hjxdl 0.0.15__py3-none-any.whl → 0.0.17__py3-none-any.whl

hdl/_version.py

@@ -12,5 +12,5 @@ __version__: str
 __version_tuple__: VERSION_TUPLE
 version_tuple: VERSION_TUPLE
 
-__version__ = version = '0.0.15'
-__version_tuple__ = version_tuple = (0, 0, 15)
+__version__ = version = '0.0.17'
+__version_tuple__ = version_tuple = (0, 0, 17)

hdl/utils/llm/chat.py

@@ -7,7 +7,7 @@ from jupyfuncs.llm.openapi import (
 )
 
 
-class GGUF_M():
+class GGUF_M(Llama):
     def __init__(
         self,
         model_path :str,
@@ -15,38 +15,43 @@ class GGUF_M():
         generation_kwargs: dict = {},
         server_ip: str = "127.0.0.1",
         server_port: int = 8000,
+        *args,
+        **kwargs
     ):
-        """Initialize the model with the provided model path and optional parameters.
+        """Initialize the model with the specified parameters.
         
         Args:
             model_path (str): The path to the model.
-            device (str, optional): The device to use for model initialization. Defaults to 'gpu'.
-            generation_kwargs (dict, optional): Additional keyword arguments for model generation. Defaults to {}.
+            device (str, optional): The device to use, either 'gpu' or 'cpu'. Defaults to 'gpu'.
+            generation_kwargs (dict, optional): Additional generation keyword arguments. Defaults to {}.
             server_ip (str, optional): The IP address of the server. Defaults to "127.0.0.1".
             server_port (int, optional): The port of the server. Defaults to 8000.
-        """ 
-        # 从本地初始化模型
-        super().__init__()
-        self.generation_kwargs = generation_kwargs
-        self.questions = []
-        self.resps = []
+            *args: Variable length argument list.
+            **kwargs: Arbitrary keyword arguments.
+        
+        Raises:
+            KeyError: If 'num_threads' or 'max_context_length' is missing in generation_kwargs.
+        """
         print("正在从本地加载模型...")
-        if device == 'cpu':
-            self.model = Llama(
+        if device.lower() == 'cpu': 
+            super().__init__(
                 model_path=model_path,
-                n_threads=self.generation_kwargs['num_threads'],
-                n_ctx=self.generation_kwargs['max_context_length'],
+                n_threads=generation_kwargs['num_threads'],
+                n_ctx=generation_kwargs['max_context_length'],
+                *args,
+                **kwargs
             )
         else:
-            self.model = Llama(
+            super().__init__(
                 model_path=model_path,
-                n_threads=self.generation_kwargs['num_threads'],
-                n_ctx=self.generation_kwargs['max_context_length'],
+                n_threads=generation_kwargs['num_threads'],
+                n_ctx=generation_kwargs['max_context_length'],
                 n_gpu_layers=-1,
-                flash_attn=True
+                flash_attn=True,
+                *args,
+                **kwargs
             )
-            
-        print("完成本地模型的加载")
+        self.generation_kwargs = generation_kwargs
 
     def invoke(
         self, 
@@ -67,7 +72,7 @@ class GGUF_M():
         """
         prompt_final = f"USER:\n{prompt}\nASSISTANT:\n" 
 
-        result = self.model.create_completion(
+        result = self.create_completion(
             prompt_final,
             repeat_penalty=self.generation_kwargs["repetition_penalty"],
             max_tokens=self.generation_kwargs["max_new_tokens"],
@@ -83,7 +88,7 @@ class GGUF_M():
         #     [prompt, resp]
         # )
         return resp
- 
+
     def stream(
         self,
         prompt: str,
@@ -101,9 +106,8 @@ class GGUF_M():
         Yields:
             str: Text responses generated by the model based on the prompt.
         """
-        self.questions.append(prompt)
         prompt = f"USER:\n{prompt}\nASSISTANT:\n" 
-        output = self.model.create_completion(
+        output = self.create_completion(
             prompt,
             stream=True,
             repeat_penalty=self.generation_kwargs["repetition_penalty"], 
@@ -123,6 +127,119 @@ class GGUF_M():
         # self.resps[-1] = "".join(self.resps[-1])
 
 
+# class GGUF_M():
+#     def __init__(
+#         self,
+#         model_path :str,
+#         device: str='gpu',
+#         generation_kwargs: dict = {},
+#         server_ip: str = "127.0.0.1",
+#         server_port: int = 8000,
+#     ):
+#         """Initialize the model with the provided model path and optional parameters.
+        
+#         Args:
+#             model_path (str): The path to the model.
+#             device (str, optional): The device to use for model initialization. Defaults to 'gpu'.
+#             generation_kwargs (dict, optional): Additional keyword arguments for model generation. Defaults to {}.
+#             server_ip (str, optional): The IP address of the server. Defaults to "127.0.0.1".
+#             server_port (int, optional): The port of the server. Defaults to 8000.
+#         """ 
+#         # 从本地初始化模型
+#         # super().__init__()
+#         self.generation_kwargs = generation_kwargs
+#         print("正在从本地加载模型...")
+#         if device == 'cpu':
+#             self.model = Llama(
+#                 model_path=model_path,
+#                 n_threads=self.generation_kwargs['num_threads'],
+#                 n_ctx=self.generation_kwargs['max_context_length'],
+#             )
+#         else:
+#             self.model = Llama(
+#                 model_path=model_path,
+#                 n_threads=self.generation_kwargs['num_threads'],
+#                 n_ctx=self.generation_kwargs['max_context_length'],
+#                 n_gpu_layers=-1,
+#                 flash_attn=True
+#             )
+            
+#         print("完成本地模型的加载")
+
+#     def invoke(
+#         self, 
+#         prompt : str,
+#         stop: list[str] | None = ["USER:", "ASSISTANT:"],
+#         # history: list = [],
+#         **kwargs: t.Any,
+#     ) -> str:
+#         """Invoke the model to generate a response based on the given prompt.
+        
+#         Args:
+#             prompt (str): The prompt to be used for generating the response.
+#             stop (list[str], optional): List of strings that indicate when the model should stop generating the response. Defaults to ["USER:", "ASSISTANT:"].
+#             **kwargs: Additional keyword arguments to be passed to the model.
+        
+#         Returns:
+#             str: The generated response based on the prompt.
+#         """
+#         prompt_final = f"USER:\n{prompt}\nASSISTANT:\n" 
+
+#         result = self.model.create_completion(
+#             prompt_final,
+#             repeat_penalty=self.generation_kwargs["repetition_penalty"],
+#             max_tokens=self.generation_kwargs["max_new_tokens"],
+#             stop=stop,
+#             echo=False,
+#             temperature=self.generation_kwargs["temperature"],
+#             mirostat_mode = 2,
+#             mirostat_tau=4.0,
+#             mirostat_eta=1.1
+#         )
+#         resp = result['choices'][0]['text']
+#         # history.append(
+#         #     [prompt, resp]
+#         # )
+#         return resp
+ 
+#     def stream(
+#         self,
+#         prompt: str,
+#         stop: list[str] | None = ["USER:", "ASSISTANT:"],
+#         # history: list = [],
+#         **kwargs: t.Any,
+#     ):
+#         """Generate text responses based on the given prompt using the model.
+        
+#         Args:
+#             prompt (str): The prompt to generate text responses.
+#             stop (list[str], optional): List of strings to stop the generation. Defaults to ["USER:", "ASSISTANT:"].
+#             **kwargs: Additional keyword arguments for the model.
+        
+#         Yields:
+#             str: Text responses generated by the model based on the prompt.
+#         """
+#         prompt = f"USER:\n{prompt}\nASSISTANT:\n" 
+#         output = self.model.create_completion(
+#             prompt,
+#             stream=True,
+#             repeat_penalty=self.generation_kwargs["repetition_penalty"], 
+#             max_tokens=self.generation_kwargs["max_new_tokens"],
+#             stop=stop,
+#             echo=False,
+#             temperature=self.generation_kwargs["temperature"],
+#             mirostat_mode = 2,
+#             mirostat_tau=4.0,
+#             mirostat_eta=1.1
+#         )
+#         # history.append([])
+#         for chunk in output:
+#             item = chunk['choices'][0]['text']
+#             # self.resps[-1].append(item)
+#             yield chunk['choices'][0]['text']
+#         # self.resps[-1] = "".join(self.resps[-1])
+
+
 class OpenAI_M():
     def __init__(
         self,

{hjxdl-0.0.15.dist-info → hjxdl-0.0.17.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: hjxdl
-Version: 0.0.15
+Version: 0.0.17
 Summary: A collection of functions for Jupyter notebooks
 Home-page: https://github.com/huluxiaohuowa/hdl
 Author: Jianxing Hu

{hjxdl-0.0.15.dist-info → hjxdl-0.0.17.dist-info}/RECORD

@@ -1,5 +1,5 @@
 hdl/__init__.py,sha256=5sZZNySv08wwfzJcSDssGTqUn9wlmDsR6R4XB8J8mFM,70
-hdl/_version.py,sha256=pzShTmESM-5FMm5QD9BQxtLcs8dS3tIY17WJZlF9dSQ,413
+hdl/_version.py,sha256=OXG6jb6QQz55miFHOqfJx8zqIQYlPcgtvsnb6XEE5Fs,413
 hdl/args/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 hdl/args/loss_args.py,sha256=s7YzSdd7IjD24rZvvOrxLLFqMZQb9YylxKeyelSdrTk,70
 hdl/controllers/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -84,11 +84,11 @@ hdl/utils/database_tools/connect.py,sha256=KUnVG-8raifEJ_N0b3c8LkTTIfn9NIyw8LX6q
 hdl/utils/general/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 hdl/utils/general/glob.py,sha256=8-RCnt6L297wMIfn34ZAMCsGCZUjHG3MGglGZI1cX0g,491
 hdl/utils/llm/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-hdl/utils/llm/chat.py,sha256=NEbWdrPpQqS5ArU7ghA6xuam8wwvWXk3LL_CiNEEuIg,7055
+hdl/utils/llm/chat.py,sha256=H2c8assJlSdZQKIfPkYrVZHqv66TsdsxtaLXv0kNe1w,11565
 hdl/utils/llm/embs.py,sha256=yCFtc25gUFas6kwgOGBFydeaHNyQMq5y1Chxl8TNEUQ,2190
 hdl/utils/schedulers/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 hdl/utils/schedulers/norm_lr.py,sha256=bDwCmdEK-WkgxQMFBiMuchv8Mm7C0-GZJ6usm-PQk14,4461
-hjxdl-0.0.15.dist-info/METADATA,sha256=uVY_O0levGtYTHXWtCIstsHzulDmM4JLwQ7Xu6uZ5sg,543
-hjxdl-0.0.15.dist-info/WHEEL,sha256=Z4pYXqR_rTB7OWNDYFOm1qRk0RX6GFP2o8LgvP453Hk,91
-hjxdl-0.0.15.dist-info/top_level.txt,sha256=-kxwTM5JPhylp06z3zAVO3w6_h7wtBfBo2zgM6YZoTk,4
-hjxdl-0.0.15.dist-info/RECORD,,
+hjxdl-0.0.17.dist-info/METADATA,sha256=MBuoruFG8xTqFc-epdvuhZWoQwbpo28-fTfi7QJPijc,543
+hjxdl-0.0.17.dist-info/WHEEL,sha256=Z4pYXqR_rTB7OWNDYFOm1qRk0RX6GFP2o8LgvP453Hk,91
+hjxdl-0.0.17.dist-info/top_level.txt,sha256=-kxwTM5JPhylp06z3zAVO3w6_h7wtBfBo2zgM6YZoTk,4
+hjxdl-0.0.17.dist-info/RECORD,,