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 12, 04:45 UTC
Update AMD-CI testing mirror (as of 2025-12-02) (#29898)
Signed-off-by: Alexei V. Ivanov <alexei.ivanov@amd.com>
Alexei-V-Ivanov-AMD committed
Dec 2, 2025 at 11:52 UTC
c77b9929a04c56d369c9f6b86fbf5d4891bab285
1 file changed
+28
-15
.buildkite/test-amd.yaml
+28
-15
@@ -51,7 +51,7 @@ steps:
51
- label: Async Engine, Inputs, Utils, Worker Test # 10min
52
timeout_in_minutes: 15
53
mirror_hardwares: [amdexperimental, amdproduction, amdtentative]
54
- agent_pool: mi325_1
54
+ agent_pool: mi355_1
55
grade: Blocking
56
source_file_dependencies:
57
- vllm/
@@ -64,7 +64,7 @@ steps:
64
- label: Async Engine, Inputs, Utils, Worker, Config Test (CPU) # 15min
65
timeout_in_minutes: 20
66
mirror_hardwares: [amdexperimental, amdproduction, amdtentative]
67
- agent_pool: mi325_1
67
+ agent_pool: mi355_1
68
grade: Blocking
69
source_file_dependencies:
70
- vllm/
@@ -99,7 +99,7 @@ steps:
99
- label: Basic Correctness Test # 20min
100
timeout_in_minutes: 30
101
mirror_hardwares: [amdexperimental, amdproduction]
102
- agent_pool: mi325_1
102
+ agent_pool: mi355_1
103
# grade: Blocking
104
fast_check: true
105
torch_nightly: true
@@ -116,7 +116,7 @@ steps:
116
117
- label: Entrypoints Unit Tests # 5min
118
mirror_hardwares: [amdexperimental, amdproduction, amdtentative]
119
- agent_pool: mi325_1
119
+ agent_pool: mi355_1
120
grade: Blocking
121
timeout_in_minutes: 10
122
working_dir: "/vllm-workspace/tests"
@@ -131,7 +131,7 @@ steps:
131
- label: Entrypoints Integration Test (LLM) # 30min
132
timeout_in_minutes: 40
133
mirror_hardwares: [amdexperimental, amdproduction]
134
- agent_pool: mi325_1
134
+ agent_pool: mi355_1
135
# grade: Blocking
136
working_dir: "/vllm-workspace/tests"
137
fast_check: true
@@ -254,7 +254,7 @@ steps:
254
255
- label: EPLB Algorithm Test # 5min
256
mirror_hardwares: [amdexperimental, amdproduction, amdtentative]
257
- agent_pool: mi325_1
257
+ agent_pool: mi355_1
258
grade: Blocking
259
timeout_in_minutes: 15
260
working_dir: "/vllm-workspace/tests"
@@ -266,7 +266,7 @@ steps:
266
267
- label: EPLB Execution Test # 10min
268
mirror_hardwares: [amdexperimental, amdproduction]
269
- agent_pool: mi325_4
269
+ agent_pool: mi355_4
270
# grade: Blocking
271
timeout_in_minutes: 20
272
working_dir: "/vllm-workspace/tests"
@@ -281,7 +281,7 @@ steps:
281
- label: Metrics, Tracing Test # 12min
282
timeout_in_minutes: 20
283
mirror_hardwares: [amdexperimental, amdproduction]
284
- agent_pool: mi325_2
284
+ agent_pool: mi355_2
285
# grade: Blocking
286
num_gpus: 2
287
source_file_dependencies:
@@ -301,7 +301,7 @@ steps:
301
- label: Regression Test # 7min
302
timeout_in_minutes: 20
303
mirror_hardwares: [amdexperimental, amdproduction, amdtentative]
304
- agent_pool: mi325_1
304
+ agent_pool: mi355_1
305
grade: Blocking
306
source_file_dependencies:
307
- vllm/
@@ -343,7 +343,7 @@ steps:
343
- label: V1 Test entrypoints # 35min
344
timeout_in_minutes: 50
345
mirror_hardwares: [amdexperimental, amdproduction, amdtentative]
346
- agent_pool: mi325_1
346
+ agent_pool: mi355_1
347
grade: Blocking
348
source_file_dependencies:
349
- vllm/
@@ -544,7 +544,7 @@ steps:
544
- label: PyTorch Fullgraph Test # 27min
545
timeout_in_minutes: 40
546
mirror_hardwares: [amdexperimental, amdproduction]
547
- agent_pool: mi325_1
547
+ agent_pool: mi355_1
548
# grade: Blocking
549
torch_nightly: true
550
source_file_dependencies:
@@ -715,6 +715,7 @@ steps:
715
# we can only upgrade after this is resolved
716
# TODO(jerryzh168): resolve the above comment
717
- uv pip install --system torchao==0.13.0
718
+ - uv pip install --system conch-triton-kernels
719
- VLLM_TEST_FORCE_LOAD_FORMAT=auto pytest -v -s quantization/ --ignore quantization/test_blackwell_moe.py
720
721
- label: LM Eval Small Models # 15min
@@ -934,6 +935,18 @@ steps:
935
commands:
936
- pytest -v -s models/language/pooling_mteb_test
937
938
+- label: Multi-Modal Processor Test (CPU)
939
+ timeout_in_minutes: 60
940
+ mirror_hardwares: [amdexperimental]
941
+ agent_pool: mi325_1
942
+ source_file_dependencies:
943
+ - vllm/
944
+ - tests/models/multimodal
945
+ no_gpu: true
946
+ commands:
947
+ - pip install git+https://github.com/TIGER-AI-Lab/Mantis.git
948
+ - pytest -v -s models/multimodal/processing --ignore models/multimodal/processing/test_tensor_schema.py
949
+
950
- label: Multi-Modal Processor Test # 44min
951
timeout_in_minutes: 60
952
mirror_hardwares: [amdexperimental]
@@ -1472,14 +1485,14 @@ steps:
1485
working_dir: "/vllm-workspace/"
1486
num_gpus: 2
1487
commands:
1475
- - pytest -v -s tests/compile/distributed/test_async_tp.py
1488
+ - VLLM_TEST_CLEAN_GPU_MEMORY=1 pytest -v -s tests/compile/distributed/test_async_tp.py
1489
- pytest -v -s tests/compile/distributed/test_sequence_parallelism.py
1490
- pytest -v -s tests/compile/distributed/test_fusion_all_reduce.py
1491
#- pytest -v -s tests/compile/distributed/test_fusions_e2e.py::test_tp2_attn_quant_allreduce_rmsnorm
1479
- - "pytest -v -s tests/compile/distributed/test_fusions_e2e.py -k 'not Llama-4'"
1480
- - pytest -v -s tests/distributed/test_sequence_parallel.py
1492
+ - "VLLM_TEST_CLEAN_GPU_MEMORY=1 pytest -v -s tests/compile/distributed/test_fusions_e2e.py -k 'not Llama-4'"
1493
+ - VLLM_TEST_CLEAN_GPU_MEMORY=1 pytest -v -s tests/distributed/test_sequence_parallel.py
1494
- pytest -v -s tests/distributed/test_context_parallel.py
1482
- - CUDA_VISIBLE_DEVICES=1,2 VLLM_ALL2ALL_BACKEND=deepep_high_throughput VLLM_USE_DEEP_GEMM=1 VLLM_LOGGING_LEVEL=DEBUG python3 examples/offline_inference/data_parallel.py --model Qwen/Qwen1.5-MoE-A2.7B --tp-size=1 --dp-size=2 --max-model-len 2048
1495
+ - HIP_VISIBLE_DEVICES=0,1 VLLM_ALL2ALL_BACKEND=deepep_high_throughput VLLM_USE_DEEP_GEMM=1 VLLM_LOGGING_LEVEL=DEBUG python3 examples/offline_inference/data_parallel.py --model Qwen/Qwen1.5-MoE-A2.7B --tp-size=1 --dp-size=2 --max-model-len 2048
1496
- pytest -v -s tests/v1/distributed/test_dbo.py
1497
1498
##### B200 test #####