Extend lint test to test/ directory (#13247)

.pre-commit-config.yaml

@@ -31,7 +31,7 @@ repos:
         args:
           - --select=F401,F821
           - --fix
-        files: ^(benchmark/|docs/|examples/|python/sglang/|sgl-router/py_*)
+        files: ^(benchmark/|docs/|examples/|python/sglang/|sgl-router/py_*|test/)
         exclude: |
           (?x)^(
           .*/__init__\.py$|

test/lang/test_anthropic_backend.py

@@ -1,4 +1,3 @@
-import json
 import unittest
 
 from sglang import Anthropic, set_default_backend

test/lang/test_litellm_backend.py

@@ -1,4 +1,3 @@
-import json
 import unittest
 
 from sglang import LiteLLM, set_default_backend

test/lang/test_separate_reasoning.py

@@ -7,7 +7,7 @@ python3 -m unittest test/lang/test_separate_reasoning.py
 
 import unittest
 
-from sglang import assistant, gen, separate_reasoning, user
+from sglang import gen, separate_reasoning
 from sglang.lang.ir import SglExprList, SglSeparateReasoning
 from sglang.test.test_utils import CustomTestCase
 

test/lang/test_separate_reasoning_execution.py

@@ -6,11 +6,9 @@ python3 -m unittest test/lang/test_separate_reasoning_execution.py
 """
 
 import threading
-import time
 import unittest
 from unittest.mock import MagicMock, patch
 
-from sglang import assistant, gen, separate_reasoning, user
 from sglang.lang.interpreter import StreamExecutor
 from sglang.lang.ir import SglGen, SglSeparateReasoning
 from sglang.test.test_utils import CustomTestCase

test/srt/ascend/test_ascend_deepep.py

@@ -6,7 +6,6 @@ from urllib.parse import urlparse
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
-    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
     CustomTestCase,
     is_in_ci,

test/srt/ascend/test_ascend_deepseek_mtp.py

@@ -6,7 +6,6 @@ from urllib.parse import urlparse
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
-    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
     CustomTestCase,
     is_in_ci,

test/srt/ascend/test_ascend_tp4_bf16.py

@@ -5,7 +5,6 @@ from urllib.parse import urlparse
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
-    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
     CustomTestCase,
     is_in_ci,

test/srt/cpu/test_activation.py

@@ -1,9 +1,7 @@
 import itertools
 import unittest
 
-import sgl_kernel
 import torch
-import torch.nn.functional as F
 from utils import GeluAndMul, SiluAndMul, precision
 
 from sglang.srt.server_args import ServerArgs, set_global_server_args_for_scheduler

test/srt/cpu/test_binding.py

@@ -1,7 +1,6 @@
 import re
 import unittest
 
-import sgl_kernel
 import torch
 
 kernel = torch.ops.sgl_kernel

test/srt/cpu/test_comm.py

@@ -5,11 +5,9 @@ import traceback
 import unittest
 from multiprocessing import Process
 
-import sgl_kernel
 import torch
 import torch.distributed as dist
 import torch.multiprocessing as mp
-from utils import precision
 
 from sglang.test.test_utils import CustomTestCase, find_available_port
 

test/srt/cpu/test_decode.py

@@ -1,6 +1,5 @@
 import unittest
 
-import sgl_kernel
 import torch
 from torch.nn.functional import scaled_dot_product_attention
 

test/srt/cpu/test_extend.py

@@ -1,6 +1,5 @@
 import unittest
 
-import sgl_kernel
 import torch
 from torch.nn.functional import scaled_dot_product_attention
 

test/srt/cpu/test_gemm.py

@@ -2,7 +2,6 @@ import itertools
 import unittest
 
 # TODO: use interface in cpu.py
-import sgl_kernel
 import torch
 import torch.nn as nn
 from utils import (

test/srt/cpu/test_mla.py

@@ -1,7 +1,5 @@
-import itertools
 import unittest
 
-import sgl_kernel
 import torch
 from torch.nn.functional import scaled_dot_product_attention
 from utils import precision

test/srt/cpu/test_moe.py

@@ -3,7 +3,6 @@ import math
 import unittest
 
 # TODO: use interface in cpu.py
-import sgl_kernel
 import torch
 
 kernel = torch.ops.sgl_kernel

test/srt/cpu/test_norm.py

@@ -2,7 +2,6 @@ import itertools
 import unittest
 from typing import Optional, Tuple, Union
 
-import sgl_kernel
 import torch
 from utils import make_non_contiguous, precision
 

test/srt/cpu/test_qkv_proj_with_rope.py

@@ -1,6 +1,5 @@
 import unittest
 
-import sgl_kernel
 import torch
 from utils import (
     convert_weight,

test/srt/cpu/test_rope.py

@@ -1,6 +1,5 @@
 import unittest
 
-import sgl_kernel
 import torch
 from utils import precision
 

test/srt/cpu/test_shared_expert.py

@@ -3,9 +3,7 @@ import math
 import unittest
 
 # TODO: use interface in cpu.py
-import sgl_kernel
 import torch
-import torch.nn as nn
 from utils import (
     BLOCK_K,
     BLOCK_N,

test/srt/cpu/test_topk.py

@@ -1,9 +1,6 @@
-import itertools
 import unittest
 
-import sgl_kernel
 import torch
-from utils import precision
 
 from sglang.srt.layers.moe.topk import (
     biased_grouped_topk_impl as native_biased_grouped_topk,

test/srt/entrypoints/http_server/test_abort_request.py

@@ -8,7 +8,6 @@ Run with:
 import threading
 import time
 import unittest
-from typing import Optional
 
 import requests
 

test/srt/ep/test_deepep_intranode.py

@@ -1,6 +1,5 @@
 # Copy from deepseek-ai/DeepEP/tests/test_intranode.py
 
-import os
 import time
 
 # noinspection PyUnresolvedReferences

test/srt/ep/test_deepep_small.py

@@ -317,7 +317,6 @@ class TestMTP(CustomTestCase):
 class TestMTPWithTBO(CustomTestCase):
     @classmethod
     def setUpClass(cls):
-        import os
 
         cls.model = DEFAULT_MODEL_NAME_FOR_TEST_MLA
         cls.base_url = DEFAULT_URL_FOR_TEST

test/srt/ep/test_eplb.py

@@ -1,4 +1,3 @@
-import os
 import tempfile
 import unittest
 from pathlib import Path

test/srt/ep/test_moe_deepep.py

@@ -1,5 +1,4 @@
 import json
-import os
 import unittest
 from types import SimpleNamespace
 

test/srt/experiment_runner.py

@@ -8,8 +8,7 @@ import threading
 import time
 from dataclasses import dataclass
 from datetime import datetime
-from pathlib import Path
-from typing import Dict, List, Optional, Tuple
+from typing import List, Optional, Tuple
 
 import psutil
 import requests

test/srt/function_call/test_json_schema_constraint.py

@@ -12,7 +12,6 @@ from sglang.srt.entrypoints.openai.protocol import (
     ToolChoice,
     ToolChoiceFuncName,
 )
-from sglang.srt.function_call.function_call_parser import FunctionCallParser
 from sglang.srt.function_call.utils import (
     _get_tool_schema_defs,
     get_json_schema_constraint,

test/srt/hicache/test_hicache_storage_3fs_backend.py

@@ -6,13 +6,10 @@ Usage:
 
 import json
 import os
-import time
 import unittest
-from types import SimpleNamespace
 
 from test_hicache_storage_file_backend import HiCacheStorageBaseMixin
 
-from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import CustomTestCase
 
 

test/srt/layers/moe/test_moe_runners.py

@@ -4,11 +4,9 @@ from types import SimpleNamespace
 from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
-    DEFAULT_AWQ_MOE_MODEL_NAME_FOR_TEST,
     DEFAULT_MODEL_NAME_FOR_TEST_FP8_WITH_MOE,
     DEFAULT_MODEL_NAME_FOR_TEST_MOE_NVFP4,
     DEFAULT_MODEL_NAME_FOR_TEST_MXFP4_WITH_MOE,
-    DEFAULT_MODEL_NAME_FOR_TEST_W8A8_WITH_MOE,
     DEFAULT_SMALL_MOE_MODEL_NAME_FOR_TEST_CHAT,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,

test/srt/lora/test_lora.py

@@ -18,7 +18,6 @@ import random
 import unittest
 from typing import List
 
-import torch
 from utils import (
     ALL_OTHER_MULTI_LORA_MODELS,
     CI_MULTI_LORA_MODELS,

test/srt/lora/test_lora_openai_api.py

@@ -6,7 +6,7 @@ that enables OpenAI-compatible LoRA adapter selection.
 """
 
 import unittest
-from unittest.mock import MagicMock, Mock
+from unittest.mock import MagicMock
 
 from sglang.srt.entrypoints.openai.serving_base import OpenAIServingBase
 from sglang.srt.server_args import ServerArgs

test/srt/lora/test_lora_radix_cache.py

@@ -13,13 +13,11 @@
 # ==============================================================================
 
 import multiprocessing as mp
-import random
 import unittest
 
 import torch
-from utils import CI_MULTI_LORA_MODELS, DEFAULT_PROMPTS, run_lora_test_one_by_one
+from utils import CI_MULTI_LORA_MODELS, run_lora_test_one_by_one
 
-from sglang.test.runners import HFRunner, SRTRunner
 from sglang.test.test_utils import CustomTestCase
 
 PROMPTS = [

test/srt/lora/utils.py

@@ -14,7 +14,7 @@
 
 import dataclasses
 import random
-from typing import List, Optional
+from typing import List
 
 import torch
 

test/srt/models/test_clip_models.py

@@ -16,10 +16,8 @@ import multiprocessing as mp
 import unittest
 
 import torch
-from transformers import AutoProcessor
 
-from sglang.srt.utils import load_image
-from sglang.test.runners import DEFAULT_PROMPTS, HFRunner, SRTRunner
+from sglang.test.runners import HFRunner, SRTRunner
 from sglang.test.test_utils import get_similarities
 
 TEXTS = "two Subway Series sandwiches with meats, cheese, lettuce, tomatoes, and onions on a black background, accompanied by the Subway Series logo, highlighting a new sandwich series."

test/srt/models/test_cross_encoder_models.py

@@ -3,7 +3,6 @@ import random
 import unittest
 
 import torch
-from transformers import AutoConfig, AutoTokenizer
 
 from sglang.test.runners import TEST_RERANK_QUERY_DOCS, HFRunner, SRTRunner
 from sglang.test.test_utils import CustomTestCase, is_in_ci

test/srt/models/test_falcon_h1_models.py

@@ -1,4 +1,3 @@
-import unittest
 from types import SimpleNamespace
 
 from sglang.srt.utils import kill_process_tree

test/srt/models/test_generation_models.py

@@ -22,7 +22,6 @@ To test a specific model locally:
 import dataclasses
 import multiprocessing as mp
 import os
-import random
 import unittest
 from typing import List
 

test/srt/models/test_llama4_models.py

@@ -1,4 +1,3 @@
-import random
 import unittest
 from types import SimpleNamespace
 

test/srt/nightly/test_gsm8k_eval_amd.py

@@ -2,7 +2,6 @@ import json
 import os
 import unittest
 import warnings
-from datetime import datetime
 from types import SimpleNamespace
 
 from sglang.srt.utils import kill_process_tree

test/srt/openai_server/basic/test_openai_embedding.py

@@ -1,8 +1,6 @@
 import json
-import os
 import unittest
 
-import numpy as np
 import openai
 
 from sglang.srt.utils import kill_process_tree

test/srt/openai_server/basic/test_openai_server.py

@@ -7,12 +7,10 @@ python3 -m unittest openai_server.basic.test_openai_server.TestOpenAIServer.test
 
 import json
 import random
-import re
 import unittest
 from concurrent.futures import ThreadPoolExecutor
 from typing import Optional
 
-import numpy as np
 import openai
 import requests
 

test/srt/openai_server/basic/test_protocol.py

@@ -13,53 +13,19 @@
 # ==============================================================================
 """Tests for OpenAI API protocol models"""
 
-import json
-import time
 import unittest
-from typing import Dict, List, Optional
+from typing import List, Optional
 
 from pydantic import BaseModel, Field, ValidationError
 
 from sglang.srt.entrypoints.openai.protocol import (
-    BatchRequest,
-    BatchResponse,
-    ChatCompletionMessageContentImagePart,
-    ChatCompletionMessageContentTextPart,
     ChatCompletionRequest,
     ChatCompletionResponse,
     ChatCompletionResponseChoice,
-    ChatCompletionResponseStreamChoice,
-    ChatCompletionStreamResponse,
-    ChatCompletionTokenLogprob,
     ChatMessage,
-    ChoiceLogprobs,
     CompletionRequest,
-    CompletionResponse,
-    CompletionResponseChoice,
-    DeltaMessage,
-    EmbeddingObject,
-    EmbeddingRequest,
-    EmbeddingResponse,
-    ErrorResponse,
-    FileDeleteResponse,
-    FileRequest,
-    FileResponse,
-    Function,
-    FunctionResponse,
-    JsonSchemaResponseFormat,
-    LogProbs,
     ModelCard,
     ModelList,
-    MultimodalEmbeddingInput,
-    ResponseFormat,
-    ScoringRequest,
-    ScoringResponse,
-    StreamOptions,
-    StructuralTagResponseFormat,
-    Tool,
-    ToolCall,
-    ToolChoice,
-    TopLogprob,
     UsageInfo,
 )
 

test/srt/openai_server/basic/test_serving_completions.py

@@ -6,7 +6,7 @@ Run with:
 
 import unittest
 from typing import Optional
-from unittest.mock import AsyncMock, Mock, patch
+from unittest.mock import AsyncMock, Mock
 
 from sglang.srt.entrypoints.openai.protocol import CompletionRequest
 from sglang.srt.entrypoints.openai.serving_completions import OpenAIServingCompletion

test/srt/openai_server/basic/test_serving_embedding.py

@@ -10,7 +10,6 @@ from fastapi import Request
 
 from sglang.srt.entrypoints.openai.protocol import (
     EmbeddingRequest,
-    EmbeddingResponse,
     MultimodalEmbeddingInput,
 )
 from sglang.srt.entrypoints.openai.serving_embedding import OpenAIServingEmbedding

test/srt/openai_server/features/test_cache_report.py

@@ -1,4 +1,3 @@
-import asyncio
 import unittest
 
 import openai

test/srt/openai_server/features/test_enable_thinking.py

@@ -6,18 +6,12 @@ python3 -m unittest openai_server.features.test_enable_thinking.TestEnableThinki
 python3 -m unittest openai_server.features.test_enable_thinking.TestEnableThinking.test_stream_chat_completion_without_reasoning
 """
 
-import asyncio
 import json
-import os
-import sys
-import time
 import unittest
 
-import openai
 import requests
 
 from sglang.srt.utils import kill_process_tree
-from sglang.srt.utils.hf_transformers_utils import get_tokenizer
 from sglang.test.test_utils import (
     DEFAULT_ENABLE_THINKING_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,

test/srt/openai_server/features/test_openai_server_hidden_states.py

@@ -1,19 +1,13 @@
-import json
-import re
-import time
 import unittest
 from abc import ABC
 
-import numpy as np
 import openai
-import torch
 
 from sglang.srt.utils import kill_process_tree
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
 from sglang.test.test_utils import (
     DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
     DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
-    DEFAULT_SMALL_EMBEDDING_MODEL_NAME_FOR_TEST,
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,

test/srt/openai_server/features/test_reasoning_content.py

@@ -9,11 +9,9 @@ python3 -m unittest openai_server.features.test_reasoning_content.TestReasoningC
 python3 -m unittest openai_server.features.test_reasoning_content.TestReasoningContentStartup.test_streaming
 """
 
-import json
 import unittest
 
 import openai
-import requests
 
 from sglang.srt.utils import kill_process_tree
 from sglang.test.test_utils import (

test/srt/openai_server/function_call/test_openai_function_calling.py

@@ -1,5 +1,4 @@
 import json
-import time
 import unittest
 
 import openai

test/srt/quant/test_awq.py

@@ -1,8 +1,6 @@
 import unittest
 from types import SimpleNamespace
 
-import requests
-
 from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (

test/srt/quant/test_triton_scaled_mm.py

@@ -1,4 +1,3 @@
-import itertools
 import unittest
 from typing import Optional
 

test/srt/rl/test_update_weights_from_distributed.py

@@ -22,7 +22,6 @@ import unittest
 import numpy as np
 import requests
 import torch
-import torch.distributed as dist
 import torch.multiprocessing as mp
 from transformers import AutoModelForCausalLM
 

test/srt/rl/test_verl_engine_2_gpu.py

@@ -19,7 +19,6 @@ from torch.distributed.fsdp.api import (
 from transformers import AutoModelForCausalLM
 
 from sglang.srt.entrypoints.verl_engine import VerlEngine
-from sglang.srt.utils import is_port_available
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
 from sglang.test.runners import (
     HFRunner,

test/srt/rl/test_verl_engine_4_gpu.py

@@ -19,7 +19,6 @@ from torch.distributed.fsdp.api import (
 from transformers import AutoModelForCausalLM
 
 from sglang.srt.entrypoints.verl_engine import VerlEngine
-from sglang.srt.utils import is_port_available
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
 from sglang.test.runners import (
     HFRunner,
@@ -125,7 +124,7 @@ class TestVerlEngine(CustomTestCase):
         if is_in_ci():
             return
 
-        for index, model_info in enumerate(ALL_OTHER_MODELS):
+        for index, model_info in enumerate(ALL_MODELS):
             self.assert_fragment_e2e_execution(index=index, **model_info)
 
     # def test_adhoc(self):

test/srt/test_abort.py

@@ -1,6 +1,4 @@
-import json
 import multiprocessing
-import os
 import time
 import unittest
 from concurrent.futures import ThreadPoolExecutor, as_completed

test/srt/test_async_dynamic_batch_tokenizer.py

@@ -8,7 +8,7 @@ including batch efficiency, timeout handling, and error cases.
 import asyncio
 import logging
 import time
-from unittest.mock import AsyncMock, Mock, patch
+from unittest.mock import Mock
 
 import pytest
 from transformers import AutoTokenizer

test/srt/test_bench_one_batch.py

@@ -1,4 +1,3 @@
-import os
 import unittest
 
 from sglang.test.test_utils import (

test/srt/test_bench_serving.py

@@ -4,7 +4,6 @@ import unittest
 
 import requests
 
-from sglang.srt.utils.hf_transformers_utils import get_tokenizer
 from sglang.test.test_utils import (
     DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
     DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,

test/srt/test_bnb.py

@@ -4,19 +4,12 @@ python3 -m unittest test_bnb.TestVisionModel.test_vlm
 python3 -m unittest test_bnb.TestLanguageModel.test_mmlu
 """
 
-import io
-import json
 import multiprocessing as mp
-import os
-import unittest
+import random
 from concurrent.futures import ThreadPoolExecutor
 from types import SimpleNamespace
 
-import numpy as np
 import openai
-import pybase64
-import requests
-from PIL import Image
 
 from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval

test/srt/test_config_integration.py

@@ -4,7 +4,6 @@ Test script to verify SGLang config file integration.
 
 import os
 import tempfile
-from pathlib import Path
 
 import pytest
 import yaml

test/srt/test_cpp_radix_cache.py

@@ -1,4 +1,3 @@
-import os
 import unittest
 from types import SimpleNamespace
 

test/srt/test_create_kvindices.py

@@ -1,4 +1,3 @@
-import itertools
 import unittest
 
 import numpy as np

test/srt/test_deepseek_chat_templates.py

@@ -6,7 +6,6 @@ both dict and string types for tool['function']['arguments'] without double-esca
 addressing issue #11700.
 """
 
-import json
 import os
 import unittest
 

test/srt/test_deepseek_v32_nsabackend.py

@@ -3,7 +3,6 @@ from types import SimpleNamespace
 
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
-from sglang.test.send_one import BenchArgs, send_one_prompt
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,

test/srt/test_deepseek_v3_basic.py

@@ -1,8 +1,6 @@
 import unittest
 from types import SimpleNamespace
 
-import requests
-
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.send_one import BenchArgs, send_one_prompt

test/srt/test_deepseek_v3_cutedsl_4gpu.py

@@ -2,8 +2,6 @@ import os
 import unittest
 from types import SimpleNamespace
 
-import requests
-
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (

test/srt/test_deterministic.py

@@ -9,19 +9,10 @@ test into unit tests so that's easily reproducible in CI.
 
 import unittest
 
-from sglang.srt.utils import kill_process_tree
-from sglang.test.test_deterministic import BenchArgs, test_deterministic
 from sglang.test.test_deterministic_utils import (
     COMMON_SERVER_ARGS,
-    DEFAULT_MODEL,
     TestDeterministicBase,
 )
-from sglang.test.test_utils import (
-    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
-    DEFAULT_URL_FOR_TEST,
-    CustomTestCase,
-    popen_launch_server,
-)
 
 
 class TestFlashinferDeterministic(TestDeterministicBase):

test/srt/test_disaggregation_hybrid_attention.py

@@ -1,8 +1,6 @@
-import os
 import unittest
 from types import SimpleNamespace
 
-from sglang.srt.environ import envs
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_disaggregation_utils import TestDisaggregationBase
 from sglang.test.test_utils import (

test/srt/test_eval_accuracy_large.py

@@ -3,8 +3,6 @@ Usage:
 python -m unittest test_eval_accuracy_large.TestEvalAccuracyLarge.test_mmlu
 """
 
-import os
-import time
 import unittest
 from types import SimpleNamespace
 

test/srt/test_expert_distribution.py

@@ -1,4 +1,3 @@
-import os
 import tempfile
 import unittest
 from pathlib import Path

test/srt/test_flash_attention_4.py

@@ -1,7 +1,6 @@
 import unittest
 from types import SimpleNamespace
 
-from sglang.srt.environ import envs
 from sglang.srt.utils import get_device_sm, kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (

test/srt/test_flashmla.py

@@ -11,15 +11,12 @@ import torch
 
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
-from sglang.test.send_one import BenchArgs, send_one_prompt
 from sglang.test.test_utils import (
     DEFAULT_MODEL_NAME_FOR_TEST_MLA,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
     CustomTestCase,
-    is_in_ci,
     popen_launch_server,
-    write_github_step_summary,
 )
 
 

test/srt/test_fp4_moe.py

@@ -5,12 +5,11 @@ import pytest
 import torch
 from flashinfer import fp4_quantize, scaled_fp4_grouped_quantize
 from flashinfer.fused_moe import cutlass_fused_moe as flashinfer_cutlass_fused_moe
-from sgl_kernel import scaled_fp4_quant
+from sgl_kernel import scaled_fp4_quant, silu_and_mul
 from torch.nn import functional as F
 
 from sglang.srt.layers.moe.cutlass_moe import cutlass_moe_fp4
 from sglang.srt.layers.moe.cutlass_moe_params import CutlassMoEParams, CutlassMoEType
-from sglang.srt.layers.moe.flashinfer_cutedsl_moe import flashinfer_cutedsl_moe_masked
 from sglang.srt.layers.moe.topk import TopKConfig, select_experts
 
 if torch.cuda.get_device_capability() < (10, 0):

test/srt/test_function_call_parser.py

@@ -12,9 +12,6 @@ from sglang.srt.function_call.llama32_detector import Llama32Detector
 from sglang.srt.function_call.mistral_detector import MistralDetector
 from sglang.srt.function_call.pythonic_detector import PythonicDetector
 from sglang.srt.function_call.qwen3_coder_detector import Qwen3CoderDetector
-from sglang.srt.function_call.qwen25_detector import Qwen25Detector
-from sglang.srt.utils.hf_transformers_utils import get_tokenizer
-from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST
 
 
 class TestPythonicDetector(unittest.TestCase):

test/srt/test_fused_moe.py

@@ -1,7 +1,6 @@
 import unittest
 
 import torch
-import torch.nn.functional as F
 from tqdm import tqdm
 
 from sglang.srt.layers.activation import SiluAndMul

test/srt/test_hybrid_attn_backend.py

@@ -1,4 +1,3 @@
-import os
 import unittest
 from types import SimpleNamespace
 

test/srt/test_kv_events.py

@@ -1,7 +1,6 @@
 import time
 import unittest
 
-import msgspec
 import requests
 import zmq
 from msgspec.msgpack import Decoder
@@ -10,8 +9,6 @@ from sglang.srt.disaggregation.kv_events import (
     AllBlocksCleared,
     BlockRemoved,
     BlockStored,
-    EventBatch,
-    KVCacheEvent,
     KVEventBatch,
 )
 from sglang.srt.utils import kill_process_tree

test/srt/test_load_weights_from_remote_instance.py

@@ -21,7 +21,6 @@ import unittest
 import numpy as np
 import requests
 import torch
-import torch.distributed as dist
 import torch.multiprocessing as mp
 
 import sglang as sgl

test/srt/test_mamba_unittest.py

@@ -1,4 +1,3 @@
-import inspect
 import os
 import unittest
 

test/srt/test_modelopt_export.py

@@ -7,7 +7,6 @@ and quantization workflow.
 
 import json
 import os
-import sys
 import tempfile
 import unittest
 from unittest.mock import Mock, patch
@@ -23,7 +22,7 @@ from sglang.srt.model_loader.loader import ModelOptModelLoader
 
 # Check if modelopt is available
 try:
-    import modelopt
+    import modelopt  # noqa: F401
 
     MODELOPT_AVAILABLE = True
 except ImportError:

test/srt/test_modelopt_loader.py

@@ -5,8 +5,6 @@ This test module verifies the functionality of ModelOptModelLoader, which
 applies NVIDIA Model Optimizer quantization to models during loading.
 """
 
-import os
-import sys
 import unittest
 from unittest.mock import MagicMock, patch
 

test/srt/test_mscclpp.py

@@ -8,28 +8,20 @@ else
 fi
 """
 
-import itertools
 import os
 import random
 import socket
 import unittest
-from contextlib import contextmanager, nullcontext
-from typing import Any, List, Optional, Union
+from typing import Any
 
 import ray
 import torch
 import torch.distributed as dist
-from torch.distributed import ProcessGroup, ReduceOp
 
 from sglang.srt.distributed import init_distributed_environment
 from sglang.srt.distributed.communication_op import (  # noqa
     tensor_model_parallel_all_reduce,
 )
-from sglang.srt.distributed.device_communicators.custom_all_reduce import (
-    CustomAllreduce,
-)
-from sglang.srt.distributed.device_communicators.pymscclpp import PyMscclppCommunicator
-from sglang.srt.distributed.device_communicators.pynccl import PyNcclCommunicator
 from sglang.srt.distributed.parallel_state import (
     get_tensor_model_parallel_group,
     graph_capture,
@@ -37,7 +29,6 @@ from sglang.srt.distributed.parallel_state import (
     set_custom_all_reduce,
     set_mscclpp_all_reduce,
 )
-from sglang.srt.distributed.utils import StatelessProcessGroup
 from sglang.test.test_utils import CustomTestCase
 
 

test/srt/test_multi_tokenizer.py

@@ -1,7 +1,6 @@
 import unittest
 from types import SimpleNamespace
 
-import sglang.srt.managers.io_struct as io_struct
 from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (

test/srt/test_original_logprobs.py

@@ -17,7 +17,6 @@ import os
 import random
 import unittest
 
-import numpy as np
 import torch
 import torch.nn.functional as F
 from transformers import AutoModelForCausalLM, AutoTokenizer

test/srt/test_priority_scheduling.py

@@ -2,7 +2,7 @@ import asyncio
 import os
 import re
 import unittest
-from typing import Any, Awaitable, Callable, List, Optional, Tuple
+from typing import Any, List, Optional, Tuple
 
 from sglang.srt.utils import kill_process_tree
 from sglang.test.test_utils import (

test/srt/test_quantization.py

@@ -1,8 +1,6 @@
 import json
-import os
 import unittest
 import warnings
-from datetime import datetime
 from types import SimpleNamespace
 
 from sglang.srt.utils import kill_process_tree

test/srt/test_schedule_policy.py

@@ -6,7 +6,7 @@ from sglang.srt.managers.schedule_policy import (
     CacheAwarePolicy,
     SchedulePolicy,
 )
-from sglang.srt.mem_cache.radix_cache import RadixCache, TreeNode
+from sglang.srt.mem_cache.radix_cache import RadixCache
 from sglang.srt.sampling.sampling_params import SamplingParams
 from sglang.test.test_utils import CustomTestCase
 

test/srt/test_tokenizer_batch_encode.py

@@ -10,12 +10,10 @@ python3 -m unittest test_tokenizer_batch_encode.TestTokenizerBatchEncodeUnit.tes
 python3 -m unittest test_tokenizer_batch_encode.TestTokenizerBatchEncodeLogic.test_batch_processing_path
 """
 
-import asyncio
 import unittest
-from typing import List
-from unittest.mock import AsyncMock, Mock, call, patch
+from unittest.mock import Mock, patch
 
-from sglang.srt.managers.io_struct import GenerateReqInput, TokenizedGenerateReqInput
+from sglang.srt.managers.io_struct import GenerateReqInput
 from sglang.srt.managers.tokenizer_manager import TokenizerManager
 from sglang.srt.server_args import PortArgs, ServerArgs
 from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST

test/srt/test_tokenizer_manager.py

@@ -12,7 +12,6 @@ python3 -m unittest test_tokenizer_manager.TestTokenizerManagerIntegration
 """
 
 import unittest
-from typing import List, Optional, Union
 from unittest.mock import Mock, patch
 
 from sglang.srt.managers.tokenizer_manager import TokenizerManager

test/srt/test_tracing.py

@@ -10,7 +10,6 @@ import requests
 import zmq
 
 from sglang import Engine
-from sglang.srt.managers.io_struct import TokenizedGenerateReqInput
 from sglang.srt.tracing.trace import *
 from sglang.srt.utils import get_zmq_socket, kill_process_tree
 from sglang.test.test_utils import (

test/srt/test_triton_fused_moe.py

@@ -1,7 +1,6 @@
 import unittest
 
 import torch
-import torch.nn.functional as F
 from tqdm import tqdm
 
 from sglang.srt.layers.activation import SiluAndMul

test/srt/test_vision_openai_server_a.py

@@ -207,11 +207,12 @@ class TestDeepseekOCRServer(TestOpenAIMLLMServerBase):
 
 
 if __name__ == "__main__":
-    del (
-        TestOpenAIMLLMServerBase,
-        ImageOpenAITestMixin,
-        VideoOpenAITestMixin,
-        AudioOpenAITestMixin,
-        OmniOpenAITestMixin,
-    )
+    # Note: Cannot delete mixin classes imported via * since they're not in local scope
+    # del (
+    #     TestOpenAIMLLMServerBase,
+    #     ImageOpenAITestMixin,
+    #     VideoOpenAITestMixin,
+    #     AudioOpenAITestMixin,
+    #     OmniOpenAITestMixin,
+    # )
     unittest.main()

test/srt/test_vlm_input_format.py

@@ -7,7 +7,6 @@ import requests
 import torch
 from PIL import Image
 from transformers import (
-    AutoModel,
     AutoProcessor,
     Gemma3ForConditionalGeneration,
     Qwen2_5_VLForConditionalGeneration,

test/srt/xpu/test_intel_xpu_backend.py

@@ -3,7 +3,6 @@ Usage:
 python3 -m unittest test_intel_xpu_backend.TestIntelXPUBackend.test_latency_qwen_model
 """
 
-import os
 import unittest
 from functools import wraps