[Benchmark] Remove re-exports from bench_serving.py (#19130)

python/sglang/bench_offline_throughput.py

@@ -23,13 +23,8 @@ from typing import Dict, List, Optional
 
 import numpy as np
 
-from sglang.bench_serving import (
-    DatasetRow,
-    get_dataset,
-    get_tokenizer,
-    sample_random_requests,
-    set_ulimit,
-)
+from sglang.benchmark.datasets import DatasetRow, get_dataset, sample_random_requests
+from sglang.benchmark.utils import get_tokenizer, set_ulimit
 from sglang.lang.backend.runtime_endpoint import Runtime
 from sglang.srt.entrypoints.engine import Engine
 from sglang.srt.server_args import ServerArgs

python/sglang/bench_serving.py

@@ -36,39 +36,15 @@ import requests
 from tqdm.asyncio import tqdm
 from transformers import AutoTokenizer, PreTrainedTokenizerBase
 
-from sglang.benchmark.datasets import (  # noqa: F401
-    ASSISTANT_SUFFIX,
-    MOONCAKE_DATASET_URL,
-    SHAREGPT_FILENAME,
-    SHAREGPT_REPO_ID,
+from sglang.benchmark.datasets import (
     DatasetRow,
-    compute_random_lens,
-    create_mm_data_row,
-    gen_mm_prompt,
-    gen_prompt,
-    get_available_tokens,
     get_dataset,
-    get_gen_prefix_cache_path,
     get_mooncake_request_over_time,
-    parse_image_resolution,
-    sample_custom_requests,
-    sample_generated_shared_prefix_requests,
-    sample_image_requests,
-    sample_mmmu_requests,
-    sample_openai_requests,
-    sample_random_requests,
-    sample_sharegpt_requests,
 )
-from sglang.benchmark.utils import (  # noqa: F401
-    download_and_cache_file,
-    download_and_cache_hf_file,
-    get_model,
-    get_processor,
+from sglang.benchmark.utils import (
     get_tokenizer,
-    is_file_valid_json,
     parse_custom_headers,
     remove_prefix,
-    remove_suffix,
     set_ulimit,
 )
 

python/sglang/test/bench_one_batch_server_internal.py

@@ -16,11 +16,8 @@ from pydantic import BaseModel
 from tabulate import tabulate
 from transformers import AutoProcessor, PreTrainedTokenizer
 
-from sglang.bench_serving import (
-    get_dataset,
-    get_processor,
-    get_tokenizer,
-)
+from sglang.benchmark.datasets import get_dataset
+from sglang.benchmark.utils import get_processor, get_tokenizer
 from sglang.profiler import run_profile
 from sglang.srt.entrypoints.http_server import launch_server
 from sglang.srt.server_args import ServerArgs

python/sglang/test/kits/cache_hit_kit.py

@@ -5,12 +5,9 @@ import time
 import aiohttp
 import requests
 
-from sglang.bench_serving import (
-    RequestFuncOutput,
-    get_tokenizer,
-    remove_prefix,
-    sample_random_requests,
-)
+from sglang.bench_serving import RequestFuncOutput
+from sglang.benchmark.datasets import sample_random_requests
+from sglang.benchmark.utils import get_tokenizer, remove_prefix
 
 AIOHTTP_TIMEOUT = aiohttp.ClientTimeout(total=20 * 60 * 60)