Add phi-4 support

fastchat/conversation.py

@@ -39,6 +39,7 @@ class SeparatorStyle(IntEnum):
     GEMMA = auto()
     CLLM = auto()
     DEFAULT = auto()
+    PHI4 = auto()
 
 
 IMAGE_PLACEHOLDER_STR = "$$<image>$$"
@@ -324,6 +325,16 @@ def get_prompt(self) -> str:
                 else:
                     ret += role + ":"
             return ret
+        elif self.sep_style == SeparatorStyle.PHI4:
+            ret = ""
+            if self.system_message:
+                ret += system_prompt
+            for i, (role, message) in enumerate(self.messages):
+                if message:
+                    ret += f"<|im_start|>{role}<|im_sep|>{message.strip()}<|im_end|>"
+                else:
+                    ret += f"<|im_start|>{role}<|im_sep|>"
+            return ret
         else:
             raise ValueError(f"Invalid style: {self.sep_style}")
 
@@ -2298,6 +2309,19 @@ def get_conv_template(name: str) -> Conversation:
     )
 )
 
+# reference: https://github.com/ggerganov/llama.cpp/pull/11148
+register_conv_template(
+    Conversation(
+        name="phi-4",
+        system_template="<|im_start|>system<|im_sep|>{system_message}<|im_end|>",
+        roles=("user", "assistant"),
+        sep_style=SeparatorStyle.PHI4,
+        sep="<|im_sep|>",
+        stop_str="<|im_end|>",
+        stop_token_ids=[100257, 100265], # <|endoftext|>: 100257, <|im_end|>": 100265,
+    )
+)
+
 
 if __name__ == "__main__":
     from fastchat.conversation import get_conv_template

fastchat/model/model_adapter.py

@@ -2500,6 +2500,16 @@ def get_default_conv_template(self, model_path: str) -> Conversation:
         return get_conv_template("api_based_default")
 
 
+class Phi4Adapter(BaseModelAdapter):
+    """The model adapter for Phi-4"""
+
+    def match(self, model_path: str):
+        return "phi-4" in model_path.lower()
+
+    def get_default_conv_template(self, model_path: str) -> Conversation:
+        return get_conv_template("phi-4")
+
+
 # Note: the registration order matters.
 # The one registered earlier has a higher matching priority.
 register_model_adapter(PeftModelAdapter)
@@ -2602,6 +2612,7 @@ def get_default_conv_template(self, model_path: str) -> Conversation:
 register_model_adapter(Llama31Adapter)
 register_model_adapter(GrokAdapter)
 register_model_adapter(NoSystemAdapter)
+register_model_adapter(Phi4Adapter)
 
 # After all adapters, try the default base adapter.
 register_model_adapter(BaseModelAdapter)

fastchat/model/model_registry.py

@@ -254,6 +254,15 @@ def get_model_info(name: str) -> ModelInfo:
     "A capable and cost-effective small language models (SLMs) by Microsoft",
 )
 
+register_model_info(
+    [
+        "phi-4"
+    ],
+    "Phi-4",
+    "https://techcommunity.microsoft.com/blog/aiplatformblog/introducing-phi-4-microsoft%E2%80%99s-newest-small-language-model-specializing-in-comple/4357090",
+    "A capable and cost-effective small language model (SLMs) by Microsoft",
+)
+
 register_model_info(
     [
         "minicpm-v-2_6",