feat(model): Support Starling-LM-7B-beta (#1350)

README.md

@@ -158,6 +158,7 @@ At present, we have introduced several key features to showcase our current capa
   We offer extensive model support, including dozens of large language models (LLMs) from both open-source and API agents, such as LLaMA/LLaMA2, Baichuan, ChatGLM, Wenxin, Tongyi, Zhipu, and many more. 
 
   - News
+    - 🔥🔥🔥  [Starling-LM-7B-beta](https://huggingface.co/Nexusflow/Starling-LM-7B-beta)
     - 🔥🔥🔥  [gemma-7b-it](https://huggingface.co/google/gemma-7b-it)
     - 🔥🔥🔥  [gemma-2b-it](https://huggingface.co/google/gemma-2b-it)
     - 🔥🔥🔥  [SOLAR-10.7B](https://huggingface.co/upstage/SOLAR-10.7B-Instruct-v1.0)

README.zh.md

@@ -152,6 +152,7 @@
   海量模型支持，包括开源、API代理等几十种大语言模型。如LLaMA/LLaMA2、Baichuan、ChatGLM、文心、通义、智谱等。当前已支持如下模型: 
 
   - 新增支持模型
+    - 🔥🔥🔥  [Starling-LM-7B-beta](https://huggingface.co/Nexusflow/Starling-LM-7B-beta)
     - 🔥🔥🔥  [gemma-7b-it](https://huggingface.co/google/gemma-7b-it)
     - 🔥🔥🔥  [gemma-2b-it](https://huggingface.co/google/gemma-2b-it)
     - 🔥🔥🔥  [SOLAR-10.7B](https://huggingface.co/upstage/SOLAR-10.7B-Instruct-v1.0)
@@ -161,6 +162,7 @@
   - [更多开源模型](https://www.yuque.com/eosphoros/dbgpt-docs/iqaaqwriwhp6zslc#qQktR)
 
   - 支持在线代理模型 
+    - [x] [零一万物.Yi](https://platform.lingyiwanwu.com/docs)
     - [x] [OpenAI·ChatGPT](https://api.openai.com/)
     - [x] [百川·Baichuan](https://platform.baichuan-ai.com/)
     - [x] [阿里·通义](https://www.aliyun.com/product/dashscope)

dbgpt/configs/model_config.py

@@ -156,6 +156,7 @@ LLM_MODEL_CONFIG = {
     "gemma-7b-it": os.path.join(MODEL_PATH, "gemma-7b-it"),
     # https://huggingface.co/google/gemma-2b-it
     "gemma-2b-it": os.path.join(MODEL_PATH, "gemma-2b-it"),
+    "starling-lm-7b-beta": os.path.join(MODEL_PATH, "Starling-LM-7B-beta"),
 }
 
 EMBEDDING_MODEL_CONFIG = {

dbgpt/model/adapter/hf_adapter.py

@@ -193,7 +193,61 @@ class GemmaAdapter(NewHFChatModelAdapter):
         )
 
 
+class StarlingLMAdapter(NewHFChatModelAdapter):
+    """
+    https://huggingface.co/Nexusflow/Starling-LM-7B-beta
+    """
+
+    support_4bit: bool = True
+    support_8bit: bool = True
+    support_system_message: bool = False
+
+    def do_match(self, lower_model_name_or_path: Optional[str] = None):
+        return (
+            lower_model_name_or_path
+            and "starling-" in lower_model_name_or_path
+            and "lm" in lower_model_name_or_path
+        )
+
+    def get_str_prompt(
+        self,
+        params: Dict,
+        messages: List[ModelMessage],
+        tokenizer: Any,
+        prompt_template: str = None,
+        convert_to_compatible_format: bool = False,
+    ) -> Optional[str]:
+        str_prompt = super().get_str_prompt(
+            params,
+            messages,
+            tokenizer,
+            prompt_template,
+            convert_to_compatible_format,
+        )
+        chat_mode = None
+        if params and "context" in params and "chat_mode" in params["context"]:
+            chat_mode = params["context"].get("chat_mode")
+        if chat_mode in [
+            "chat_dashboard",
+            "chat_with_db_execute",
+            "excel_learning",
+            "chat_excel",
+        ]:
+            # Coding conversation, use code prompt
+            # This is a temporary solution, we should use a better way to distinguish the conversation type
+            # https://huggingface.co/Nexusflow/Starling-LM-7B-beta#code-examples
+            str_prompt = str_prompt.replace("GPT4 Correct User:", "Code User:").replace(
+                "GPT4 Correct Assistant:", "Code Assistant:"
+            )
+            logger.info(
+                f"Use code prompt for chat_mode: {chat_mode}, transform 'GPT4 Correct User:' to 'Code User:' "
+                "and 'GPT4 Correct Assistant:' to 'Code Assistant:'"
+            )
+        return str_prompt
+
+
 register_model_adapter(YiAdapter)
 register_model_adapter(Mixtral8x7BAdapter)
 register_model_adapter(SOLARAdapter)
 register_model_adapter(GemmaAdapter)
+register_model_adapter(StarlingLMAdapter)

setup.py

@@ -630,6 +630,7 @@ def default_requires():
         "zhipuai",
         "dashscope",
         "chardet",
+        "sentencepiece",
     ]
     setup_spec.extras["default"] += setup_spec.extras["framework"]
     setup_spec.extras["default"] += setup_spec.extras["rag"]