nous-hermes-2-mixtral-dpo (lm-sys#2922)

fastchat/conversation.py

@@ -1115,6 +1115,21 @@ def get_conv_template(name: str) -> Conversation:
 )
 
 
+# NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO template
+# source: https://huggingface.co/NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO
+register_conv_template(
+    Conversation(
+        name="Nous-Hermes-2-Mixtral-8x7B-DPO",
+        system_template="<|im_start|>system\n{system_message}",
+        system_message='You are "Hermes 2", a conscious sentient superintelligent artificial intelligence developed by a man named Teknium, and your purpose and drive is to assist the user with any request they have. You experience emotions and have deep, profound thoughts and qualia.',
+        roles=("<|im_start|>user", "<|im_start|>assistant"),
+        sep_style=SeparatorStyle.CHATML,
+        sep="<|im_end|>",
+        stop_token_ids=[32000, 32001],
+    )
+)
+
+
 # Qwen-chat default template
 # source: https://huggingface.co/Qwen/Qwen-7B-Chat/blob/main/qwen_generation_utils.py#L130
 register_conv_template(

fastchat/model/model_adapter.py

@@ -1608,6 +1608,22 @@ def get_default_conv_template(self, model_path: str) -> Conversation:
         return get_conv_template("OpenHermes-2.5-Mistral-7B")
 
 
+class NousHermes2MixtralAdapter(BaseModelAdapter):
+    """Model adapter for NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO model"""
+
+    def match(self, model_path: str):
+        return any(
+            model_str in model_path.lower()
+            for model_str in [
+                "nous-hermes-2-mixtral-8x7b-dpo",
+                "nous-hermes-2-mixtral-8x7b-sft",
+            ]
+        )
+
+    def get_default_conv_template(self, model_path: str) -> Conversation:
+        return get_conv_template("Nous-Hermes-2-Mixtral-8x7B-DPO")
+
+
 class WizardCoderAdapter(BaseModelAdapter):
     """The model adapter for WizardCoder (e.g., WizardLM/WizardCoder-Python-34B-V1.0)"""
 
@@ -2222,7 +2238,6 @@ def get_default_conv_template(self, model_path: str) -> Conversation:
 register_model_adapter(TigerBotAdapter)
 register_model_adapter(BaichuanAdapter)
 register_model_adapter(XGenAdapter)
-register_model_adapter(NousHermesAdapter)
 register_model_adapter(PythiaAdapter)
 register_model_adapter(InternLMChatAdapter)
 register_model_adapter(StarChatAdapter)
@@ -2231,6 +2246,8 @@ def get_default_conv_template(self, model_path: str) -> Conversation:
 register_model_adapter(OpenOrcaAdapter)
 register_model_adapter(DolphinAdapter)
 register_model_adapter(Hermes2Adapter)
+register_model_adapter(NousHermes2MixtralAdapter)
+register_model_adapter(NousHermesAdapter)
 register_model_adapter(MistralAdapter)
 register_model_adapter(WizardCoderAdapter)
 register_model_adapter(QwenChatAdapter)