From 90fad86337987f4793e7c79ed4f982b3a29793c3 Mon Sep 17 00:00:00 2001 From: SZCJW <792430652@qq.com> Date: Fri, 27 Mar 2026 01:28:45 +0800 Subject: [PATCH] x --- scripts/start_vllm.py | 19 ++++++++++++++++--- 1 file changed, 16 insertions(+), 3 deletions(-) diff --git a/scripts/start_vllm.py b/scripts/start_vllm.py index 135b674..8c849a8 100644 --- a/scripts/start_vllm.py +++ b/scripts/start_vllm.py @@ -137,8 +137,8 @@ def nuke_vllm_cache(): except Exception as e: print(f" Failed: {e}") -def configure_and_launch(model_idx, gpu_count): - print(f"DEBUG: configure_and_launch called with model_idx={model_idx}, gpu_count={gpu_count}") +def configure_and_launch(model_idx, gpu_count, use_default=False): + print(f"DEBUG: configure_and_launch called with model_idx={model_idx}, gpu_count={gpu_count}, use_default={use_default}") model_id = MODELS_TO_RUN[model_idx] config = MODEL_TABLE[model_id] print(f"DEBUG: model_id={model_id}") @@ -215,6 +215,15 @@ def configure_and_launch(model_idx, gpu_count): name = model_id.split("/")[-1] + # If use_default is True, skip interactive menu and launch directly + if use_default: + print(f"DEBUG: use_default=True, skipping interactive menu") + print(f"DEBUG: Using default config: TP={current_tp}, Seqs={current_seqs}, Ctx={current_ctx}") + # Jump directly to launch + launch_server = True + else: + launch_server = False + while True: cache_status = "YES" if clear_cache else "NO" eager_status = "YES" if use_eager else "NO" @@ -314,6 +323,10 @@ def configure_and_launch(model_idx, gpu_count): elif choice == "8": # Launch + launch_server = True + break + + if launch_server: break # Build Command @@ -385,7 +398,7 @@ def main(): try: default_idx = MODELS_TO_RUN.index(DEFAULT_MODEL) print(f"Using default model: {DEFAULT_MODEL}") - configure_and_launch(default_idx, gpu_count) + configure_and_launch(default_idx, gpu_count, use_default=True) except ValueError: print(f"Error: Default model {DEFAULT_MODEL} not found in configuration") sys.exit(1)