Revert removing the unused imports (#385)

python/sglang/srt/models/yivl.py

 """Inference-only Yi-VL model."""
 
-from typing import Optional
+import os
+from typing import List, Optional
 
 import torch
 import torch.nn as nn
@@ -12,6 +13,7 @@ from vllm.model_executor.weight_utils import (
 
 from sglang.srt.models.llava import (
     LlavaLlamaForCausalLM,
+    clip_vision_embed_forward,
     monkey_path_clip_vision_embed_forward,
 )
 

python/sglang/srt/server.py

@@ -10,6 +10,9 @@ import threading
 import time
 from typing import List, Optional, Union
 
+# Fix a Python bug
+setattr(threading, "_register_atexit", lambda *args, **kwargs: None)
+
 import aiohttp
 import psutil
 import pydantic
@@ -55,9 +58,6 @@ from sglang.srt.managers.tokenizer_manager import TokenizerManager
 from sglang.srt.server_args import PortArgs, ServerArgs
 from sglang.srt.utils import enable_show_time_cost, handle_port_init
 
-# Fix a Python bug
-setattr(threading, "_register_atexit", lambda *args, **kwargs: None)
-
 asyncio.set_event_loop_policy(uvloop.EventLoopPolicy())
 
 API_KEY_HEADER_NAME = "X-API-Key"
@@ -619,7 +619,7 @@ def launch_server(server_args, pipe_finish_writer):
             try:
                 requests.get(url + "/get_model_info", timeout=5, headers=headers)
                 break
-            except requests.exceptions.RequestException:
+            except requests.exceptions.RequestException as e:
                 pass
         else:
             if pipe_finish_writer is not None:

python/sglang/srt/utils.py

@@ -157,6 +157,7 @@ def get_exception_traceback():
 
 
 def get_int_token_logit_bias(tokenizer, vocab_size):
+    from transformers import LlamaTokenizer, LlamaTokenizerFast
 
     # a bug when model's vocab size > tokenizer.vocab_size
     vocab_size = tokenizer.vocab_size

test/lang/run_all.py

 import argparse
 import glob
 import multiprocessing
+import os
 import time
 import unittest
 

test/lang/test_anthropic_backend.py

+import json
 import unittest
 
 from sglang import Anthropic, set_default_backend

test/lang/test_srt_backend.py

@@ -2,6 +2,7 @@
 python3 -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port 30000
 """
 
+import json
 import unittest
 
 import sglang as sgl
@@ -12,6 +13,8 @@ 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/lang/test_tracing.py

@@ -110,7 +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("detailed_tip")
+                forks[i] += sgl.gen(f"detailed_tip")
 
             s += "Tip 1:" + forks[0]["detailed_tip"] + "\n"
             s += "Tip 2:" + forks[1]["detailed_tip"] + "\n"

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

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.model_runner import ModelRunner
+from sglang.srt.managers.router.infer_batch import ForwardMode
+from sglang.srt.managers.router.model_runner import InputMetadata, ModelRunner
 from sglang.srt.model_config import ModelConfig
 from sglang.srt.utils import load_image
 

test/srt/test_httpserver_concurrent.py

@@ -9,8 +9,11 @@ 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

@@ -10,6 +10,7 @@ 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

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