ChatOpenAI为了判断token有没有超过模型的context上下文长度，每个模型的token算法不一样 ，所以这里应该自己实现token长度计算

第一次初始化的时候，openai的类会强制使用3.5，

server/minx_chat_openai.py

@@ -0,0 +1,51 @@
+from typing import (
+    TYPE_CHECKING,
+    Any,
+    Tuple
+)
+import sys
+import logging
+
+logger = logging.getLogger(__name__)
+
+if TYPE_CHECKING:
+    import tiktoken
+
+
+class MinxChatOpenAI:
+
+    @staticmethod
+    def import_tiktoken() -> Any:
+        try:
+            import tiktoken
+        except ImportError:
+            raise ValueError(
+                "Could not import tiktoken python package. "
+                "This is needed in order to calculate get_token_ids. "
+                "Please install it with `pip install tiktoken`."
+            )
+        return tiktoken
+
+    @staticmethod
+    def get_encoding_model(self) -> Tuple[str, "tiktoken.Encoding"]:
+        tiktoken_ = MinxChatOpenAI.import_tiktoken()
+        if self.tiktoken_model_name is not None:
+            model = self.tiktoken_model_name
+        else:
+            model = self.model_name
+            if model == "gpt-3.5-turbo":
+                # gpt-3.5-turbo may change over time.
+                # Returning num tokens assuming gpt-3.5-turbo-0301.
+                model = "gpt-3.5-turbo-0301"
+            elif model == "gpt-4":
+                # gpt-4 may change over time.
+                # Returning num tokens assuming gpt-4-0314.
+                model = "gpt-4-0314"
+        # Returns the number of tokens used by a list of messages.
+        try:
+            encoding = tiktoken_.encoding_for_model(model)
+        except Exception as e:
+            logger.warning("Warning: model not found. Using cl100k_base encoding.")
+            model = "cl100k_base"
+            encoding = tiktoken_.get_encoding(model)
+        return model, encoding

server/utils.py

@@ -12,10 +12,23 @@ from concurrent.futures import ThreadPoolExecutor, as_completed
 from langchain.chat_models import ChatOpenAI
 from langchain.llms import OpenAI
 import httpx
-from typing import Literal, Optional, Callable, Generator, Dict, Any, Awaitable, Union, Tuple
+from typing import (
+    TYPE_CHECKING,
+    Literal,
+    Optional,
+    Callable,
+    Generator,
+    Dict,
+    Any,
+    Awaitable,
+    Union,
+    Tuple
+)
 import logging
 import torch
 
+from server.minx_chat_openai import MinxChatOpenAI
+
 
 async def wrap_done(fn: Awaitable, event: asyncio.Event):
     """Wrap an awaitable with a event to signal when it's done or an exception is raised."""
@@ -44,7 +57,7 @@ def get_ChatOpenAI(
     config = get_model_worker_config(model_name)
     if model_name == "openai-api":
         model_name = config.get("model_name")
-
+    ChatOpenAI._get_encoding_model = MinxChatOpenAI.get_encoding_model
     model = ChatOpenAI(
         streaming=streaming,
         verbose=verbose,
@@ -153,6 +166,7 @@ class ChatMessage(BaseModel):
 
 def torch_gc():
     try:
+        import torch
         if torch.cuda.is_available():
             # with torch.cuda.device(DEVICE):
             torch.cuda.empty_cache()
@@ -500,58 +514,29 @@ def set_httpx_config(
     # 自动检查torch可用的设备。分布式部署时，不运行LLM的机器上可以不装torch
 
 
-def is_mps_available():
-    return hasattr(torch.backends, "mps") and torch.backends.mps.is_available()
-
-
-def is_cuda_available():
-    return torch.cuda.is_available()
-
-
 def detect_device() -> Literal["cuda", "mps", "cpu"]:
     try:
+        import torch
         if torch.cuda.is_available():
             return "cuda"
-        if is_mps_available():
+        if torch.backends.mps.is_available():
             return "mps"
     except:
         pass
     return "cpu"
 
 
-def llm_device(device: str = None) -> Literal["cuda", "mps", "cpu", "xpu"]:
+def llm_device(device: str = None) -> Literal["cuda", "mps", "cpu"]:
     device = device or LLM_DEVICE
-    if device not in ["cuda", "mps", "cpu", "xpu"]:
-        logging.warning(f"device not in ['cuda', 'mps', 'cpu','xpu'], device = {device}")
+    if device not in ["cuda", "mps", "cpu"]:
         device = detect_device()
-    elif device == 'cuda' and not is_cuda_available() and is_mps_available():
-        logging.warning("cuda is not available, fallback to mps")
-        return "mps"
-    if device == 'mps' and not is_mps_available() and is_cuda_available():
-        logging.warning("mps is not available, fallback to cuda")
-        return "cuda"
-
-    # auto detect device if not specified
-    if device not in ["cuda", "mps", "cpu", "xpu"]:
-        return detect_device()
     return device
 
 
-def embedding_device(device: str = None) -> Literal["cuda", "mps", "cpu", "xpu"]:
-    device = device or LLM_DEVICE
+def embedding_device(device: str = None) -> Literal["cuda", "mps", "cpu"]:
+    device = device or EMBEDDING_DEVICE
     if device not in ["cuda", "mps", "cpu"]:
-        logging.warning(f"device not in ['cuda', 'mps', 'cpu','xpu'], device = {device}")
         device = detect_device()
-    elif device == 'cuda' and not is_cuda_available() and is_mps_available():
-        logging.warning("cuda is not available, fallback to mps")
-        return "mps"
-    if device == 'mps' and not is_mps_available() and is_cuda_available():
-        logging.warning("mps is not available, fallback to cuda")
-        return "cuda"
-
-    # auto detect device if not specified
-    if device not in ["cuda", "mps", "cpu"]:
-        return detect_device()
     return device