[Bugfix] Restore user config/runtime stage init timeout (#2519)

Signed-off-by: yuanheng <jonathan.zhaoyh@gmail.com>
Signed-off-by: Yuanheng Zhao <jonathan.zhaoyh@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Co-authored-by: SYLAR <125541396+lishunyang12@users.noreply.github.com>
This commit is contained in:
authored and GitHub committed 2026-04-12 00:48:47 +08:00
1 parent 6e93595822
commit c20cac86ce
16 files changed
+207 -29

No files matched your search

+3 -2
View File
@@ -114,8 +114,8 @@ class AccuracyServerConfig:
params = self.generate_params
model = self.model_prefix + params.model
server_args = params.server_args or []
if params.use_omni:
server_args = ["--stage-init-timeout", "120", *server_args]
if params.use_omni and params.stage_init_timeout is not None:
server_args = ["--stage-init-timeout", str(params.stage_init_timeout), *server_args]
with OmniServer(
model,
server_args,
@@ -226,6 +226,7 @@ def _build_accuracy_server_config(
server_args=generate_server_args,
env_dict={"CUDA_VISIBLE_DEVICES": shared_gpu},
use_omni=True,
stage_init_timeout=300,
),
judge_params=OmniServerParams(
model=judge_model,
@@ -47,7 +47,7 @@ test_params = [
OmniServerParams(
model=MODEL,
stage_config_path=STAGE_CONFIGS_PATH,
server_args=["--stage-init-timeout", "300"],
stage_init_timeout=300,
),
]
@@ -30,7 +30,7 @@ T2I_PROMPT = "A high quality detailed living room interior photo."
T2S_PROMPT = "Please read this sentence naturally: Hello from Dynin-Omni online serving."
I2I_PROMPT = "Transform this outdoor nature boardwalk scene into a painting style with vivid colors."
TEST_PARAMS = [OmniServerParams(model=MODEL, stage_config_path=STAGE_CONFIG)]
TEST_PARAMS = [OmniServerParams(model=MODEL, stage_config_path=STAGE_CONFIG, stage_init_timeout=600)]
_STAGE_COUNT = 3
_I2I_STAGE_SAMPLING = {"max_tokens": 1, "temperature": 0.0, "top_p": 1.0, "detokenize": False}
@@ -30,6 +30,7 @@ from tests.conftest import (
from tests.utils import hardware_test
MODEL = "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice"
STAGE_INIT_TIMEOUT_S = 120
def get_stage_config(name: str = "qwen3_tts.yaml"):
@@ -47,7 +48,7 @@ def omni_server():
"--stage-configs-path",
stage_config_path,
"--stage-init-timeout",
"120",
str(STAGE_INIT_TIMEOUT_S),
"--trust-remote-code",
"--enforce-eager",
"--disable-log-stats",
@@ -337,7 +338,7 @@ def omni_server_batch2():
"--stage-configs-path",
config_path,
"--stage-init-timeout",
"120",
str(STAGE_INIT_TIMEOUT_S),
"--trust-remote-code",
"--enforce-eager",
"--disable-log-stats",
@@ -23,6 +23,7 @@ from tests.utils import hardware_test
MODEL_BASE = "Qwen/Qwen3-TTS-12Hz-0.6B-Base"
MODEL_BASE_1_7B = "Qwen/Qwen3-TTS-12Hz-1.7B-Base"
STAGE_INIT_TIMEOUT_S = 120
# A synthetic 1024-dim speaker embedding (all 0.1 — not a real voice, but
# exercises the full code path through the talker's _build_prompt_embeds).
@@ -47,7 +48,7 @@ def _server_args():
"--stage-configs-path",
get_stage_config(),
"--stage-init-timeout",
"120",
str(STAGE_INIT_TIMEOUT_S),
"--trust-remote-code",
"--enforce-eager",
"--disable-log-stats",
@@ -19,6 +19,7 @@ os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
os.environ["VLLM_TEST_CLEAN_GPU_MEMORY"] = "0"
MODEL = "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice"
STAGE_INIT_TIMEOUT_S = 120
def get_stage_config() -> str:
@@ -37,7 +38,7 @@ def omni_server():
"--stage-configs-path",
stage_config_path,
"--stage-init-timeout",
"120",
str(STAGE_INIT_TIMEOUT_S),
"--trust-remote-code",
"--enforce-eager",
"--disable-log-stats",