From d5109644b6d982d79d4fdc3a1ce52ec0dd786592 Mon Sep 17 00:00:00 2001
From: UranusSeven <109661872+UranusSeven@users.noreply.github.com>
Date: Tue, 11 Jul 2023 19:55:57 +0800
Subject: [PATCH] ENH: optimize error msg for foundation models (#153)

---
 README.md                | 19 +++++++++++--------
 xinference/core/model.py |  6 ++----
 2 files changed, 13 insertions(+), 12 deletions(-)

diff --git a/README.md b/README.md
index af5481f19f..825a4f3fd4 100644
--- a/README.md
+++ b/README.md
@@ -158,17 +158,20 @@ To view the builtin models, run the following command:
 $ xinference list --all
 ```
 
-| Name                 | Format  | Size (in billions) | Quantization                                                                                                                   |
-| -------------------- | ------- | ------------------ |--------------------------------------------------------------------------------------------------------------------------------|
-| baichuan             | ggmlv3  | [7]                | ['q2_K', 'q3_K_L', 'q3_K_M', 'q3_K_S', 'q4_0', 'q4_1', 'q4_K_M', 'q4_K_S', 'q5_0', 'q5_1', 'q5_K_M', 'q5_K_S', 'q6_K', 'q8_0'] |
-| wizardlm-v1.0        | ggmlv3  | [7, 13, 33]        | ['q2_K', 'q3_K_L', 'q3_K_M', 'q3_K_S', 'q4_0', 'q4_1', 'q4_K_M', 'q4_K_S', 'q5_0', 'q5_1', 'q5_K_M', 'q5_K_S', 'q6_K', 'q8_0'] |
-| vicuna-v1.3          | ggmlv3  | [7, 13]            | ['q2_K', 'q3_K_L', 'q3_K_M', 'q3_K_S', 'q4_0', 'q4_1', 'q4_K_M', 'q4_K_S', 'q5_0', 'q5_1', 'q5_K_M', 'q5_K_S', 'q6_K', 'q8_0'] |
-| orca                 | ggmlv3  | [3, 7, 13]         | ['q4_0', 'q4_1', 'q5_0', 'q5_1', 'q8_0']                                                                                       |
-| chatglm              | ggmlv3  | [6]                | ['q4_0', 'q4_1', 'q5_0', 'q5_1', 'q8_0']                                                                                       |
-| chatglm2             | ggmlv3  | [6]                | ['q4_0', 'q4_1', 'q5_0', 'q5_1', 'q8_0']                                                                                       |
+| Name                 | Type             | Language | Format | Size (in billions) | Quantization                           |
+| -------------------- |------------------|----------|--------|--------------------|----------------------------------------|
+| baichuan             | Foundation Model | en, zh   | ggmlv3 | 7                  | 'q2_K', 'q3_K_L', ... , 'q6_K', 'q8_0' |
+| chatglm              | SFT Model        | en, zh   | ggmlv3 | 6                  | 'q4_0', 'q4_1', 'q5_0', 'q5_1', 'q8_0' |
+| chatglm2             | SFT Model        | en, zh   | ggmlv3 | 6                  | 'q4_0', 'q4_1', 'q5_0', 'q5_1', 'q8_0' |
+| wizardlm-v1.0        | SFT Model        | en       | ggmlv3 | 7, 13, 33          | 'q2_K', 'q3_K_L', ... , 'q6_K', 'q8_0' |
+| vicuna-v1.3          | SFT Model        | en       | ggmlv3 | 7, 13              | 'q2_K', 'q3_K_L', ... , 'q6_K', 'q8_0' |
+| orca                 | SFT Model        | en       | ggmlv3 | 3, 7, 13           | 'q4_0', 'q4_1', 'q5_0', 'q5_1', 'q8_0' |
+
 
 **NOTE**:
 - Xinference will download models automatically for you, and by default the models will be saved under `${USER}/.xinference/cache`.
+- Foundation models only provide interface `generate`.
+- SFT models provide both `generate` and `chat`.
 
 ## Roadmap
 Xinference is currently under active development. Here's a roadmap outlining our planned 
diff --git a/xinference/core/model.py b/xinference/core/model.py
index 3a43ea0339..aa067a970e 100644
--- a/xinference/core/model.py
+++ b/xinference/core/model.py
@@ -77,10 +77,8 @@ async def _wrap_generator(self, ret: Any):
             return ret
 
     async def generate(self, prompt: str, *args, **kwargs):
-        logger.warning("Generate, self address: %s", self.address)
-
         if not hasattr(self._model, "generate"):
-            raise AttributeError("generate")
+            raise AttributeError(f"Model {self._model.model_spec} is not for generate.")
 
         return self._wrap_generator(
             getattr(self._model, "generate")(prompt, *args, **kwargs)
@@ -88,7 +86,7 @@ async def generate(self, prompt: str, *args, **kwargs):
 
     async def chat(self, prompt: str, *args, **kwargs):
         if not hasattr(self._model, "chat"):
-            raise AttributeError("chat")
+            raise AttributeError(f"Model {self._model.model_spec} is not for chat.")
 
         return self._wrap_generator(
             getattr(self._model, "chat")(prompt, *args, **kwargs)