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 6, 08:15 UTC
[ROCm][CI] Remove deepep DBO tests on gfx90a (#37614)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Andreas Karatzas committed
Mar 20, 2026 at 04:07 UTC
37cd9fc107211931d1d69d3d79c93a8c408778c0
1 file changed
+1
-8
.buildkite/test-amd.yaml
+1
-8
@@ -1402,7 +1402,7 @@ steps:
1402
- label: Distributed Tests (2 GPUs)(H100-MI250) # TBD
1403
timeout_in_minutes: 180
1404
mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
1405
- agent_pool: mi250_2
1405
+ agent_pool: mi325_2
1406
num_gpus: 2
1407
working_dir: "/vllm-workspace/"
1408
source_file_dependencies:
@@ -1412,7 +1412,6 @@ steps:
1412
- vllm/v1/attention/backends/
1413
- vllm/v1/attention/selector.py
1414
- tests/distributed/test_context_parallel.py
1415
- - tests/v1/distributed/test_dbo.py
1415
- examples/offline_inference/data_parallel.py
1416
- vllm/_aiter_ops.py
1417
- vllm/platforms/rocm.py
@@ -1420,7 +1419,6 @@ steps:
1419
- export TORCH_NCCL_BLOCKING_WAIT=1
1420
- pytest -v -s tests/distributed/test_context_parallel.py
1421
- VLLM_LOGGING_LEVEL=DEBUG python3 examples/offline_inference/data_parallel.py --model=Qwen/Qwen1.5-MoE-A2.7B -tp=1 -dp=2 --max-model-len=2048 --all2all-backend=allgather_reducescatter --disable-nccl-for-dp-synchronization
1423
- - pytest -v -s tests/v1/distributed/test_dbo.py
1422
1423
1424
#####################################################################################################################################
@@ -2594,21 +2592,16 @@ steps:
2592
mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi325]
2593
agent_pool: mi325_2
2594
num_gpus: 2
2597
- optional: true
2595
working_dir: "/vllm-workspace/"
2596
source_file_dependencies:
2597
- vllm/distributed/
2598
- vllm/v1/distributed/
2599
- vllm/model_executor/layers/fused_moe/
2603
- - tests/distributed/test_context_parallel.py
2600
- tests/v1/distributed/test_dbo.py
2605
- - examples/offline_inference/data_parallel.py
2601
- vllm/_aiter_ops.py
2602
- vllm/platforms/rocm.py
2603
commands:
2604
- export TORCH_NCCL_BLOCKING_WAIT=1
2610
- - pytest -v -s tests/distributed/test_context_parallel.py
2611
- - VLLM_LOGGING_LEVEL=DEBUG python3 examples/offline_inference/data_parallel.py --model=Qwen/Qwen1.5-MoE-A2.7B -tp=1 -dp=2 --max-model-len=2048 --all2all-backend=deepep_high_throughput
2605
- pytest -v -s tests/v1/distributed/test_dbo.py
2606
2607