chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

CI: use 'sglang serve' in CI tests (#18597)

Browse Source 
Co-authored-by: Mick <mickjagger19@icloud.com>
Co-authored-by: sglang-bot <sglangbot@gmail.com>
This commit is contained in:
yrk111222
2026-02-28 06:00:41 +08:00
committed by GitHub
parent 776709efe8
commit e6da514c2c
3 changed files with 47 additions and 37 deletions
Download Patch File Download Diff File Expand all files Collapse all files

40
python/sglang/test/test_utils.py
View File

@@ -879,18 +879,22 @@ def popen_launch_server(
use_mixed_pd_engine = not pd_separated and num_replicas is not None
if pd_separated or use_mixed_pd_engine:
command = "sglang.launch_pd_server"
command = [
"python3",
"-m",
"sglang.launch_pd_server",
"--model-path",
model,
*[str(x) for x in other_args],
]
else:
command = "sglang.launch_server"
command = [
"python3",
"-m",
command,
"--model-path",
model,
*[str(x) for x in other_args],
]
command = [
"sglang",
"serve",
"--model-path",
model,
*[str(x) for x in other_args],
]
if pd_separated or use_mixed_pd_engine:
command.extend(["--lb-host", host, "--lb-port", port])
@@ -1308,7 +1312,7 @@ def run_score_benchmark(
json=warmup_data,
timeout=aiohttp.ClientTimeout(total=30),
)
except:
except Exception:
pass # Ignore warmup errors
test_requests = []
@@ -1376,17 +1380,9 @@ def run_embeddings_benchmark(
async def _run_benchmark():
# Load tokenizer for generating test data
from sglang.srt.utils.hf_transformers_utils import get_tokenizer
tokenizer = get_tokenizer(model)
def generate_text_with_token_count(num_tokens):
"""Generate text with precise token count using special tokens."""
# Use a token that reliably produces 1 token
special_token = "<|im_start|>"
# Verify it's a single token
test_tokens = tokenizer.encode(special_token, add_special_tokens=False)
text = special_token * num_tokens
return text
@@ -1406,7 +1402,7 @@ def run_embeddings_benchmark(
json=warmup_data,
timeout=aiohttp.ClientTimeout(total=30),
)
except:
except Exception:
pass # Ignore warmup errors
test_requests = []
@@ -1822,7 +1818,7 @@ def run_mulit_request_test(
},
},
)
ret = response.json()
response.json()
with ThreadPoolExecutor(2) as executor:
list(executor.map(run_one, list(range(4))))