fix

Author: zhhangBian

.gitignore

@@ -6,4 +6,3 @@ dist
 .idea
 .vscode
 tmp/
-ref/

lightllm/server/httpserver/manager.py

@@ -272,18 +272,12 @@ async def generate(
                 original_multimodal_params = copy.deepcopy(multimodal_params)
 
             if self.pd_mode.is_P_or_NORMAL():
-                print(f"[debug] generate verify_and_preload: {multimodal_params.to_dict()}")
                 await multimodal_params.verify_and_preload(request)
 
             # 记录请求到达的相关信息
             await self._log_req_header(request_headers, group_request_id)
             # 监控
 
-            print(
-                f"[debug] generate request: {prompt}, \
-                sampling_params: {sampling_params.to_dict()}, \
-                multimodal_params: {multimodal_params.to_dict()}"
-            )
             # 给img id
             prompt_ids = await self._encode(prompt, multimodal_params, sampling_params)
             prompt_tokens = len(prompt_ids)
@@ -402,8 +396,6 @@ async def _encode(
                 if multimodal_params.audios:
                     assert self.args.enable_multimodal_audio, "audio multimodal not enabled"
                 await self._alloc_multimodal_resources(multimodal_params, sampling_params)
-                print(f"[debug] _encode: {prompt}, multimodal_params: {multimodal_params.to_dict()}")
-                print(f"[debug] model_name: {self.args.model_name}, model_path: {self.args.model_dir}")
                 prompt_ids = self.tokenizer.encode(
                     prompt, multimodal_params, add_special_tokens=sampling_params.add_special_tokens
                 )

lightllm/server/tokenizer.py

@@ -87,7 +87,6 @@ def get_tokenizer(
     elif model_type == "qwen" and "visual" in model_cfg:
         tokenizer = QWenVLTokenizer(tokenizer, model_cfg)
     elif model_type == "mineru2_qwen":
-        print(f"[debug] use mineru2_qwen tokenizer, model_cfg: {model_cfg}")
         tokenizer = Mineru2QwenTokenizer(tokenizer, model_cfg)
     elif model_type in ["qwen2_vl", "qwen2_5_vl"] and "vision_config" in model_cfg:
         from transformers import AutoProcessor

lightllm/server/visualserver/manager.py

@@ -110,7 +110,6 @@ async def loop_for_fwd(self):
                 processing_group_reqs = []
                 images_need_infer = []
                 while len(self.waiting_reqs) > 0:
-                    print(f"[debug] loop_for_fwd waiting_reqs={len(self.waiting_reqs)}")
                     group_req_indexes = self.waiting_reqs.pop(0)
                     shm_req = self.shm_req_manager.get_req_obj_by_index(group_req_indexes.shm_req_indexes[0])
                     is_aborted = shm_req.is_aborted