Add Microsoft/Orca-2-7b and update model support docs

docs/model_support.md

@@ -52,6 +52,8 @@
 - [HuggingFaceH4/zephyr-7b-alpha](https://huggingface.co/HuggingFaceH4/zephyr-7b-alpha)
 - [Xwin-LM/Xwin-LM-7B-V0.1](https://huggingface.co/Xwin-LM/Xwin-LM-70B-V0.1)
 - [OpenLemur/lemur-70b-chat-v1](https://huggingface.co/OpenLemur/lemur-70b-chat-v1)
+- [allenai/tulu-2-dpo-7b](https://huggingface.co/allenai/tulu-2-dpo-7b)
+- [Microsoft/Orca-2-7b](https://huggingface.co/microsoft/Orca-2-7b)
 - 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

@@ -1190,6 +1190,19 @@ def get_conv_template(name: str) -> Conversation:
     )
 )
 
+# Orca-2 template
+# reference: https://huggingface.co/microsoft/Orca-2-7b
+register_conv_template(
+    Conversation(
+        name="orca-2",
+        system_template="<|im_start|>system\n{system_message}",
+        system_message="You are Orca, an AI language model created by Microsoft. You are a cautious assistant. You carefully follow instructions. You are helpful and harmless and you follow ethical guidelines and promote positive behavior.",
+        roles=("<|im_start|>user", "<|im_start|>assistant"),
+        sep_style=SeparatorStyle.CHATML,
+        sep="<|im_end|>",
+        stop_str="<|im_end|>",
+    )
+)
 
 if __name__ == "__main__":
     from fastchat.conversation import get_conv_template

fastchat/model/model_adapter.py

@@ -1827,6 +1827,18 @@ def get_default_conv_template(self, model_path: str) -> Conversation:
         return get_conv_template("metharme")
 
 
+class MicrosoftOrcaAdapter(BaseModelAdapter):
+    """The model adapter for Microsoft/Orca-2 series of models (e.g. Microsoft/Orca-2-7b, Microsoft/Orca-2-13b)"""
+
+    use_fast_tokenizer = False  # Flag neeeded since tokenizers>=0.13.3 is required for a normal functioning of this module
+
+    def match(self, model_path: str):
+        return "orca-2" in model_path.lower()
+
+    def get_default_conv_template(self, model_path: str) -> Conversation:
+        return get_conv_template("orca-2")
+
+
 # Note: the registration order matters.
 # The one registered earlier has a higher matching priority.
 register_model_adapter(PeftModelAdapter)
@@ -1893,7 +1905,7 @@ def get_default_conv_template(self, model_path: str) -> Conversation:
 register_model_adapter(XwinLMAdapter)
 register_model_adapter(LemurAdapter)
 register_model_adapter(PygmalionAdapter)
-
+register_model_adapter(MicrosoftOrcaAdapter)
 
 # After all adapters, try the default base adapter.
 register_model_adapter(BaseModelAdapter)