升级 lmdeploy==0.9.1 vllm==0.9.2 sglang[all]>=0.4.9.post3 transformers==4.53.3 xgrammar==0.1.21

Author: shell-nlp

gpt_server/model_worker/embedding_infinity.py

@@ -48,9 +48,10 @@ def __init__(
             device = "cuda"
         logger.warning(f"使用{device}加载...")
         model_type = getattr(self.model_config, "model_type", None)
-        bettertransformer = True
-        if model_type is not None and "deberta" in model_type:
-            bettertransformer = False
+        bettertransformer = False
+        # TODO bettertransformer = True  transformer 出问题
+        # if model_type is not None and "deberta" in model_type:
+        #     bettertransformer = False
         engine_args = EngineArgs(
             model_name_or_path=model_path,
             engine="torch",

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "gpt_server"
-version = "0.5.3"
+version = "0.5.4"
 description = "gpt_server是一个用于生产级部署LLMs或Embedding的开源框架。"
 readme = "README.md"
 license = { text = "Apache 2.0" }
@@ -12,20 +12,20 @@ dependencies = [
     "ffmpy",
     "fschat==0.2.36",
     "infinity-emb[all]==0.0.76",
-    "lmdeploy==0.9.0",
+    "lmdeploy==0.9.1",
     "loguru>=0.7.2",
     "openai==1.86.0",
     "setuptools==75.2.0",
     "streamlit==1.39.0",
     "torch==2.6.0",
     "torchvision==0.20.1",
-    "vllm==0.9.1",
+    "vllm==0.9.2",
     "qwen_vl_utils",
     "evalscope[perf,rag]==0.16.1",
     "modelscope==1.26.0",
     "edge-tts>=7.0.0",
     "funasr>=1.2.6",
-    "sglang[all]>=0.4.8.post1",
+    "sglang[all]>=0.4.9.post3",
     "flashinfer-python",
     "flashtts>=0.1.7",
     "diffusers>=0.33.1",
@@ -39,10 +39,9 @@ override-dependencies = [
     "torchaudio==2.7.1",
     "torch==2.7.0",
     "triton",
-    "outlines==0.1.11",
-    "transformers==4.51.3", #  infinity-emb
+    "transformers==4.53.3", #  infinity-emb
     "soundfile==0.13.1", # infinity
-    "xgrammar==0.1.19", #  sglang[all]==0.4.5 depends on xgrammar==0.1.17
+    "xgrammar==0.1.21", #  sglang[all]==0.4.5 depends on xgrammar==0.1.17
 ]
 
 [project.scripts]