Thuật ngữ AI
Từ điển đầy đủ về Trí tuệ nhân tạo
RAPIDS
Open-source software suite from NVIDIA for GPU-accelerated data analytics and machine learning, including cuDF, cuML, cuGraph, and other integrated libraries.
cuDF
GPU-accelerated tabular data manipulation library offering a pandas-compatible API for processing massive datasets in GPU memory.
cuML
GPU-accelerated machine learning library providing fast implementations of algorithms like random forests, k-means, linear regression, and SVM with scikit-learn compatible API.
cuGraph
GPU-accelerated graph analytics library implementing centrality algorithms, community detection, graph traversal, and PageRank on GPU-optimized data structures.
Numba
Just-in-time compiler for Python that transforms Python and NumPy functions into optimized machine code for CPU and GPU via @jit and @cuda.jit annotations.
NCCL
Multi-GPU and multi-node collective communication library optimized for NVIDIA GPUs implementing all-reduce, broadcast, reduce, and gather operations with minimal latency.
Thrust
C++ parallel programming library inspired by STL providing parallel algorithms, containers, and iterators for high-performance CUDA application development.
cuRAND
CUDA random number generation library offering optimized pseudo-random and quasi-random generators for GPU with various statistical distributions.
Dask-CUDA
Dask extension enabling distributed computation orchestration across multiple NVIDIA GPUs with automatic data transfer management and inter-GPU parallelism.
XGBoost GPU
GPU-accelerated implementation of the XGBoost algorithm using CUDA for fast training of gradient boosting models on large datasets.
LightGBM GPU
GPU version of the LightGBM framework optimizing gradient boosting training via parallel histograms and feature-wise parallelism on CUDA architectures.
cuSPARSE
CUDA sparse linear algebra library providing optimized operations on sparse matrices and vectors for scientific and analytical applications.
cuSOLVER
CUDA linear system solver library implementing optimized LU, QR, SVD algorithms and eigenvalue decompositions for GPUs.