add an openai chat example (#33)

Author: yejunjin

documents/CN/examples_python.md

@@ -357,6 +357,9 @@ docker run -d \
     -m /workspace/qwen/Qwen-7B-Chat \
     /workspace/config_qwen_v10_7b.json
 ```
+
+你还可以使用[openai_chat.py](../../examples/python/4_fastchat/openai_chat.py)来测试使用openai api的聊天客户端。
+
 # 模型配置文件
 
 `<path_to_dashinfer>/examples/python/model_config`目录下提供了一些config示例。

documents/EN/examples_python.md

@@ -357,6 +357,8 @@ docker run -d \
     /workspace/config_qwen_v10_7b.json
 ```
 
+You can also use [openai_chat.py](../../examples/python/4_fastchat/openai_chat.py) to test the chat client using the OpenAI API.
+
 # Model Configuration Files
 
 The `<path_to_dashinfer>/examples/python/model_config` directory provides several configuration examples.

examples/python/4_fastchat/openai_chat.py

@@ -0,0 +1,48 @@
+#!/usr/bin/env python3
+
+import openai
+
+openai.api_key = "EMPTY"
+openai.base_url = "http://localhost:8000/v1/"
+
+def test_list_models():
+    model_list = openai.models.list()
+    names = [x.id for x in model_list.data]
+    return names
+
+def test_chat_completion_stream(model):
+    messages = [{"role": "user", "content": "Talk about the impact of artificial intelligence on different aspects of society. Please talk at least 1000 words"}]
+    res = openai.chat.completions.create(
+        model=model, messages=messages, stream=True, temperature=0
+    )
+    for chunk in res:
+        try:
+            content = chunk.choices[0].delta.content
+            if content is None:
+                content = ""
+        except Exception as e:
+            content = chunk.choices[0].delta.get("content", "")
+        print(content, end="", flush=True)
+    print()
+
+def test_chat_completion(model):
+    completion = openai.chat.completions.create(
+        model=model,
+        messages=[{"role": "user", "content": "Hello! What is your name?"}],
+        temperature=0,
+    )
+    print(completion.choices[0].message.content)
+
+if __name__ == "__main__":
+    model = "Qwen-7B-Chat" # the model name which is typically the last component of model path
+
+    print("List models:")
+    print(test_list_models())
+
+    print("\n\n====================================")
+    print("Chat completion:")
+    test_chat_completion(model)
+
+    print("\n\n====================================")
+    print("Chat completion stream:")
+    test_chat_completion_stream(model)

scripts/docker/fschat_ubuntu_x86.Dockerfile

@@ -8,6 +8,11 @@ COPY ./fschat_entrypoint.sh ./
 SHELL ["conda", "run", "-n", "py38env", "/bin/bash", "-c"]
 RUN pip install \
   -i https://mirrors.aliyun.com/pypi/simple/ \
-  "fschat[model_worker]"
+  "fschat[model_worker]==0.2.36"
+
+# fastchat has a bug in pydantic v2: https://github.com/lm-sys/FastChat/pull/3356
+# downgrade to v1.10.13
+RUN pip uninstall pydantic -y \
+    && pip install -i https://mirrors.aliyun.com/pypi/simple pydantic==1.10.13
 
 ENTRYPOINT ["conda", "run", "--no-capture-output", "-n", "py38env", "./fschat_entrypoint.sh"]