[Auto Sync] Update detokenizer_manager.py, io_struct.py, mu... (20260120) (#17442)

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Wangfan Fu <wangfan@x.ai>

test/registered/openai_server/basic/test_openai_server.py

@@ -89,7 +89,9 @@ class TestOpenAIServer(CustomTestCase):
 
         if logprobs:
             assert response.choices[0].logprobs
-            assert isinstance(response.choices[0].logprobs.tokens[0], str)
+            assert isinstance(
+                response.choices[0].logprobs.tokens[0], str
+            ), f"{response=}"
             assert isinstance(response.choices[0].logprobs.top_logprobs[1], dict)
             ret_num_top_logprobs = len(response.choices[0].logprobs.top_logprobs[1])
 
@@ -155,6 +157,9 @@ class TestOpenAIServer(CustomTestCase):
 
             if logprobs:
                 assert response.choices[0].logprobs, f"no logprobs in response"
+                print(
+                    f"{response=}, {response.choices[0]=}, {response.choices[0].logprobs=}"
+                )
                 assert isinstance(
                     response.choices[0].logprobs.tokens[0], str
                 ), f"{response.choices[0].logprobs.tokens[0]} is not a string"

test/registered/quant/test_gptqmodel_dynamic.py

@@ -54,11 +54,11 @@ def check_quant_method(model_path: str, use_marlin_kernel: bool):
         model_config=model_config, load_config=load_config, device_config=device_config
     )
 
-    from sglang.srt.layers.linear import UnquantizedLinearMethod
     from sglang.srt.layers.quantization.gptq import (
         GPTQLinearMethod,
         GPTQMarlinLinearMethod,
     )
+    from sglang.srt.layers.quantization.unquant import UnquantizedLinearMethod
 
     linear_method_cls = (
         GPTQMarlinLinearMethod if use_marlin_kernel else (GPTQLinearMethod)