From 67a2c5bbb7cc545e5c73c25bc105a2d91ff0b10c Mon Sep 17 00:00:00 2001 From: LittleMouse Date: Tue, 18 Nov 2025 15:30:57 +0800 Subject: [PATCH] [update] support qwen3-vl --- backend/llm_client_backend.py | 3 ++- services/model_list.py | 3 +++ 2 files changed, 5 insertions(+), 1 deletion(-) diff --git a/backend/llm_client_backend.py b/backend/llm_client_backend.py index 9d6a218..6b89d42 100644 --- a/backend/llm_client_backend.py +++ b/backend/llm_client_backend.py @@ -106,7 +106,8 @@ class LlmClientBackend(BaseModelBackend): "max_token_len": request.max_tokens, "temperature": request.temperature, "top_p": request.top_p, - "prompt": parsed_prompt + "prompt": parsed_prompt, + "b_video": True } ) ) diff --git a/services/model_list.py b/services/model_list.py index 96178be..ac86139 100644 --- a/services/model_list.py +++ b/services/model_list.py @@ -93,6 +93,9 @@ class GetModelList: if precompute_len is not None: new_entry['max_context_length'] = precompute_len + if '-vl-' in mode: + new_entry['max_context_length'] = 1152 + elif model_type == 'tts': mode_param = model_data.get("mode_param", {}) precompute_len = None