Fix roles in chat interface

xorbitsai · Jan 30, 2024 · 39aa2a6 · 39aa2a6
1 parent 5dad18a
commit 39aa2a6
Show file tree

Hide file tree

Showing 3 changed files with 11 additions and 2 deletions.
diff --git a/xinference/core/chat_interface.py b/xinference/core/chat_interface.py
@@ -98,9 +98,16 @@ def flatten(matrix: List[List[str]]) -> List[str]:
             return flat_list
 
         def to_chat(lst: List[str]) -> List[ChatCompletionMessage]:
+            from ..model.llm import BUILTIN_LLM_PROMPT_STYLE
+
             res = []
+            prompt_style = BUILTIN_LLM_PROMPT_STYLE.get(self.model_name)
+            if prompt_style is None:
+                roles = ["assistant", "user"]
+            else:
+                roles = prompt_style.roles
             for i in range(len(lst)):
-                role = "assistant" if i % 2 == 1 else "user"
+                role = roles[0] if i % 2 == 1 else roles[1]
                 res.append(ChatCompletionMessage(role=role, content=lst[i]))
             return res
 

diff --git a/xinference/model/llm/core.py b/xinference/model/llm/core.py
@@ -135,6 +135,8 @@ def to_dict(self):
             "model_description": self._llm_family.model_description,
             "model_format": self._llm_spec.model_format,
             "model_size_in_billions": self._llm_spec.model_size_in_billions,
+            "model_family": self._llm_family.model_family
+            or self._llm_family.model_name,
             "quantization": self._quantization,
             "model_hub": self._llm_spec.model_hub,
             "revision": self._llm_spec.model_revision,

diff --git a/xinference/web/ui/src/scenes/running_models/index.js b/xinference/web/ui/src/scenes/running_models/index.js
@@ -193,7 +193,7 @@ const RunningModels = () => {
                         },
                         body: JSON.stringify({
                           model_type: row.model_type,
-                          model_name: row.model_name,
+                          model_name: row.model_family,
                           model_size_in_billions: row.model_size_in_billions,
                           model_format: row.model_format,
                           quantization: row.quantization,