← 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 71 of 200 (9,967 entries)

global memory, hardware

Main GPU memory.

global pooling, graph neural networks

Global pooling aggregates all node features into graph-level representations using operations like sum mean or attention-weighted averaging.

global variation, design & verification

Global variation affects entire die uniformly from wafer-level process differences.

globally asynchronous locally synchronous, gals, design

Hybrid timing approach.

gloo, infrastructure

PyTorch's collective communication library.

glossary,define terms,jargon,concept

I can act as a living glossary: define terms, connect them to each other, and give concrete examples.

glove box, manufacturing operations

Glove boxes provide isolated atmospheres for handling sensitive materials.

glow discharge mass spectrometry, gdms, metrology

Bulk impurity analysis.

glowtts, audio & speech

Glow-TTS combines normalizing flows with monotonic alignment for fast high-quality speech synthesis.

glu variants, glu, neural architecture

Different gated activation functions.

glu, glu, llm architecture

Gating mechanism for sequences.

glue (general language understanding evaluation),glue,general language understanding evaluation,evaluation

Suite of language understanding tasks.

glue benchmark, glue, evaluation

Standard language understanding evaluation.

glue, glue, evaluation

General Language Understanding Evaluation tests models on diverse NLP tasks.

gmf, gmf, recommendation systems

Generalized Matrix Factorization applies element-wise product followed by neural layers generalizing linear matrix factorization models.

gmlp (gated mlp),gmlp,gated mlp,llm architecture

MLP with gating for language.

gmlp for vision, computer vision

Gated MLP for images.

gmt, gmt, graph neural networks

Graph Multiset Transformer uses multiset attention and virtual nodes for expressive graph-level representation learning.

gnn expressiveness, gnn, graph neural networks

GNN expressiveness theory studies which graph properties can be distinguished by different message passing architectures.

go-explore, reinforcement learning advanced

Go-Explore maintains archive of interesting states and returns to them enabling exploration of sparse reward environments.

goal achievement, ai agents

Goal achievement detection recognizes when objectives are satisfied enabling termination.

goal stack, ai agents

Goal stacks organize objectives hierarchically tracking completion dependencies.

goal-conditioned rl, reinforcement learning

Learn to reach specified goals.

goal-conditioned rl, reinforcement learning advanced

Goal-conditioned reinforcement learning trains agents to reach diverse goals specified as part of state representation.

goat,arithmetic,fine tune

GOAT is arithmetic fine-tuned LLM. Perfect on many arithmetic tasks.

god class detection, code ai

Find classes doing too much.

gold standard,data quality

High-quality reference annotations.

gold wire bonding, packaging

Use gold wire.

golden chamber,production

Best-performing chamber used as reference.

golden wafer,metrology

Reference wafer with known good properties for calibration.

goodbye,bye,farewell

Goodbye! Feel free to come back anytime with questions about AI, chips, or LLMs.

goodness-of-fit, quality & reliability

Goodness-of-fit tests compare observed distributions to theoretical expectations.

gopher,foundation model

DeepMind's large language model.

gorilla,ai agent

LLM trained to use APIs and tools effectively.

gorilla,api,calling

Gorilla is LLM trained for API calling. Reduces hallucination on APIs.

gowning procedure, manufacturing operations

Gowning procedures specify sequence for donning cleanroom apparel.

gowning procedures, facility

Protocols for cleanroom garments.

gowning room,facility

Area where workers change into cleanroom suits before entering fab.

gpt (generative pre-trained transformer),gpt,generative pre-trained transformer,foundation model

Autoregressive model family for text generation.

gpt engineer,code,generate

GPT Engineer generates codebases from specs. Agentic coding.

gpt j,eleuther,6b

GPT-J is EleutherAI 6B model. Popular open model.

gpt neox,eleuther,20b

GPT-NeoX is EleutherAI 20B model. Open source, open weights.

gpt-4,foundation model

OpenAI's multimodal large language model.

gpt-4v (gpt-4 vision),gpt-4v,gpt-4 vision,foundation model

OpenAI's multimodal GPT-4.

gpt4all,local,desktop

GPT4All runs local models on desktop. Nomic AI. Privacy-focused.

gptq,quantization,method

GPTQ is post-training quantization method. Uses calibration data. Popular for 4-bit models.

gpu (graphics processing unit),gpu,graphics processing unit,hardware

Parallel processor widely used for deep learning.

gpu clusters for training, gpu, infrastructure

Interconnected GPUs for parallel training.

gpu direct rdma, gpu, infrastructure

Remote direct memory access for GPUs.

gpu direct, gpu, infrastructure

NVIDIA technology for direct data transfer.