升级sglang vllm

Author: shell-nlp

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "gpt_server"
-version = "0.4.2"
+version = "0.4.3"
 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" }
@@ -17,15 +17,15 @@ dependencies = [
     "openai==1.55.3",
     "setuptools==75.2.0",
     "streamlit==1.39.0",
-    "torch==2.5.1",
+    "torch==2.6.0",
     "torchvision==0.20.1",
-    "vllm==0.8.4",
+    "vllm==0.8.5",
     "qwen_vl_utils",
     "evalscope[perf]==0.10.1",
     "modelscope==1.20.1",
     "edge-tts>=7.0.0",
     "funasr>=1.2.6",
-    "sglang[all]>=0.4.5",
+    "sglang[all]>=0.4.6.post1",
     "flashinfer-python",
     "flashtts>=0.1.0",
 ]
@@ -38,7 +38,7 @@ override-dependencies = [
     "torch==2.6.0",
     "triton",
     "outlines==0.1.11",
-    "transformers==4.50.0",
+    "transformers==4.51.3", #  infinity-emb
     "soundfile==0.13.1", # infinity
     "xgrammar==0.1.18", #  sglang[all]==0.4.5 depends on xgrammar==0.1.17
 ]

requirements.txt

@@ -86,7 +86,7 @@ cachetools==5.5.2
     #   evalscope
     #   streamlit
     #   vllm
-certifi==2025.1.31
+certifi==2025.4.26
     # via
     #   edge-tts
     #   httpcore
@@ -234,7 +234,7 @@ flashinfer-python==0.2.3+cu124torch2.5
     # via
     #   gpt-server (pyproject.toml)
     #   sglang
-flashtts==0.1.0
+flashtts==0.1.4
     # via gpt-server (pyproject.toml)
 flatbuffers==25.2.10
     # via onnxruntime
@@ -279,7 +279,7 @@ hf-transfer==0.1.9
     # via
     #   infinity-emb
     #   sglang
-hf-xet==1.0.4
+hf-xet==1.0.5
     # via huggingface-hub
 httpcore==1.0.9
     # via httpx
@@ -334,7 +334,7 @@ interegular==0.3.3
     #   outlines
     #   outlines-core
     #   sglang
-ipython==8.35.0
+ipython==8.36.0
     # via sglang
 jaconv==0.4.0
     # via funasr
@@ -436,8 +436,6 @@ modelscope==1.20.1
     #   evalscope
     #   funasr
     #   sglang
-monotonic==1.6
-    # via posthog
 more-itertools==10.7.0
     # via openai-whisper
 mpmath==1.3.0
@@ -456,7 +454,7 @@ multiprocess==0.70.16
     # via
     #   datasets
     #   evaluate
-narwhals==1.36.0
+narwhals==1.37.1
     # via
     #   altair
     #   plotly
@@ -630,7 +628,7 @@ opentelemetry-semantic-conventions-ai==0.4.3
     # via vllm
 optimum==1.24.0
     # via infinity-emb
-orjson==3.10.16
+orjson==3.10.17
     # via
     #   infinity-emb
     #   sglang
@@ -714,7 +712,7 @@ pooch==1.8.2
     # via librosa
 portalocker==3.1.1
     # via sacrebleu
-posthog==4.0.0
+posthog==4.0.1
     # via infinity-emb
 prometheus-client==0.21.1
     # via
@@ -757,7 +755,7 @@ pure-eval==0.2.3
     # via stack-data
 py-cpuinfo==9.0.0
     # via vllm
-pyarrow==19.0.1
+pyarrow==20.0.0
     # via
     #   datasets
     #   evalscope
@@ -954,7 +952,7 @@ sentencepiece==0.2.0
     #   mistral-common
     #   vllm
     #   xgrammar
-setproctitle==1.3.5
+setproctitle==1.3.6
     # via sglang
 setuptools==75.2.0
     # via
@@ -965,9 +963,9 @@ setuptools==75.2.0
     #   torch
     #   triton
     #   vllm
-sgl-kernel==0.0.9.post2
+sgl-kernel==0.1.0
     # via sglang
-sglang==0.4.5.post3
+sglang==0.4.6.post1
     # via gpt-server (pyproject.toml)
 shellingham==1.5.4
     # via typer
@@ -1132,7 +1130,7 @@ traitlets==5.14.3
     # via
     #   ipython
     #   matplotlib-inline
-transformers==4.50.0
+transformers==4.51.3
     # via
     #   --override (workspace)
     #   colpali-engine
@@ -1213,7 +1211,7 @@ uvloop==0.21.0
     # via
     #   sglang
     #   uvicorn
-vllm==0.8.4
+vllm==0.8.5
     # via gpt-server (pyproject.toml)
 watchdog==5.0.3
     # via streamlit