAG, Laplace approximation, uncertainty calibration — pure Rust, zero CUDA SDK dependency.

Audio/Speech ML architectures for OxiCUDA: Conformer encoder, Wav2Vec2 CNN feature extractor, CTC forward+beam-search, WaveNet dilated stack, SpecAugment, speaker embeddings (x-vector/TDNN) — pure Rust, zero CUDA SDK dependency.

Anomaly detection primitives for OxiCUDA — DeepSVDD, AE/VAE reconstruction, LOF, COPOD, isolation scoring, statistical methods, ensemble

OxiCUDA Launch - Type-safe GPU kernel launch infrastructure

Approximate Nearest Neighbor & vector-search primitives for OxiCUDA — HNSW, IVF, Product Quantization, IVFPQ, LSH, k-NN graph, top-K selection

Adversarial robustness primitives for OxiCUDA: FGSM, PGD, MIM, CW, AutoPGD attacks; TRADES, MART, randomized smoothing, certified bounds — pure Rust, zero CUDA SDK dependency.

Vim-modal terminal editor: standalone TUI built on the hjkl engine.

Host-agnostic application layer for the hjkl editor. Used by apps/hjkl (TUI).

Ratatui adapter for hjkl-syntax: converts renderer-agnostic RenderOutput spans to ratatui Span/Line and routes DiagSigns to hjkl_buffer::Sign.

OxiCUDA Graph — CUDA Graph execution engine with operator fusion, buffer lifetime analysis, stream partitioning, and optimized execution planning