[NeuralChat] Sync RESTful API with latest OpenAI protocol (#1205)

intel_extension_for_transformers/neural_chat/chatbot.py

@@ -115,25 +115,25 @@ def build_chatbot(config: PipelineConfig=None):
     # create model adapter
     if "llama" in config.model_name_or_path.lower():
         from .models.llama_model import LlamaModel
-        adapter = LlamaModel()
+        adapter = LlamaModel(config.model_name_or_path, config.task)
     elif "mpt" in config.model_name_or_path.lower():
         from .models.mpt_model import MptModel
-        adapter = MptModel()
+        adapter = MptModel(config.model_name_or_path, config.task)
     elif "neural-chat" in config.model_name_or_path.lower():
         from .models.neuralchat_model import NeuralChatModel
-        adapter = NeuralChatModel()
+        adapter = NeuralChatModel(config.model_name_or_path, config.task)
     elif "chatglm" in config.model_name_or_path.lower():
         from .models.chatglm_model import ChatGlmModel
-        adapter = ChatGlmModel()
+        adapter = ChatGlmModel(config.model_name_or_path, config.task)
     elif "qwen" in config.model_name_or_path.lower():
         from .models.qwen_model import QwenModel
-        adapter = QwenModel()
+        adapter = QwenModel(config.model_name_or_path, config.task)
     elif "mistral" in config.model_name_or_path.lower():
         from .models.mistral_model import MistralModel
-        adapter = MistralModel()
+        adapter = MistralModel(config.model_name_or_path, config.task)
     elif "solar" in config.model_name_or_path.lower():
         from .models.solar_model import SolarModel
-        adapter = SolarModel()
+        adapter = SolarModel(config.model_name_or_path, config.task)
     elif "opt" in config.model_name_or_path.lower() or \
          "gpt" in config.model_name_or_path.lower() or \
          "flan-t5" in config.model_name_or_path.lower() or \
@@ -144,12 +144,13 @@ def build_chatbot(config: PipelineConfig=None):
          "mixtral" in config.model_name_or_path.lower() or \
          "phi-2" in config.model_name_or_path.lower():
         from .models.base_model import BaseModel
-        adapter = BaseModel()
+        adapter = BaseModel(config.model_name_or_path, config.task)
     else:
         set_latest_error(ErrorCodes.ERROR_MODEL_NOT_SUPPORTED)
         logging.error("build_chatbot: unknown model")
         return
-
+    from .models.base_model import register_model_adapter
+    register_model_adapter(adapter)
     # register plugin instance in model adaptor
     if config.plugins:
         for plugin_name, plugin_value in config.plugins.items():

intel_extension_for_transformers/neural_chat/config.py

@@ -450,6 +450,7 @@ def __init__(self,
                  tokenizer_name_or_path=None,
                  hf_access_token=None,
                  device="auto",
+                 task="",
                  plugins=plugins,
                  loading_config=None,
                  optimization_config=None,
@@ -462,7 +463,7 @@ def __init__(self,
             self.device = get_device_type()
         else:
             self.device = device
-
+        self.task = task
         self.plugins = plugins
 
         self.loading_config = loading_config if loading_config is not None else \

intel_extension_for_transformers/neural_chat/models/base_model.py

@@ -63,11 +63,11 @@ class BaseModel(ABC):
     A base class for LLM.
     """
 
-    def __init__(self):
+    def __init__(self, model_name, task="chat"):
         """
         Initializes the BaseModel class.
         """
-        self.model_name = ""
+        self.model_name = model_name
         self.asr = None
         self.tts = None
         self.face_animation = None
@@ -81,16 +81,14 @@ def __init__(self):
         self.device = None
         self.conv_template = None
         self.ipex_int8 = None
+        self.get_conv_template(task)
 
-    def match(self, model_path: str):
+    def match(self):
         """
-        Check if the provided model_path matches the current model.
-
-        Args:
-            model_path (str): Path to a model.
+        Check if the provided model_name matches the current model.
 
         Returns:
-            bool: True if the model_path matches, False otherwise.
+            bool: True if the model_name matches, False otherwise.
         """
         return True
 
@@ -167,7 +165,6 @@ def predict_stream(self, query, origin_query="", config=None):
                 raise ValueError(f"The audio file path {query} is invalid.")
 
         query_include_prompt = False
-        self.get_conv_template(self.model_name, config.task)
         if (self.conv_template.roles[0] in query and self.conv_template.roles[1] in query) or \
               "starcoder" in self.model_name.lower() or "codellama" in self.model_name.lower() or \
               "codegen" in self.model_name.lower() or "magicoder" in self.model_name.lower() or \
@@ -219,7 +216,7 @@ def predict_stream(self, query, origin_query="", config=None):
         assert query is not None, "Query cannot be None."
 
         if not query_include_prompt and not is_plugin_enabled("retrieval"):
-            query = self.prepare_prompt(query, self.model_name, config.task)
+            query = self.prepare_prompt(query, config.task)
 
         # Phind/Phind-CodeLlama-34B-v2 model accpects Alpaca/Vicuna instruction format.
         if "phind" in self.model_name.lower():
@@ -278,7 +275,6 @@ def predict(self, query, origin_query="", config=None):
                 raise ValueError(f"The audio file path {query} is invalid.")
 
         query_include_prompt = False
-        self.get_conv_template(self.model_name, config.task)
         if (self.conv_template.roles[0] in query and self.conv_template.roles[1] in query) or \
                "starcoder" in self.model_name.lower() or "codellama" in self.model_name.lower() or \
                "codegen" in self.model_name.lower() or "magicoder" in self.model_name.lower():
@@ -326,7 +322,7 @@ def predict(self, query, origin_query="", config=None):
 
         if not query_include_prompt and not is_plugin_enabled("retrieval") \
             and not 'vllm' in str(MODELS[self.model_name]['model']):
-            query = self.prepare_prompt(query, self.model_name, config.task)
+            query = self.prepare_prompt(query, config.task)
 
         # Phind/Phind-CodeLlama-34B-v2 model accpects Alpaca/Vicuna instruction format.
         if "phind" in self.model_name.lower():
@@ -406,7 +402,7 @@ def face_animate(self, image_path, audio_path=None, text=None, voice=None) -> st
             os.remove(audio_path)
         return video_path
 
-    def get_default_conv_template(self, model_path: str) -> Conversation:
+    def get_default_conv_template(self) -> Conversation:
         """
         Get the default conversation template for the given model path.
 
@@ -432,22 +428,22 @@ def get_conv_template(self, model_path: str, task: str = "") -> Conversation:
         if self.conv_template:
             return
         if not task:
-            self.conv_template = PromptTemplate(self.get_default_conv_template(model_path).name, clear_history=True)
+            self.conv_template = PromptTemplate(self.get_default_conv_template().name, clear_history=True)
         else:
             clear_history = True
             if task == "completion":
                 name = "alpaca_without_input"
             elif task == "chat":
-                name = "neural-chat-7b-v2"
                 clear_history = False
+                name = self.get_default_conv_template().name
             elif task == "summarization":
                 name = "summarization"
             else:
                 raise NotImplementedError(f"Unsupported task {task}.")
             self.conv_template = PromptTemplate(name, clear_history=clear_history)
 
-    def prepare_prompt(self, prompt: str, model_path: str, task: str = ""):
-        self.get_conv_template(model_path, task)
+    def prepare_prompt(self, prompt: str, task: str = ""):
+        self.get_conv_template(task)
         self.conv_template.append_message(self.conv_template.roles[0], prompt)
         self.conv_template.append_message(self.conv_template.roles[1], None)
         return self.conv_template.get_prompt()
@@ -485,15 +481,15 @@ def register_plugin_instance(self, plugin_name, instance):
 
 def register_model_adapter(cls):
     """Register a model adapter."""
-    model_adapters.append(cls())
+    model_adapters.append(cls)
 
 
 def get_model_adapter(model_name_path: str) -> BaseModel:
     """Get a model adapter for a model_name_path."""
     model_path_basename = os.path.basename(os.path.normpath(model_name_path)).lower()
 
     for adapter in model_adapters:
-        if adapter.match(model_path_basename) and type(adapter) != BaseModel:
+        if adapter.match(model_path_basename):
             return adapter
 
     raise ValueError(f"No valid model adapter for {model_name_path}")

intel_extension_for_transformers/neural_chat/models/chatglm_model.py

@@ -15,7 +15,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from .base_model import BaseModel, register_model_adapter
+from .base_model import BaseModel
 from fastchat.conversation import get_conv_template, Conversation
 import logging
 
@@ -27,7 +27,7 @@
 logger = logging.getLogger(__name__)
 
 class ChatGlmModel(BaseModel):
-    def match(self, model_path: str):
+    def match(self):
         """
         Check if the provided model_path matches the current model.
 
@@ -37,9 +37,9 @@ def match(self, model_path: str):
         Returns:
             bool: True if the model_path matches, False otherwise.
         """
-        return "chatglm" in model_path.lower()
+        return "chatglm" in self.model_name.lower()
 
-    def get_default_conv_template(self, model_path: str) -> Conversation:
+    def get_default_conv_template(self) -> Conversation:
         """
         Get the default conversation template for the given model path.
 
@@ -49,9 +49,6 @@ def get_default_conv_template(self, model_path: str) -> Conversation:
         Returns:
             Conversation: A default conversation template.
         """
-        model_path = model_path.lower()
-        if "chatglm2" in model_path.lower():
+        if "chatglm2" in self.model_name.lower():
             return get_conv_template("chatglm2")
         return get_conv_template("chatglm")
-
-register_model_adapter(ChatGlmModel)

intel_extension_for_transformers/neural_chat/models/llama_model.py

@@ -15,7 +15,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from .base_model import BaseModel, register_model_adapter
+from .base_model import BaseModel
 import logging
 from fastchat.conversation import get_conv_template, Conversation
 
@@ -27,28 +27,20 @@
 logger = logging.getLogger(__name__)
 
 class LlamaModel(BaseModel):
-    def match(self, model_path: str):
+    def match(self):
         """
         Check if the provided model_path matches the current model.
 
-        Args:
-            model_path (str): Path to a model.
-
         Returns:
             bool: True if the model_path matches, False otherwise.
         """
-        return "llama" in model_path.lower()
+        return "llama" in self.model_name.lower()
 
-    def get_default_conv_template(self, model_path: str) -> Conversation:
+    def get_default_conv_template(self) -> Conversation:
         """
         Get the default conversation template for the given model path.
 
-        Args:
-            model_path (str): Path to the model.
-
         Returns:
             Conversation: A default conversation template.
         """
         return get_conv_template("llama-2")
-
-register_model_adapter(LlamaModel)

intel_extension_for_transformers/neural_chat/models/mistral_model.py

@@ -15,7 +15,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from .base_model import BaseModel, register_model_adapter
+from .base_model import BaseModel
 import logging
 from fastchat.conversation import get_conv_template, Conversation
 
@@ -27,28 +27,20 @@
 logger = logging.getLogger(__name__)
 
 class MistralModel(BaseModel):
-    def match(self, model_path: str):
+    def match(self):
         """
         Check if the provided model_path matches the current model.
 
-        Args:
-            model_path (str): Path to a model.
-
         Returns:
             bool: True if the model_path matches, False otherwise.
         """
-        return "mistral" in model_path.lower()
+        return "mistral" in self.model_name.lower()
 
-    def get_default_conv_template(self, model_path: str) -> Conversation:
+    def get_default_conv_template(self) -> Conversation:
         """
         Get the default conversation template for the given model path.
 
-        Args:
-            model_path (str): Path to the model.
-
         Returns:
             Conversation: A default conversation template.
         """
         return get_conv_template("mistral")
-
-register_model_adapter(MistralModel)

intel_extension_for_transformers/neural_chat/models/mpt_model.py

@@ -15,7 +15,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from .base_model import BaseModel, register_model_adapter
+from .base_model import BaseModel
 import logging
 from fastchat.conversation import get_conv_template, Conversation
 
@@ -27,28 +27,20 @@
 logger = logging.getLogger(__name__)
 
 class MptModel(BaseModel):
-    def match(self, model_path: str):
+    def match(self):
         """
         Check if the provided model_path matches the current model.
 
-        Args:
-            model_path (str): Path to a model.
-
         Returns:
             bool: True if the model_path matches, False otherwise.
         """
-        return "mpt" in model_path.lower()
+        return "mpt" in self.model_name.lower()
 
-    def get_default_conv_template(self, model_path: str) -> Conversation:
+    def get_default_conv_template(self) -> Conversation:
         """
         Get the default conversation template for the given model path.
 
-        Args:
-            model_path (str): Path to the model.
-
         Returns:
             Conversation: A default conversation template.
         """
         return get_conv_template("mpt-7b-chat")
-
-register_model_adapter(MptModel)