1/20
Looks like no tags are added yet.
Artificial Intelligence (AI)
Broad field of computer science focused on building systems that perform tasks requiring human-like intelligence.
Machine Learning (ML)
Subset of AI where systems learn patterns directly from data instead of following hand-written rules.
Deep Learning (DL)
Subset of ML using neural networks with many layers, requiring large datasets and heavy computation.
Generative AI
Branch of artificial intelligence focused on models that create original content such as text, images, audio, and code.
Large Language Model (LLM)
Transformer-based model trained on vast amounts of text to understand, process, and generate natural language.
Training
Process of adjusting a model's weights by learning from data, typically requiring heavy computation on multi-GPU clusters.
Inference
Stage of using a fully trained model in production to generate predictions or outputs continuously for users.
Fine-tuning
Process of further training a pretrained model on smaller, task-specific datasets to adapt it for a distinct use case.
Accelerated Computing
Use of specialized processors (like GPUs) alongside traditional CPUs to speed up compute-intensive workloads.
CPU vs. GPU Architecture
Hardware distinction where sequential processing uses a few powerful cores, whereas parallel processing uses thousands of simpler cores built for matrix math.
CUDA
NVIDIA's parallel computing platform and programming model that forms the foundation of its software stack.
CUDA-X
Collection of GPU-accelerated libraries built on top of parallel computing platforms, including cuDNN, cuBLAS, and NCCL.
cuDNN
GPU-accelerated library providing deep neural network building blocks for frameworks like PyTorch and TensorFlow.
NCCL
NVIDIA Collective Communications Library providing fast multi-GPU data exchange across single or multi-node systems.
NGC
Catalog of GPU-optimized software containing containers, pretrained models, Helm charts, and SDKs.
NVIDIA AI Enterprise
Licensed, enterprise-supported software suite for building and deploying production AI applications.
Triton Inference Server
Open-source software designed to serve models from multiple frameworks in production with features like dynamic batching.
TensorRT
SDK that optimizes trained models for high-speed execution on GPUs via layer fusion and lower precision (FP16/INT8).
NIM
NVIDIA Inference Microservices that package optimized models with standard APIs inside prebuilt containers for rapid deployment.
RAPIDS
Suite of GPU-accelerated software libraries for data science and data processing.
On-Premises vs. Cloud Infrastructure
Deployment model contrasting owned hardware (capex, full control) with rented, elastic hardware (opex, pay-as-you-go).