add support for iei yuan2.0 (https://huggingface.co/IEITYuan) (lm-sys#2919)

Author: wangpengfei1013

docs/model_support.md

@@ -61,6 +61,7 @@
 - [WizardLM/WizardLM-13B-V1.0](https://huggingface.co/WizardLM/WizardLM-13B-V1.0)
 - [WizardLM/WizardCoder-15B-V1.0](https://huggingface.co/WizardLM/WizardCoder-15B-V1.0)
 - [Xwin-LM/Xwin-LM-7B-V0.1](https://huggingface.co/Xwin-LM/Xwin-LM-70B-V0.1)
+- [IEITYuan/Yuan2-2B/51B/102B-hf](https://huggingface.co/IEITYuan)
 - Any [EleutherAI](https://huggingface.co/EleutherAI) pythia model such as [pythia-6.9b](https://huggingface.co/EleutherAI/pythia-6.9b)
 - Any [Peft](https://github.com/huggingface/peft) adapter trained on top of a
   model above.  To activate, must have `peft` in the model path.  Note: If

fastchat/conversation.py

@@ -1440,6 +1440,22 @@ def get_conv_template(name: str) -> Conversation:
 )
 
 
+#yuan 2.0 template
+#reference:https://github.com/IEIT-Yuan/Yuan-2.0
+#reference:https://huggingface.co/IEITYuan
+register_conv_template(
+    Conversation(
+        name="yuan",
+        system_template="",
+        roles=("", ""),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep="<sep>",
+        stop_str="<eod>",
+
+    )
+)
+
+
 if __name__ == "__main__":
     from fastchat.conversation import get_conv_template
 

fastchat/model/model_adapter.py

@@ -2159,6 +2159,22 @@ def match(self, model_path: str):
     def get_default_conv_template(self, model_path: str) -> Conversation:
         return get_conv_template("solar")
 
+class Yuan2Adapter(BaseModelAdapter):
+    """The model adapter for Yuan """
+
+    def match(self, model_path: str):
+        return "yuan" in model_path.lower()
+
+    def load_model(self, model_path: str, from_pretrained_kwargs: dict):
+        model, tokenizer = super().load_model(model_path, from_pretrained_kwargs)
+        tokenizer.add_tokens(
+            ['<sep>', '<pad>', '<mask>', '<predict>', '<FIM_SUFFIX>', '<FIM_PREFIX>', '<FIM_MIDDLE>', '<commit_before>',
+             '<commit_msg>', '<commit_after>', '<jupyter_start>', '<jupyter_text>', '<jupyter_code>',
+             '<jupyter_output>', '<empty_output>'], special_tokens=True)
+        return model, tokenizer
+
+    def get_default_conv_template(self, model_path: str) -> Conversation:
+        return get_conv_template("yuan")
 
 # Note: the registration order matters.
 # The one registered earlier has a higher matching priority.
@@ -2246,6 +2262,7 @@ def get_default_conv_template(self, model_path: str) -> Conversation:
 register_model_adapter(MetaMathAdapter)
 register_model_adapter(BagelAdapter)
 register_model_adapter(SolarAdapter)
+register_model_adapter(Yuan2Adapter)
 
 # After all adapters, try the default base adapter.
 register_model_adapter(BaseModelAdapter)

fastchat/model/model_registry.py

@@ -543,3 +543,10 @@ def get_model_info(name: str) -> ModelInfo:
     "https://huggingface.co/meta-math",
     "MetaMath is a finetune of Llama2 on [MetaMathQA](https://huggingface.co/datasets/meta-math/MetaMathQA) that specializes in mathematical reasoning.",
 )
+
+register_model_info(
+    ["Yuan2-2B-hf","Yuan2-51B-hf","Yuan2-102B-hf"],
+    "IEIYuan",
+    "https://huggingface.co/IEITYuan",
+    "Yuan2 is a Basemodel developed by IEI.",
+)