diff --git a/stacks/llm/docker-compose.yml b/stacks/llm/docker-compose.yml index 6b17f6d..743e5b5 100644 --- a/stacks/llm/docker-compose.yml +++ b/stacks/llm/docker-compose.yml @@ -35,13 +35,19 @@ services: - "-ngl" - "99" - "--ctx-size" + - "65536" + - "--rope-scaling" + - "yarn" + - "--rope-scale" + - "2" + - "--yarn-orig-ctx" - "32768" - "--flash-attn" - "on" - "--cache-type-k" - "q8_0" - "--cache-type-v" - - "q8_0" + - "q4_0" - "--host" - "0.0.0.0" - "--port"