AI term
Decoding FLOPs
What is Decoding FLOPs?
Floating-point operations used while a model generates its output.
In other languages
- 한국어디코딩 부동소수점 연산량
- 출력을 생성하는 디코딩 과정에서 수행되는 부동소수점 연산의 양입니다.
- 日本語デコード時の浮動小数点演算数
- モデルが出力を生成する際に行う浮動小数点演算の量。
Related Terms
- FLOPsFloating-point operations per second, a measure of an AI system's computational power and processing scale
- Model FLOP utilizationThe share of a system’s theoretical floating-point computing capacity that a model training run actually uses.
- Model FLOPs UtilizationA measure of how much of a system’s theoretical floating-point computing capacity is used during model computation.
- Peak FLOPSMaximum floating-point operations per second a computing system can theoretically perform under ideal conditions
- FP8An 8-bit floating-point format that allows models to train and run faster with lower memory requirements.
- Gflop/sA unit of computing speed measuring one billion floating-point operations per second
- MXFP4Microscaling 4-bit floating-point format designed to reduce memory use while preserving numerical quality in large model inference
- FP8 quantizationA technique that reduces the precision of model weights and activations to 8-bit floating point format to decrease memory usage and accelerate inference
- FP4A 4-bit floating-point number format used for highly compressed low-precision AI computation
- Floating-point operationAn arithmetic calculation involving numbers with fractional parts, used as a fundamental measure of computational effort in computer science.
The fastest, most reliable AI news, straight to your inbox