← Back to AI Factory Chat

AI Factory Glossary

9,967 technical terms and definitions

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z All
Showing page 75 of 200 (9,967 entries)

grouped convolution, model optimization

Grouped convolutions partition input channels into groups processing each independently reducing parameters.

grouped query attention,gqa,kv

GQA shares key-value heads across query heads. Reduces KV cache size. Llama 2 70B uses GQA.

grouped-query attention (gqa),grouped-query attention,gqa,llm architecture

Middle ground between MQA and multi-head attention.

grouped-query kv cache, optimization

Share K/V within groups.

groupnorm, neural architecture

Normalize within groups of channels.

grover's algorithm, quantum ai

Quantum search algorithm.

growth,adoption,engagement

Drive AI adoption: onboarding, education, quick wins. Measure engagement and retention. Iterate on UX.

grpc,rpc,streaming

gRPC is high-performance RPC framework. HTTP/2, protobuf. Streaming support.

gru4rec, recommendation systems

GRU4Rec applies gated recurrent units to session-based recommendation by modeling sequential user behavior within sessions.

gscan, evaluation

Grounded compositional generalization.

gsm8k, gsm8k, evaluation

Math word problems.

gsm8k, gsm8k, evaluation

Grade School Math 8K tests mathematical reasoning on word problems.

gsm8k,evaluation

Grade school math word problems.

gtn, gtn, graph neural networks

Graph Transformer Networks learn new graph structures through soft edge selection for heterogeneous graphs.

guanaco,qlora,efficient

Guanaco is QLoRA fine-tuned model. Efficient training on single GPU.

guard band,design

Safety margin in specifications.

guard ring effectiveness,design

Measure latchup protection.

guard rings,design

Structures to prevent latch-up.

guardbanding, advanced test & probe

Guardbanding tightens test limits beyond specifications to reduce test escapes and improve shipped quality.

guardbanding, design

Add margin to account for variation.

guardrails ai,framework

Framework for adding structure validation and safety to LLM outputs.

guardrails, ai safety

Guardrails constrain model behavior preventing specific undesired outputs.

guardrails,boundary,limit

Guardrails prevent unwanted model behavior. Topic restrictions, format requirements, safety filters.

guidance scale, generative models

Strength of conditioning guidance.

guidance scale, multimodal ai

Guidance scale controls trade-off between prompt adherence and sample diversity in guided generation.

guidance,framework

Language for controlling generation from language models.

guidance,structured,microsoft

Guidance controls LLM output structure. Constrained generation. Microsoft.

guided backprop, interpretability

Guided backpropagation modifies gradient flow to visualize features detected by neural networks.

guided backpropagation, explainable ai

Modified backprop for visualization.

gull-wing leads, packaging

Leads bent outward and down.

h-gate,design

Body contact on both sides.

h-tree, design & verification

H-tree topology creates balanced clock distribution through recursive H-shaped branching.

h-tree,design

Symmetric tree structure for clock.

h100,a100,datacenter gpu

A100/H100 are NVIDIA data center GPUs for AI. Huge VRAM + fast HBM + strong Tensor Cores = ideal for large LLM training and high-throughput inference.

h2o cache, h2o, optimization

Keep most important KV pairs.

h3 (hungry hungry hippos),h3,hungry hungry hippos,llm architecture

Hybrid SSM+attention architecture.

half-pitch,lithography

Half the pitch used as resolution metric.

halide, model optimization

Halide separates algorithm from schedule enabling portable high-performance image processing.

hall effect measurement, metrology

Determine carrier concentration and mobility.

hallucination detection, ai safety

Identify false statements.

hallucination in llms, challenges

Generating false information.

hallucination, evaluation

Hallucination generates plausible-sounding but false or unsupported information.

hallucination,confabulation,grounding

Hallucination: model generates false info confidently. Mitigate with RAG, grounding, citations, and temperature 0.

halo implant, process integration

Halo implants are angled pocket implants placed asymmetrically to suppress drain-induced barrier lowering.

halo implant,process

Angled implant to reduce short-channel effects.

halstead metrics, code ai

Software complexity measures.

halt (highly accelerated life test),halt,highly accelerated life test,reliability

Extreme stress test to find failure modes.

halt vs hass, halt, reliability

Discovery vs screening.

halt, halt, business & standards

Highly Accelerated Life Test discovers design weaknesses through extreme combined stresses.

ham, ham, reinforcement learning

Hierarchical RL framework.