hAIgopala › AI terms › Rapid Fire (3 terms per short) › Day 46

Compute / FLOPs

Infrastructure & HardwareChapter 13: The Engine RoomTerm 225 of 253

Amount of computation used for training or inference.

At Sunrise School

Gopal asks

How much compute did training take?

Bit answers

It's measured in FLOPs or GPU-hours. Our classifier: 2 GPU-hours. A frontier LLM: millions.

In the real world

Training cost measured in GPU-hours.

Watch the short

Short coming soonSeason 5 · Day 46Follow @haigopala

Quick check

Quick check. Does a KV cache make long replies faster? Yes or no?

Taught together with

Distributed Training · KV Cache

Practise it

Try this term and all 253 others in the free Model Training Arena.

Open the Arena →

Follow Bit daily: YouTube English · తెలుగు · हिंदी  |  Instagram English · తెలుగు · हिंदी