[CI] Modify some CI test cases to run on L4 environment to reduce H100 resource usage. (#1543)
Signed-off-by: yenuo26 <410167048@qq.com> Signed-off-by: wangyu <53896905+yenuo26@users.noreply.github.com>
This commit is contained in:
1 parent
b0156d8e1c
commit
cd2234a1bf
13 files changed
+103
-129
No files matched your search
@@ -3,36 +3,18 @@ from pathlib import Path
|
||||
|
||||
import pytest
|
||||
|
||||
from tests.conftest import OmniServer
|
||||
from tests.utils import hardware_test
|
||||
|
||||
models = ["Qwen/Qwen3-Omni-30B-A3B-Instruct"]
|
||||
stage_configs = [str(Path(__file__).parent.parent / "e2e" / "stage_configs" / "qwen3_omni_ci.yaml")]
|
||||
models = ["Qwen/Qwen2.5-Omni-7B"]
|
||||
stage_configs = [str(Path(__file__).parent.parent / "e2e" / "stage_configs" / "qwen2_5_omni_ci.yaml")]
|
||||
|
||||
# Create parameter combinations for model and stage config
|
||||
test_params = [(model, stage_config) for model in models for stage_config in stage_configs]
|
||||
|
||||
|
||||
@pytest.fixture(scope="module")
|
||||
def omni_server(request):
|
||||
"""Start vLLM-Omni server as a subprocess with actual model weights.
|
||||
Uses session scope so the server starts only once for the entire test session.
|
||||
Multi-stage initialization can take 10-20+ minutes.
|
||||
"""
|
||||
model, stage_config_path = request.param
|
||||
|
||||
print(f"Starting OmniServer with model: {model}")
|
||||
print("This may take 10-20+ minutes for initialization...")
|
||||
|
||||
with OmniServer(model, ["--stage-configs-path", stage_config_path, "--stage-init-timeout", "120"]) as server:
|
||||
print("OmniServer started successfully")
|
||||
yield server
|
||||
print("OmniServer stopped")
|
||||
|
||||
|
||||
@pytest.mark.core_model
|
||||
@pytest.mark.benchmark
|
||||
@hardware_test(res={"cuda": "H100"}, num_cards=2)
|
||||
@hardware_test(res={"cuda": "L4"}, num_cards=3)
|
||||
@pytest.mark.parametrize("omni_server", test_params, indirect=True)
|
||||
def test_bench_serve_chat(omni_server):
|
||||
command = [
|
||||
|
||||
@@ -71,7 +71,7 @@ def get_system_prompt():
|
||||
|
||||
def get_prompt(prompt_type="text_only"):
|
||||
prompts = {
|
||||
"text_only": "What is the capital of China?",
|
||||
"text_only": "What is the capital of China? Answer in 20 words.",
|
||||
"mix": "What is recited in the audio? What is in this image? What is in this video?",
|
||||
"text_video": "What is in this video? ",
|
||||
"text_image": "What is in this image? ",
|
||||
|
||||
@@ -22,6 +22,7 @@ stage_args:
|
||||
trust_remote_code: true
|
||||
engine_output_type: latent
|
||||
enable_prefix_caching: false
|
||||
mm_processor_cache_gb: 0
|
||||
is_comprehension: true
|
||||
final_output: true
|
||||
final_output_type: text
|
||||
|
||||
@@ -0,0 +1,31 @@
|
||||
stage_args:
|
||||
- stage_id: 0
|
||||
runtime:
|
||||
process: true # Run this stage in a separate process
|
||||
devices: "0" # Visible devices for this stage (CUDA_VISIBLE_DEVICES/torch.cuda.set_device)
|
||||
max_batch_size: 1
|
||||
engine_args:
|
||||
model_stage: thinker
|
||||
model_arch: Qwen2_5OmniForConditionalGeneration
|
||||
worker_type: ar
|
||||
scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler
|
||||
max_model_len: 16384
|
||||
max_num_batched_tokens: 16384
|
||||
max_num_seqs: 1
|
||||
gpu_memory_utilization: 0.9
|
||||
skip_mm_profiling: true
|
||||
enforce_eager: true # Now we only support eager mode
|
||||
trust_remote_code: true
|
||||
engine_output_type: latent
|
||||
enable_prefix_caching: false
|
||||
is_comprehension: true
|
||||
final_output: true
|
||||
final_output_type: text
|
||||
default_sampling_params:
|
||||
temperature: 0.0
|
||||
top_p: 1.0
|
||||
top_k: -1
|
||||
max_tokens: 128
|
||||
seed: 42
|
||||
detokenize: True
|
||||
repetition_penalty: 1.1
|
||||
@@ -1,31 +0,0 @@
|
||||
# The following config has been verified on 2x H100-80G GPUs.
|
||||
stage_args:
|
||||
- stage_id: 0
|
||||
runtime:
|
||||
devices: "0,1"
|
||||
max_batch_size: 5
|
||||
engine_args:
|
||||
model_stage: thinker
|
||||
model_arch: Qwen3OmniMoeForConditionalGeneration
|
||||
worker_type: ar
|
||||
scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler
|
||||
gpu_memory_utilization: 0.6
|
||||
enforce_eager: false
|
||||
trust_remote_code: true
|
||||
engine_output_type: latent # Output hidden states for talker
|
||||
distributed_executor_backend: "mp"
|
||||
enable_prefix_caching: false
|
||||
hf_config_name: thinker_config
|
||||
tensor_parallel_size: 2
|
||||
load_format: dummy
|
||||
final_output: true
|
||||
final_output_type: text
|
||||
is_comprehension: true
|
||||
default_sampling_params:
|
||||
temperature: 0.4
|
||||
top_p: 0.9
|
||||
top_k: 1
|
||||
max_tokens: 100
|
||||
seed: 42
|
||||
detokenize: True
|
||||
repetition_penalty: 1.05
|
||||
@@ -15,8 +15,8 @@ os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
|
||||
|
||||
SEED = 42
|
||||
|
||||
stage_config = str(Path(__file__).parent.parent / "e2e" / "stage_configs" / "qwen3_omni_thinker_ci.yaml")
|
||||
model = "Qwen/Qwen3-Omni-30B-A3B-Instruct"
|
||||
stage_config = str(Path(__file__).parent.parent / "e2e" / "stage_configs" / "qwen2_5_omni_thinker_ci.yaml")
|
||||
model = "Qwen/Qwen2.5-Omni-7B"
|
||||
|
||||
|
||||
async def generate(
|
||||
@@ -60,7 +60,7 @@ async def generate(
|
||||
|
||||
@pytest.mark.core_model
|
||||
@pytest.mark.omni
|
||||
@hardware_test(res={"cuda": "H100", "rocm": "MI325"}, num_cards=2)
|
||||
@hardware_test(res={"cuda": "L4", "rocm": "MI325"}, num_cards=1)
|
||||
@pytest.mark.asyncio
|
||||
async def test_abort():
|
||||
with ExitStack() as after:
|
||||
|
||||
@@ -236,7 +236,6 @@ def test_modality_control_003(omni_server) -> None:
|
||||
# TODO: Verify the E2E latency after confirmation baseline.
|
||||
|
||||
|
||||
@pytest.mark.skip(reason="There is a known issue with stream error.")
|
||||
@pytest.mark.advanced_model
|
||||
@pytest.mark.omni
|
||||
@hardware_test(res={"cuda": "L4", "rocm": "MI325"}, num_cards={"cuda": 4, "rocm": 2})
|
||||
|
||||
Reference in new issue
Block a user