Packages

Native Erlang/OTP wrapper around llama.cpp with token-exact tiered KV cache; cache more warm state than fits in RAM

Retired package: Renamed - Renamed to barrel_inference: https://github.com/barrel-platform/barrel_inference

Current section

Files

Jump to
erllama config sys.config
Raw

config/sys.config

[
{erllama, [
{min_tokens, 512},
{cold_min_tokens, 512},
{cold_max_tokens, 30000},
{continued_interval, 2048},
{boundary_trim_tokens, 32},
{boundary_align_tokens, 2048},
{evict_save_timeout_ms, 30000},
{session_resume_wait_ms, 500},
{fingerprint_mode, safe}
]}
].