Skip to main content

Module run

Module run 

Source
Expand description

Run command - Interactive chat with a model (ollama-style)

Structs§

RunCommand

Enums§

OutputFormat
Output format for ferrum run. JSONL mode emits one record per event (assistant generation result, user input, exit) on stdout — used by integration tests and scripting. Text mode is the default interactive UX.

Functions§

apply_kv_dtype_override
Apply the resolved --kv-dtype / runtime-config override to an engine config, validating early. Default is FP16 (the production-validated path on every backend); selecting INT8 / FP8 is rejected with a helpful message until model integration ships.
execute
select_device