NVIDIA-Certified Associate: Generative AI LLMs: key terms

Flashcards for this exam NVIDIA-Certified Associate: Generative AI LLMs guide Class dates

• Activation FunctionA nonlinear function applied to a neuron's output that lets networks learn complex patterns.Foundational
• AdamAdaptive Moment Estimation: an optimizer combining momentum with per-parameter adaptive learning rates.Intermediate
• Attention MechanismA technique that lets a model weight the relevance of different input elements when producing each output.Intermediate
• AutoencoderA neural network trained to reconstruct its input through a compressed bottleneck representation.Intermediate
• Autoregressive GenerationProducing output one token at a time, with each new token conditioned on all previously generated tokens.Intermediate
• BackpropagationThe algorithm that computes gradients of the loss with respect to every weight by applying the chain rule backward through the network.Intermediate
• Base ModelA pretrained model before instruction tuning or alignment, which continues text rather than following instructions.Intermediate
• Batch NormalizationNormalizing each layer's inputs across the mini-batch to stabilize and speed up training.Intermediate
• Batch SizeThe number of training examples processed together before the model's parameters are updated.Foundational
• BenchmarkA standardized dataset and metric used to compare model capabilities.Intermediate
• BERTBidirectional Encoder Representations from Transformers: Google's 2018 encoder-only model pretrained with masked language modeling.Intermediate
• BERTScoreA text-similarity metric that compares contextual embeddings of generated and reference text.Advanced
• BF16Brain Floating Point 16: a 16-bit format with FP32's exponent range but less mantissa precision.Intermediate
• BLEUBilingual Evaluation Understudy: a metric scoring generated text by n-gram precision overlap with reference text.Intermediate
• Byte-Pair Encoding (BPE)A subword tokenization algorithm that iteratively merges the most frequent adjacent symbol pairs into new tokens.Advanced
• Catastrophic ForgettingWhen fine-tuning on new data causes a model to lose previously learned capabilities.Advanced
• Chain-of-Thought Prompting (CoT)Prompting a model to reason step by step before giving a final answer.Intermediate
• Context WindowThe maximum number of tokens (input plus output) a model can consider in a single request.Foundational
• Cosine SimilarityA similarity measure based on the angle between two vectors, ranging from -1 to 1.Intermediate
• Cross-Entropy LossA loss that measures the difference between predicted class probabilities and the true labels; standard for classification and LLMs.Intermediate
• CUDACompute Unified Device Architecture: NVIDIA's parallel computing platform and programming model for GPUs.Foundational
• Decoder-Only ArchitectureA transformer that uses only the decoder stack with causal (masked) attention to generate text one token at a time.Intermediate
• Decoding StrategyThe method used to choose each next token from the model's probability distribution.Intermediate
• Deep Learning (DL)A subset of machine learning that uses multi-layer neural networks to learn hierarchical representations directly from raw data.Foundational
• Diffusion ModelA generative model that learns to create data by reversing a gradual noising process.Intermediate
• Direct Preference Optimization (DPO)A preference-tuning method that optimizes a model directly on preferred vs. rejected responses without a separate reward model.Advanced
• Discriminative ModelA model that learns the boundary between classes, estimating P(label | input).Intermediate
• Dot ProductThe sum of element-wise products of two vectors, used as a similarity score in retrieval and attention.Intermediate
• DropoutA regularization technique that randomly deactivates a fraction of neurons during each training step.Intermediate
• EmbeddingA dense numeric vector representing the meaning of text, images or other data so that similar items are close together.Foundational
• Embedding ModelA model specialized in converting inputs into embedding vectors for search and similarity tasks.Intermediate
• Encoder-Decoder ArchitectureA transformer with an encoder that reads the input and a decoder that generates output while attending to the encoder.Intermediate
• Encoder-Only ArchitectureA transformer that uses only the encoder stack with bidirectional attention, suited to understanding tasks.Intermediate
• EpochOne complete pass of the training algorithm through the entire training dataset.Foundational
• Exploding GradientA training problem where gradients grow uncontrollably large, causing unstable updates and NaN losses.Intermediate
• FAISSFacebook AI Similarity Search: Meta's open-source library for efficient vector similarity search, with GPU support.Advanced
• Few-Shot PromptingIncluding several examples of inputs and desired outputs in the prompt to guide the model.Foundational
• Fine-TuningFurther training a pretrained model on a smaller task- or domain-specific dataset to adapt its behavior.Intermediate
• FlashAttentionAn IO-aware exact attention algorithm that reduces GPU memory traffic to speed up attention.Advanced
• Foundation Model (FM)A large model pretrained on broad data that can be adapted to many downstream tasks.Foundational
• FP1616-bit half-precision floating point, halving memory versus FP32 with a narrower numeric range.Intermediate
• FP3232-bit single-precision floating point, the traditional default for neural network math.Intermediate
• Full Fine-TuningUpdating all of a model's weights during fine-tuning.Intermediate
• Generative Adversarial Network (GAN)A generative model in which a generator and a discriminator are trained against each other.Intermediate
• Generative AI (GenAI)AI that creates new content such as text, images, code, audio or video based on patterns learned from training data.Foundational
• Generative Pre-trained Transformer (GPT)OpenAI's family of decoder-only transformer models pretrained on large text corpora for generation.Foundational
• Gradient AccumulationSumming gradients over several mini-batches before updating weights to simulate a larger batch.Advanced
• Gradient DescentAn optimization algorithm that iteratively adjusts parameters in the direction that most reduces the loss.Intermediate
• Graphics Processing Unit (GPU)A massively parallel processor, originally for graphics, that accelerates the matrix math behind AI training and inference.Foundational
• Hugging FaceA platform and open-source community hosting models, datasets and libraries such as Transformers.Foundational
• HumanEvalA code-generation benchmark of programming problems scored by whether generated code passes unit tests.Intermediate
• In-Context Learning (ICL)A model's ability to learn a task from instructions and examples in the prompt without weight updates.Intermediate
• InferenceUsing a trained model to make predictions or generate outputs on new data.Foundational
• Instruction TuningSupervised fine-tuning on many instruction-response examples so a model follows natural-language instructions.Intermediate
• Knowledge CutoffThe date after which a model has no information from its training data.Foundational
• Knowledge DistillationTraining a smaller student model to reproduce the outputs or behavior of a larger teacher model.Intermediate
• KV CacheStored attention keys and values for previous tokens, reused so each new token does not recompute them.Advanced
• Large Language Model (LLM)A very large transformer model trained on massive text data to understand and generate natural language and code.Foundational
• Latent SpaceA compressed, learned representation space where similar inputs map to nearby points.Intermediate
• Layer NormalizationNormalizing activations across the features of each individual example rather than across a batch.Advanced
• Learning RateThe hyperparameter that sets how large a step the optimizer takes when updating weights.Foundational
• Long Short-Term Memory (LSTM)An RNN variant with gates that control what information to keep, forget and output, enabling long-range memory.Intermediate
• Loss FunctionA function that measures how far a model's predictions are from the true values; training minimizes it.Intermediate
• Low-Rank Adaptation (LoRA)A PEFT method that freezes base weights and trains small low-rank matrices added to selected layers.Intermediate
• Machine Translation (MT)Automatically translating text or speech from one language to another.Foundational
• Masked Language Modeling (MLM)A pretraining objective in which random tokens are hidden and the model predicts them from surrounding context.Advanced
• Mixed Precision TrainingTraining with lower-precision number formats (FP16/BF16/FP8) for most operations while keeping critical values in FP32.Intermediate
• Mixture of Experts (MoE)An architecture where a router activates only a few expert sub-networks per token, increasing capacity without proportional compute.Advanced
• Model ParametersThe internal values, such as weights and biases, that a model learns from training data.Foundational
• Multi-Head Attention (MHA)Running several attention operations in parallel so the model can attend to different relationships at once.Advanced
• Multilayer Perceptron (MLP)A feedforward neural network with one or more hidden layers of fully connected neurons.Intermediate
• Named Entity Recognition (NER)An NLP task that identifies and classifies entities such as people, organizations, locations and dates in text.Foundational
• Natural Language Processing (NLP)The field of AI that enables computers to understand, interpret and generate human language.Foundational
• NeMo GuardrailsNVIDIA's open-source toolkit for adding programmable safety, topic and security rails to LLM applications.Advanced
• Neural Network (NN)A model made of layers of interconnected artificial neurons that learn weighted transformations of input data.Foundational
• Next-Token PredictionThe training objective of predicting the next token in a sequence given the preceding tokens.Intermediate
• NVIDIA NeMoNVIDIA's framework and microservices for building, customizing, evaluating and guarding generative AI models and agents.Intermediate
• NVIDIA NIMNVIDIA Inference Microservices: prebuilt, optimized containers that serve AI models through industry-standard APIs.Intermediate
• Open-Weight ModelA model whose trained weights are publicly downloadable for self-hosting and customization.Intermediate
• OptimizerThe algorithm that updates model weights using computed gradients, such as SGD, Adam or AdamW.Intermediate
• Parameter-Efficient Fine-Tuning (PEFT)Fine-tuning methods that update only a small set of added or selected parameters while freezing the base model.Intermediate
• PerplexityA language-model metric measuring how well a model predicts text; lower perplexity means better prediction.Intermediate
• Positional EncodingInformation added to token embeddings so a transformer knows the order of tokens.Advanced
• PretrainingThe initial, compute-intensive training of a model on vast unlabeled data to learn general representations.Intermediate
• Prompt TuningA PEFT method that learns continuous 'soft prompt' embeddings prepended to inputs while the model stays frozen.Advanced
• PruningRemoving less important weights, neurons or layers from a model to make it smaller and faster.Advanced
• PyTorchAn open-source deep learning framework, originally from Meta and now governed by the PyTorch Foundation, known for dynamic computation graphs.Foundational
• QLoRALoRA fine-tuning on top of a base model quantized to 4-bit precision, sharply reducing memory.Advanced
• QuantizationReducing the numeric precision of model weights (and sometimes activations) to shrink memory and speed inference.Intermediate
• Recurrent Neural Network (RNN)A neural network that processes sequences by passing a hidden state from one step to the next.Intermediate
• Reinforcement Learning (RL)Learning in which an agent takes actions in an environment and improves its policy by maximizing cumulative reward.Foundational
• Reinforcement Learning from Human Feedback (RLHF)Aligning a model by training a reward model on human preference rankings and optimizing the model against it.Intermediate
• ReLURectified Linear Unit: an activation that outputs the input if positive and zero otherwise, f(x)=max(0,x).Foundational
• Residual ConnectionA skip connection that adds a layer's input to its output, easing gradient flow in deep networks.Advanced
• Retrieval-Augmented Generation (RAG)Retrieving relevant information from external sources at query time and adding it to the prompt so the model answers from it.Foundational
• Reward ModelA model trained to score outputs according to human or AI preferences, used to guide RL fine-tuning.Advanced
• ROUGERecall-Oriented Understudy for Gisting Evaluation: n-gram and sequence overlap metrics for summarization.Intermediate
• Scaling LawsEmpirical relationships showing model loss improves predictably with more parameters, data and compute.Advanced
• Self-AttentionAttention in which each token in a sequence attends to every other token in the same sequence.Intermediate
• Self-Supervised Learning (SSL)Learning in which the training signal is generated automatically from the data itself, without human labels.Intermediate
• Sigmoid FunctionAn S-shaped activation that maps any real number to the range 0 to 1.Foundational
• Small Language Model (SLM)A language model with relatively few parameters (typically under ~15B) designed for efficiency, edge or cost-sensitive use.Intermediate
• SoftmaxA function that converts a vector of raw scores (logits) into a probability distribution summing to 1.Intermediate
• SummarizationCondensing a document into a shorter version that preserves key information.Foundational
• Supervised Fine-Tuning (SFT)Fine-tuning a model on labeled input-output pairs, such as instruction-response examples.Intermediate
• TemperatureAn inference parameter that scales randomness in token selection; low values are more deterministic, high values more creative.Foundational
• TensorA multi-dimensional array of numbers, the core data structure in deep learning frameworks.Foundational
• Tensor CoreSpecialized NVIDIA GPU units that accelerate matrix multiply-accumulate operations used in deep learning.Intermediate
• TensorRTNVIDIA's SDK for optimizing trained models into high-performance inference engines.Intermediate
• TensorRT-LLMNVIDIA's open-source library for optimized LLM inference on NVIDIA GPUs.Advanced
• TokenThe basic unit of text an LLM processes, typically a word, subword, character or punctuation mark.Foundational
• Tokenization (NLP)Splitting text into tokens that a model can convert to numeric IDs.Foundational
• TokenizerThe component that converts text to token IDs and back, defined by its vocabulary and splitting rules.Intermediate
• Top-k SamplingSampling the next token only from the k most probable candidates.Intermediate
• Top-p SamplingNucleus sampling: choose the next token from the smallest set whose cumulative probability exceeds p.Intermediate
• Transfer LearningReusing a model trained on one task as the starting point for a related task.Intermediate
• TransformerA neural network architecture based on self-attention that processes all tokens in a sequence in parallel; the basis of modern LLMs.Foundational
• Triton Inference ServerNVIDIA's open-source inference server (now Dynamo-Triton) for serving models from many frameworks on GPUs and CPUs.Intermediate
• Vanishing GradientA training problem where gradients shrink toward zero in early layers, so they learn very slowly or not at all.Intermediate
• Variational Autoencoder (VAE)A generative autoencoder that learns a probabilistic latent space from which new samples can be drawn.Advanced
• WeightsLearned numeric parameters that scale the strength of connections between neurons.Foundational
• Zero-Shot PromptingAsking a model to perform a task with instructions only and no examples.Foundational