tldr/pages/linux/llama-cli.md

llama-cli

Command-line interface for LLM inference. More information: https://github.com/ggml-org/llama.cpp#llama-cli.

• Start up LLM [m]odel and interface:

llama-cli {{[-m|--model]}} {{path/to/model}}

• Start with setup prompt [c]ontext size:

llama-cli {{[-m|--model]}} {{path/to/model}} {{[-c|--ctx-size]}} {{number}}

• Start with set offload [n]umber of layer [g]pu [l]ayers:

llama-cli {{[-m|--model]}} {{path/to/model}} {{[-ngl|--gpu-layers|--n-gpu-layers]}} {{auto|all|number}}

• Start with sending content as [sys]tem prompt from a [f]ile:

llama-cli {{[-m|--model]}} {{path/to/model}} {{[-sysf|--system-prompt-file]}} {{path/to/prompt}}

• Start with sending content in [f]ile as first prompt:

llama-cli {{[-m|--model]}} {{path/to/model}} {{[-f|--file]}} {{path/to/prompt}}