test_enable_thinking.py 6.9 KB
Newer Older
1
2
"""
Usage:
3
4
5
6
python3 -m unittest openai_server.features.test_enable_thinking.TestEnableThinking.test_chat_completion_with_reasoning
python3 -m unittest openai_server.features.test_enable_thinking.TestEnableThinking.test_chat_completion_without_reasoning
python3 -m unittest openai_server.features.test_enable_thinking.TestEnableThinking.test_stream_chat_completion_with_reasoning
python3 -m unittest openai_server.features.test_enable_thinking.TestEnableThinking.test_stream_chat_completion_without_reasoning
7
8
9
10
11
12
13
14
15
"""

import asyncio
import json
import os
import sys
import time
import unittest

16
import openai
17
18
import requests

19
from sglang.srt.hf_transformers_utils import get_tokenizer
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
from sglang.srt.utils import kill_process_tree
from sglang.test.test_utils import (
    DEFAULT_ENABLE_THINKING_MODEL_NAME_FOR_TEST,
    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
    DEFAULT_URL_FOR_TEST,
    CustomTestCase,
    popen_launch_server,
)


class TestEnableThinking(CustomTestCase):
    @classmethod
    def setUpClass(cls):
        cls.model = DEFAULT_ENABLE_THINKING_MODEL_NAME_FOR_TEST
        cls.base_url = DEFAULT_URL_FOR_TEST
        cls.api_key = "sk-1234"
        cls.process = popen_launch_server(
            cls.model,
            cls.base_url,
            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
            api_key=cls.api_key,
            other_args=[
                "--reasoning-parser",
                "qwen3",
            ],
        )

    @classmethod
    def tearDownClass(cls):
        kill_process_tree(cls.process.pid)

    def test_chat_completion_with_reasoning(self):
        # Test non-streaming with "enable_thinking": True, reasoning_content should not be empty
        client = requests.post(
            f"{self.base_url}/v1/chat/completions",
            headers={"Authorization": f"Bearer {self.api_key}"},
            json={
                "model": self.model,
                "messages": [{"role": "user", "content": "Hello"}],
                "temperature": 0,
                "separate_reasoning": True,
                "chat_template_kwargs": {"enable_thinking": True},
            },
        )

        self.assertEqual(client.status_code, 200, f"Failed with: {client.text}")
        data = client.json()

        self.assertIn("choices", data)
        self.assertTrue(len(data["choices"]) > 0)
        self.assertIn("message", data["choices"][0])
        self.assertIn("reasoning_content", data["choices"][0]["message"])
        self.assertIsNotNone(data["choices"][0]["message"]["reasoning_content"])

    def test_chat_completion_without_reasoning(self):
        # Test non-streaming with "enable_thinking": False, reasoning_content should be empty
        client = requests.post(
            f"{self.base_url}/v1/chat/completions",
            headers={"Authorization": f"Bearer {self.api_key}"},
            json={
                "model": self.model,
                "messages": [{"role": "user", "content": "Hello"}],
                "temperature": 0,
                "separate_reasoning": True,
                "chat_template_kwargs": {"enable_thinking": False},
            },
        )

        self.assertEqual(client.status_code, 200, f"Failed with: {client.text}")
        data = client.json()

        self.assertIn("choices", data)
        self.assertTrue(len(data["choices"]) > 0)
        self.assertIn("message", data["choices"][0])

        if "reasoning_content" in data["choices"][0]["message"]:
            self.assertIsNone(data["choices"][0]["message"]["reasoning_content"])

    def test_stream_chat_completion_with_reasoning(self):
        # Test streaming with "enable_thinking": True, reasoning_content should not be empty
        response = requests.post(
            f"{self.base_url}/v1/chat/completions",
            headers={"Authorization": f"Bearer {self.api_key}"},
            json={
                "model": self.model,
                "messages": [{"role": "user", "content": "Hello"}],
                "temperature": 0,
                "separate_reasoning": True,
                "stream": True,
                "chat_template_kwargs": {"enable_thinking": True},
            },
            stream=True,
        )

        self.assertEqual(response.status_code, 200, f"Failed with: {response.text}")

        has_reasoning = False
        has_content = False

        print("\n=== Stream With Reasoning ===")
        for line in response.iter_lines():
            if line:
                line = line.decode("utf-8")
                if line.startswith("data:") and not line.startswith("data: [DONE]"):
                    data = json.loads(line[6:])
                    if "choices" in data and len(data["choices"]) > 0:
                        delta = data["choices"][0].get("delta", {})

                        if "reasoning_content" in delta and delta["reasoning_content"]:
                            has_reasoning = True

                        if "content" in delta and delta["content"]:
                            has_content = True

        self.assertTrue(
            has_reasoning,
            "The reasoning content is not included in the stream response",
        )
        self.assertTrue(
            has_content, "The stream response does not contain normal content"
        )

    def test_stream_chat_completion_without_reasoning(self):
        # Test streaming with "enable_thinking": False, reasoning_content should  be empty
        response = requests.post(
            f"{self.base_url}/v1/chat/completions",
            headers={"Authorization": f"Bearer {self.api_key}"},
            json={
                "model": self.model,
                "messages": [{"role": "user", "content": "Hello"}],
                "temperature": 0,
                "separate_reasoning": True,
                "stream": True,
                "chat_template_kwargs": {"enable_thinking": False},
            },
            stream=True,
        )

        self.assertEqual(response.status_code, 200, f"Failed with: {response.text}")

        has_reasoning = False
        has_content = False

        print("\n=== Stream Without Reasoning ===")
        for line in response.iter_lines():
            if line:
                line = line.decode("utf-8")
                if line.startswith("data:") and not line.startswith("data: [DONE]"):
                    data = json.loads(line[6:])
                    if "choices" in data and len(data["choices"]) > 0:
                        delta = data["choices"][0].get("delta", {})

                        if "reasoning_content" in delta and delta["reasoning_content"]:
                            has_reasoning = True

                        if "content" in delta and delta["content"]:
                            has_content = True

        self.assertFalse(
            has_reasoning,
            "The reasoning content should not be included in the stream response",
        )
        self.assertTrue(
            has_content, "The stream response does not contain normal content"
        )


if __name__ == "__main__":
    unittest.main()