Intro to Large Language Models

:material-circle-edit-outline: 约 70 个字

what is a Large Language Models

• LLM
  • parameters with weight
  • run program
• like chatGPT ,gemini ,their architecture are NEVER released

exp.

a LLM called llama-2-70b

it contains a parameters file with 140G - each parameter domains only 8 byte, and a run file - exactly a source code file, such C, with only approximate 500 lines

yes, just 2 files

but it performs as well as gpt3.5