From fc79606b6de80474acfc70e5da04f1ce73de7458 Mon Sep 17 00:00:00 2001 From: Wing Lian Date: Wed, 30 Apr 2025 09:11:25 -0400 Subject: [PATCH] only import vllm serve cli if its being called (#2597) [skip ci] --- src/axolotl/cli/main.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/axolotl/cli/main.py b/src/axolotl/cli/main.py index 593614733..de4fb6cbe 100644 --- a/src/axolotl/cli/main.py +++ b/src/axolotl/cli/main.py @@ -28,7 +28,6 @@ from axolotl.cli.utils import ( fetch_from_github, filter_none_kwargs, ) -from axolotl.cli.vllm_serve import do_vllm_serve from axolotl.integrations.lm_eval.cli import lm_eval from axolotl.utils import set_pytorch_cuda_alloc_conf from axolotl.utils.schemas.config import AxolotlInputConfig @@ -327,6 +326,8 @@ def fetch(directory: str, dest: Optional[str]) -> None: @add_options_from_dataclass(VllmServeCliArgs) @filter_none_kwargs def vllm_serve(config: str, **cli_args: VllmServeCliArgs): + from axolotl.cli.vllm_serve import do_vllm_serve + do_vllm_serve(config, cli_args)