A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 23:45 UTC
[Misc] Add dummy maverick test to CI (#21324)
Signed-off-by: Ming Yang <minos.future@gmail.com> Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
Ming Yang committed
Jul 23, 2025 at 20:22 UTC
772ce5af9745393d90408eaa1b8b090014ab551d
2 files changed
+4
.buildkite/test-pipeline.yaml
+1
@@ -718,6 +718,7 @@ steps:
718
- VLLM_USE_V1=0 CUDA_VISIBLE_DEVICES=0,1 pytest -v -s test_sharded_state_loader.py
719
- VLLM_USE_V1=0 CUDA_VISIBLE_DEVICES=0,1 pytest -v -s kv_transfer/test_disagg.py
720
- CUDA_VISIBLE_DEVICES=0,1 pytest -v -s v1/shutdown
721
+ - pytest -v -s models/multimodal/generation/test_maverick.py
722
723
- label: Plugin Tests (2 GPUs) # 40min
724
mirror_hardwares: [amdexperimental]
tests/models/multimodal/generation/test_maverick.py
+3
@@ -23,6 +23,8 @@ from transformers import (AutoConfig, AutoProcessor, AutoTokenizer,
23
24
from vllm import LLM, SamplingParams
25
26
+from ....utils import multi_gpu_test
27
+
28
# Sample prompts for testing
29
PROMPTS: list[str] = [
30
"Hello, my name is",
@@ -541,6 +543,7 @@ def run_reduced_model(model_path: str,
543
print("-" * 40)
544
545
546
+@multi_gpu_test(num_gpus=2)
547
@pytest.mark.parametrize(
548
"original_model_name,text_layers,num_experts,vision_layers,",
549
[("meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8", 4, 4, 2)])