EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[CI] Fix ci tests (#5769)

Browse Source
This commit is contained in:
Lianmin Zheng
2025-04-27 07:18:10 -07:00
committed by GitHub
parent 35ca04d2fa
commit 621e96bf9b
18 changed files with 126 additions and 295 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
python/sglang/test/test_utils.py
View File

@@ -33,33 +33,44 @@ from sglang.srt.utils import (
from sglang.test.run_eval import run_eval
from sglang.utils import get_exception_traceback
DEFAULT_FP8_MODEL_NAME_FOR_TEST = "neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8"
DEFAULT_FP8_MODEL_NAME_FOR_ACCURACY_TEST = "neuralmagic/Meta-Llama-3-8B-Instruct-FP8"
DEFAULT_FP8_MODEL_NAME_FOR_DYNAMIC_QUANT_ACCURACY_TEST = (
# General test models
DEFAULT_MODEL_NAME_FOR_TEST = "meta-llama/Llama-3.1-8B-Instruct"
DEFAULT_SMALL_MODEL_NAME_FOR_TEST = "meta-llama/Llama-3.2-1B-Instruct"
DEFAULT_MOE_MODEL_NAME_FOR_TEST = "mistralai/Mixtral-8x7B-Instruct-v0.1"
DEFAULT_SMALL_MOE_MODEL_NAME_FOR_TEST = "Qwen/Qwen1.5-MoE-A2.7B"
# MLA test models
DEFAULT_MLA_MODEL_NAME_FOR_TEST = "deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct"
DEFAULT_MLA_FP8_MODEL_NAME_FOR_TEST = "neuralmagic/DeepSeek-Coder-V2-Lite-Instruct-FP8"
DEFAULT_MODEL_NAME_FOR_TEST_MLA = "lmsys/sglang-ci-dsv3-test"
DEFAULT_MODEL_NAME_FOR_TEST_MLA_NEXTN = "lmsys/sglang-ci-dsv3-test-NextN"
# FP8 models
DEFAULT_MODEL_NAME_FOR_TEST_FP8 = "neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8"
DEFAULT_MODEL_NAME_FOR_ACCURACY_TEST_FP8 = "neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8"
DEFAULT_MODEL_NAME_FOR_DYNAMIC_QUANT_ACCURACY_TEST_FP8 = (
"neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8-dynamic"
)
DEFAULT_FP8_MODEL_NAME_FOR_MODELOPT_QUANT_ACCURACY_TEST = (
DEFAULT_MODEL_NAME_FOR_MODELOPT_QUANT_ACCURACY_TEST_FP8 = (
"nvidia/Llama-3.1-8B-Instruct-FP8"
)
DEFAULT_MODEL_NAME_FOR_TEST = "meta-llama/Llama-3.1-8B-Instruct"
# EAGLE
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST = "meta-llama/Llama-2-7b-chat-hf"
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST = "lmsys/sglang-EAGLE-llama2-chat-7B"
DEFAULT_MODEL_NAME_FOR_TEST_EAGLE3 = "jamesliu1/sglang-EAGLE3-Llama-3.1-Instruct-8B"
DEFAULT_MODEL_NAME_FOR_TEST_MLA = "lmsys/sglang-ci-dsv3-test"
DEFAULT_MODEL_NAME_FOR_TEST_MLA_NEXTN = "lmsys/sglang-ci-dsv3-test-NextN"
DEFAULT_SMALL_MODEL_NAME_FOR_TEST = "meta-llama/Llama-3.2-1B-Instruct"
# Other use cases
DEFAULT_MODEL_NAME_FOR_TEST_LOCAL_ATTENTION = (
"meta-llama/Llama-4-Scout-17B-16E-Instruct"
)
DEFAULT_MOE_MODEL_NAME_FOR_TEST = "mistralai/Mixtral-8x7B-Instruct-v0.1"
DEFAULT_SMALL_MOE_MODEL_NAME_FOR_TEST = "Qwen/Qwen1.5-MoE-A2.7B"
DEFAULT_SMALL_EMBEDDING_MODEL_NAME_FOR_TEST = "Alibaba-NLP/gte-Qwen2-1.5B-instruct"
DEFAULT_MLA_MODEL_NAME_FOR_TEST = "deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct"
DEFAULT_MLA_FP8_MODEL_NAME_FOR_TEST = "neuralmagic/DeepSeek-Coder-V2-Lite-Instruct-FP8"
DEFAULT_REASONING_MODEL_NAME_FOR_TEST = "deepseek-ai/DeepSeek-R1-Distill-Qwen-7B"
DEFAULT_AWQ_MOE_MODEL_NAME_FOR_TEST = (
"hugging-quants/Mixtral-8x7B-Instruct-v0.1-AWQ-INT4"
)
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH = 1000
# Nightly tests
DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_TP1 = "meta-llama/Llama-3.1-8B-Instruct,mistralai/Mistral-7B-Instruct-v0.3,deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct,google/gemma-2-27b-it"
DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_TP2 = "meta-llama/Llama-3.1-70B-Instruct,mistralai/Mixtral-8x7B-Instruct-v0.1,Qwen/Qwen2-57B-A14B-Instruct"
DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_FP8_TP1 = "neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8,neuralmagic/Mistral-7B-Instruct-v0.3-FP8,neuralmagic/DeepSeek-Coder-V2-Lite-Instruct-FP8,neuralmagic/gemma-2-2b-it-FP8"
@@ -68,12 +79,11 @@ DEFAULT_MODEL_NAME_FOR_NIGHTLY_EVAL_QUANT_TP1 = "hugging-quants/Meta-Llama-3.1-8
DEFAULT_SMALL_MODEL_NAME_FOR_TEST_QWEN = "Qwen/Qwen2.5-1.5B-Instruct"
DEFAULT_SMALL_VLM_MODEL_NAME = "Qwen/Qwen2-VL-2B"
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST = "meta-llama/Llama-2-7b-chat-hf"
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST = "lmsys/sglang-EAGLE-llama2-chat-7B"
DEFAULT_IMAGE_URL = "https://github.com/sgl-project/sglang/blob/main/test/lang/example_image.png?raw=true"
DEFAULT_VIDEO_URL = "https://raw.githubusercontent.com/EvolvingLMMs-Lab/sglang/dev/onevision_local/assets/jobs.mp4"
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH = 1000
def is_in_ci():
"""Return whether it is in CI runner."""
@@ -499,7 +509,7 @@ def run_unittest_files(files: List[TestFile], timeout_per_file: float):
tic = time.time()
success = True
for file in files:
for i, file in enumerate(files):
filename, estimated_time = file.name, file.estimated_time
process = None
@@ -507,7 +517,10 @@ def run_unittest_files(files: List[TestFile], timeout_per_file: float):
nonlocal process
filename = os.path.join(os.getcwd(), filename)
print(f".\n.\nBegin:\npython3 {filename}\n.\n.\n", flush=True)
print(
f".\n.\nBegin ({i}/{len(files)}):\npython3 {filename}\n.\n.\n",
flush=True,
)
tic = time.time()
process = subprocess.Popen(
@@ -517,7 +530,7 @@ def run_unittest_files(files: List[TestFile], timeout_per_file: float):
elapsed = time.time() - tic
print(
f".\n.\nEnd:\n{filename=}, {elapsed=:.0f}, {estimated_time=}\n.\n.\n",
f".\n.\nEnd ({i}/{len(files)}):\n{filename=}, {elapsed=:.0f}, {estimated_time=}\n.\n.\n",
flush=True,
)
return process.returncode