add `.isort.cfg` (#378)

test/lang/test_anthropic_backend.py

-import json
 import unittest
 
-from sglang.test.test_programs import test_mt_bench, test_stream
-
 from sglang import Anthropic, set_default_backend
+from sglang.test.test_programs import test_mt_bench, test_stream
 
 
 class TestAnthropicBackend(unittest.TestCase):

test/lang/test_bind_pin.py

 import unittest
 
-from sglang.backend.runtime_endpoint import RuntimeEndpoint
-
 import sglang as sgl
+from sglang.backend.runtime_endpoint import RuntimeEndpoint
 
 
 class TestBind(unittest.TestCase):

test/lang/test_openai_backend.py

 import unittest
 
+from sglang import OpenAI, set_default_backend
 from sglang.test.test_programs import (
     test_decode_int,
     test_decode_json,
@@ -15,8 +16,6 @@ from sglang.test.test_programs import (
     test_tool_use,
 )
 
-from sglang import OpenAI, set_default_backend
-
 
 class TestOpenAIBackend(unittest.TestCase):
     backend = None

test/lang/test_srt_backend.py

@@ -2,9 +2,9 @@
 python3 -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port 30000
 """
 
-import json
 import unittest
 
+import sglang as sgl
 from sglang.test.test_programs import (
     test_decode_int,
     test_decode_json_regex,
@@ -12,16 +12,12 @@ from sglang.test.test_programs import (
     test_few_shot_qa,
     test_mt_bench,
     test_parallel_decoding,
-    test_parallel_encoding,
-    test_react,
     test_regex,
     test_select,
     test_stream,
     test_tool_use,
 )
 
-import sglang as sgl
-
 
 class TestSRTBackend(unittest.TestCase):
     backend = None

test/lang/test_tracing.py

 import unittest
 
+import sglang as sgl
 from sglang.backend.base_backend import BaseBackend
 from sglang.lang.chat_template import get_chat_template
 
-import sglang as sgl
-
 
 class TestTracing(unittest.TestCase):
     def test_few_shot_qa(self):
@@ -111,7 +110,7 @@ class TestTracing(unittest.TestCase):
             forks = s.fork(3)
             for i in range(3):
                 forks[i] += f"Now, expand tip {i+1} into a paragraph:\n"
-                forks[i] += sgl.gen(f"detailed_tip")
+                forks[i] += sgl.gen("detailed_tip")
 
             s += "Tip 1:" + forks[0]["detailed_tip"] + "\n"
             s += "Tip 2:" + forks[1]["detailed_tip"] + "\n"

test/lang/test_vertexai_backend.py

 import unittest
 
+from sglang import VertexAI, set_default_backend
 from sglang.test.test_programs import (
     test_expert_answer,
     test_few_shot_qa,
@@ -10,8 +11,6 @@ from sglang.test.test_programs import (
     test_stream,
 )
 
-from sglang import VertexAI, set_default_backend
-
 
 class TestVertexAIBackend(unittest.TestCase):
     backend = None

test/srt/model/bench_llama_low_api.py

@@ -4,6 +4,7 @@ from dataclasses import dataclass
 
 import torch
 import torch.distributed as dist
+
 from sglang.srt.managers.router.model_runner import ModelRunner
 from sglang.srt.model_config import ModelConfig
 
@@ -66,9 +67,9 @@ class BenchBatch:
             p_idx = prefix_req_idx[i // fork_num].item()
             n_idx = self.req_pool_indices[i].item()
             req_to_token[n_idx, :prefix_len] = req_to_token[p_idx, :prefix_len]
-            req_to_token[
-                n_idx, prefix_len : prefix_len + extend_len
-            ] = self.out_cache_loc[i * extend_len : (i + 1) * extend_len]
+            req_to_token[n_idx, prefix_len : prefix_len + extend_len] = (
+                self.out_cache_loc[i * extend_len : (i + 1) * extend_len]
+            )
 
     def update_decode(self, predict_ids, batch_size):
         assert predict_ids.shape[0] == batch_size
@@ -81,9 +82,9 @@ class BenchBatch:
             self.out_cache_cont_start,
             self.out_cache_cont_end,
         ) = self.token_to_kv_pool.alloc_contiguous(batch_size)
-        self.req_to_token_pool.req_to_token[
-            self.req_pool_indices, self.seq_lens
-        ] = self.out_cache_loc
+        self.req_to_token_pool.req_to_token[self.req_pool_indices, self.seq_lens] = (
+            self.out_cache_loc
+        )
         self.seq_lens.add_(1)
 
 

test/srt/model/reference_hf.py

 import argparse
-import os
 
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer

test/srt/model/test_llama_extend.py

 import multiprocessing
 import os
-import time
 
-import numpy as np
-import torch
-import torch.distributed as dist
 import transformers
+
 from sglang.srt.managers.router.infer_batch import Batch, ForwardMode, Req
 from sglang.srt.managers.router.model_runner import ModelRunner
 from sglang.srt.model_config import ModelConfig

test/srt/model/test_llama_low_api.py

@@ -4,6 +4,7 @@ import time
 import numpy as np
 import torch
 import torch.distributed as dist
+
 from sglang.srt.managers.router.model_runner import ModelRunner
 from sglang.srt.model_config import ModelConfig
 

test/srt/model/test_llava_low_api.py

 import multiprocessing
-import time
 
 import numpy as np
 import torch
-import torch.distributed as dist
+
 from sglang.srt.hf_transformers_utils import get_processor
-from sglang.srt.managers.router.infer_batch import ForwardMode
-from sglang.srt.managers.router.model_runner import InputMetadata, ModelRunner
+from sglang.srt.managers.router.model_runner import ModelRunner
 from sglang.srt.model_config import ModelConfig
 from sglang.srt.utils import load_image
 

test/srt/test_flashinfer.py

 import flashinfer
 import pytest
 import torch
+
 from sglang.srt.layers.extend_attention import extend_attention_fwd
 from sglang.srt.layers.token_attention import token_attention_fwd
 

test/srt/test_httpserver_concurrent.py

@@ -9,11 +9,8 @@ The capital of the United Kindom is London.\nThe capital of the United Kingdom i
 
 import argparse
 import asyncio
-import json
-import time
 
 import aiohttp
-import requests
 
 
 async def send_request(url, data, delay=0):

test/srt/test_httpserver_llava.py

@@ -10,7 +10,6 @@ The image features a man standing on the back of a yellow taxi cab, holding
 import argparse
 import asyncio
 import json
-import time
 
 import aiohttp
 import requests

test/srt/test_httpserver_reuse.py

@@ -6,7 +6,6 @@ The capital of France is Paris.\nThe capital of the United States is Washington,
 """
 
 import argparse
-import time
 
 import requests
 

test/srt/test_jump_forward.py

@@ -2,14 +2,14 @@ import argparse
 from enum import Enum
 
 from pydantic import BaseModel, constr
+
+import sglang as sgl
 from sglang.srt.constrained import build_regex_from_object
 from sglang.test.test_utils import (
     add_common_sglang_args_and_parse,
     select_sglang_backend,
 )
 
-import sglang as sgl
-
 IP_REGEX = r"((25[0-5]|2[0-4]\d|[01]?\d\d?)\.){3}(25[0-5]|2[0-4]\d|[01]?\d\d?)"
 
 ip_jump_forward = (

test/srt/test_robust.py

@@ -2,13 +2,13 @@ import argparse
 import random
 import string
 
+from vllm.transformers_utils.tokenizer import get_tokenizer
+
+import sglang as sgl
 from sglang.test.test_utils import (
     add_common_sglang_args_and_parse,
     select_sglang_backend,
 )
-from vllm.transformers_utils.tokenizer import get_tokenizer
-
-import sglang as sgl
 
 TOKENIZER = None
 RANDOM_PREFILL_LEN = None