Clean up unit tests (#1020)

test/srt/run_suite.py

@@ -6,9 +6,9 @@ from sglang.test.test_utils import run_unittest_files
 suites = {
     "minimal": [
         "test_eval_accuracy.py",
-        "test_embedding_openai_server.py",
         "test_openai_server.py",
         "test_vision_openai_server.py",
+        "test_embedding_openai_server.py",
         "test_chunked_prefill.py",
         "test_torch_compile.py",
         "test_models_from_modelscope.py",

test/srt/test_chunked_prefill.py

@@ -37,9 +37,4 @@ class TestAccuracy(unittest.TestCase):
 
 
 if __name__ == "__main__":
-    unittest.main(warnings="ignore")
-
-    # t = TestAccuracy()
-    # t.setUpClass()
-    # t.test_mmlu()
-    # t.tearDownClass()
+    unittest.main()

test/srt/test_embedding_openai_server.py

@@ -1,11 +1,8 @@
-import json
-import time
 import unittest
 
 import openai
 
 from sglang.srt.hf_transformers_utils import get_tokenizer
-from sglang.srt.openai_api.protocol import EmbeddingObject
 from sglang.srt.utils import kill_child_process
 from sglang.test.test_utils import popen_launch_server
 
@@ -65,12 +62,12 @@ class TestOpenAIServer(unittest.TestCase):
         ), f"{response.usage.total_tokens} vs {num_prompt_tokens}"
 
     def run_batch(self):
-        # FIXME not implemented
+        # FIXME: not implemented
         pass
 
     def test_embedding(self):
-        # TODO the fields of encoding_format, dimensions, user are skipped
-        # TODO support use_list_input
+        # TODO: the fields of encoding_format, dimensions, user are skipped
+        # TODO: support use_list_input
         for use_list_input in [False, True]:
             for token_input in [False, True]:
                 self.run_embedding(use_list_input, token_input)
@@ -80,9 +77,4 @@ class TestOpenAIServer(unittest.TestCase):
 
 
 if __name__ == "__main__":
-    unittest.main(warnings="ignore")
-
-    # t = TestOpenAIServer()
-    # t.setUpClass()
-    # t.test_embedding()
-    # t.tearDownClass()
+    unittest.main()

test/srt/test_eval_accuracy.py

@@ -32,9 +32,4 @@ class TestAccuracy(unittest.TestCase):
 
 
 if __name__ == "__main__":
-    unittest.main(warnings="ignore")
-
-    # t = TestAccuracy()
-    # t.setUpClass()
-    # t.test_mmlu()
-    # t.tearDownClass()
+    unittest.main()

test/srt/test_models_from_modelscope.py

@@ -44,4 +44,4 @@ class TestDownloadFromModelScope(unittest.TestCase):
 
 
 if __name__ == "__main__":
-    unittest.main(warnings="ignore")
+    unittest.main()

test/srt/test_openai_server.py

@@ -399,9 +399,4 @@ class TestOpenAIServer(unittest.TestCase):
 
 
 if __name__ == "__main__":
-    unittest.main(warnings="ignore")
-
-    # t = TestOpenAIServer()
-    # t.setUpClass()
-    # t.test_completion()
-    # t.tearDownClass()
+    unittest.main()

test/srt/test_skip_tokenizer_init.py

@@ -1,18 +1,13 @@
 import json
-import os
-import sys
 import unittest
 
 import requests
 
 from sglang.srt.utils import kill_child_process
-from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import DEFAULT_MODEL_NAME_FOR_TEST, popen_launch_server
 
-# os.environ["CUDA_VISIBLE_DEVICES"] = "1"
 
-
-class TestSRTEndpoint(unittest.TestCase):
+class TestSkipTokenizerInit(unittest.TestCase):
 
     @classmethod
     def setUpClass(cls):
@@ -26,9 +21,7 @@ class TestSRTEndpoint(unittest.TestCase):
     def tearDownClass(cls):
         kill_child_process(cls.process.pid)
 
-    def run_decode(
-        self, return_logprob=False, top_logprobs_num=0, return_text=False, n=1
-    ):
+    def run_decode(self, return_logprob=False, top_logprobs_num=0, n=1):
         response = requests.post(
             self.base_url + "/generate",
             json={
@@ -50,7 +43,6 @@ class TestSRTEndpoint(unittest.TestCase):
                 "stream": False,
                 "return_logprob": return_logprob,
                 "top_logprobs_num": top_logprobs_num,
-                "return_text_in_logprobs": return_text,
                 "logprob_start_len": 0,
             },
         )
@@ -65,13 +57,11 @@ class TestSRTEndpoint(unittest.TestCase):
 
     def test_logprob(self):
         for top_logprobs_num in [0, 3]:
-            for return_text in [False, False]:
-                self.run_decode(
-                    return_logprob=True,
-                    top_logprobs_num=top_logprobs_num,
-                    return_text=return_text,
-                )
+            self.run_decode(
+                return_logprob=True,
+                top_logprobs_num=top_logprobs_num,
+            )
 
 
 if __name__ == "__main__":
-    unittest.main(warnings="ignore")
+    unittest.main()

test/srt/test_srt_endpoint.py

@@ -4,7 +4,6 @@ import unittest
 import requests
 
 from sglang.srt.utils import kill_child_process
-from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import DEFAULT_MODEL_NAME_FOR_TEST, popen_launch_server
 
 
@@ -59,4 +58,4 @@ class TestSRTEndpoint(unittest.TestCase):
 
 
 if __name__ == "__main__":
-    unittest.main(warnings="ignore")
+    unittest.main()

test/srt/test_torch_compile.py

@@ -34,9 +34,4 @@ class TestAccuracy(unittest.TestCase):
 
 
 if __name__ == "__main__":
-    unittest.main(warnings="ignore")
-
-    # t = TestAccuracy()
-    # t.setUpClass()
-    # t.test_mmlu()
-    # t.tearDownClass()
+    unittest.main()

test/srt/test_vision_openai_server.py

@@ -113,9 +113,4 @@ class TestOpenAIVisionServer(unittest.TestCase):
 
 
 if __name__ == "__main__":
-    unittest.main(warnings="ignore")
-
-    # t = TestOpenAIVisionServer()
-    # t.setUpClass()
-    # t.test_chat_completion()
-    # t.tearDownClass()
+    unittest.main()