修复Azure 不设置Max token的bug (#2254)

configs/__init__.py

@@ -5,4 +5,4 @@ from .server_config import *
 from .prompt_config import *
 
 
-VERSION = "v0.2.8-preview"
+VERSION = "v0.2.9-preview"

requirements.txt

@@ -5,7 +5,7 @@ langchain-experimental>=0.0.42
 pydantic==1.10.13
 fschat>=0.2.33
 xformers>=0.0.22.post7
-openai>=1.3.6
+openai>=1.3.7
 sentence_transformers
 transformers>=4.35.2
 torch==2.1.0  ##on Windows system, install the cuda version manually from https://pytorch.org/
@@ -53,7 +53,7 @@ vllm==0.2.2; sys_platform == "linux"
 
 arxiv>=2.0.0
 youtube-search>=2.1.2
-duckduckgo-search>=4.9.3
+duckduckgo-search>=3.9.9
 metaphor-python>=0.1.23
 
 # WebUI requirements

requirements_api.txt

@@ -5,7 +5,7 @@ langchain-experimental>=0.0.42
 pydantic==1.10.13
 fschat>=0.2.33
 xformers>=0.0.22.post7
-openai>=1.3.6
+openai>=1.3.7
 sentence_transformers
 transformers>=4.35.2
 torch==2.1.0  ##on Windows system, install the cuda version manually from https://pytorch.org/
@@ -37,6 +37,7 @@ pandas~=2.0.3
 einops>=0.7.0
 transformers_stream_generator==0.0.4
 vllm==0.2.2; sys_platform == "linux"
+httpx[brotli,http2,socks]>=0.25.2
 
 
 # Online api libs dependencies
@@ -53,5 +54,5 @@ vllm==0.2.2; sys_platform == "linux"
 
 arxiv>=2.0.0
 youtube-search>=2.1.2
-duckduckgo-search>=4.9.3
+duckduckgo-search>=3.9.9
 metaphor-python>=0.1.23

requirements_lite.txt

@@ -1,7 +1,7 @@
 langchain==0.0.344
 pydantic==1.10.13
 fschat>=0.2.33
-openai>=1.3.6
+openai>=1.3.7
 fastapi>=0.104.1
 python-multipart
 nltk~=3.8.1
@@ -39,14 +39,15 @@ zhipuai>=1.0.7 # zhipu
 
 numpy~=1.24.4
 pandas~=2.0.3
-streamlit~=1.28.1
+streamlit>=1.29.0
 streamlit-option-menu>=0.3.6
-streamlit-antd-components>=0.1.11
-streamlit-chatbox==1.1.11
+streamlit-antd-components>=0.2.3
+streamlit-chatbox>=1.1.11
+streamlit-modal>=0.1.0
 streamlit-aggrid>=0.3.4.post3
-httpx~=0.24.1
-watchdog
-tqdm
+httpx[brotli,http2,socks]>=0.25.2
+watchdog>=3.0.0
+tqdm>=4.66.1
 websockets
 einops>=0.7.0
 
@@ -58,5 +59,5 @@ einops>=0.7.0
 
 arxiv>=2.0.0
 youtube-search>=2.1.2
-duckduckgo-search>=4.9.3
+duckduckgo-search>=3.9.9
 metaphor-python>=0.1.23

server/model_workers/azure.py

@@ -1,4 +1,5 @@
 import sys
+import os
 from fastchat.conversation import Conversation
 from server.model_workers.base import *
 from server.utils import get_httpx_client
@@ -19,16 +20,16 @@ class AzureWorker(ApiModelWorker):
             **kwargs,
     ):
         kwargs.update(model_names=model_names, controller_addr=controller_addr, worker_addr=worker_addr)
-        kwargs.setdefault("context_len", 8000) #TODO 16K模型需要改成16384
         super().__init__(**kwargs)
         self.version = version
 
     def do_chat(self, params: ApiChatParams) -> Dict:
         params.load_config(self.model_names[0])
+
         data = dict(
             messages=params.messages,
             temperature=params.temperature,
-            max_tokens=params.max_tokens,
+            max_tokens=params.max_tokens if params.max_tokens else None,
             stream=True,
         )
         url = ("https://{}.openai.azure.com/openai/deployments/{}/chat/completions?api-version={}"
@@ -47,6 +48,7 @@ class AzureWorker(ApiModelWorker):
 
         with get_httpx_client() as client:
             with client.stream("POST", url, headers=headers, json=data) as response:
+                print(data)
                 for line in response.iter_lines():
                     if not line.strip() or "[DONE]" in line:
                         continue
@@ -60,6 +62,7 @@ class AzureWorker(ApiModelWorker):
                                     "error_code": 0,
                                     "text": text
                                 }
+                        print(text)
                     else:
                         self.logger.error(f"请求 Azure API 时发生错误：{resp}")
 

server/utils.py

@@ -43,6 +43,7 @@ def get_ChatOpenAI(
     config = get_model_worker_config(model_name)
     if model_name == "openai-api":
         model_name = config.get("model_name")
+
     model = ChatOpenAI(
         streaming=streaming,
         verbose=verbose,