[ROCm][CI] Move remaining mi250_2 tests out of the MI250 queue (#46905)
Signed-off-by: Codex <codex@example.invalid> Co-authored-by: Codex <codex@example.invalid>
Andreas Karatzas committed
Jun 27, 2026 at 04:08 UTC
9fd00ee006ccd4996bbc756397b039343d2fde94
1 file changed
+64
-150
.buildkite/test-amd.yaml
+64
-150
@@ -112,33 +112,6 @@ steps:
112
# #
113
#########################################################################################################################################
114
115
-#----------------------------------------------------- mi250 · basic_correctness -----------------------------------------------------#
116
-
117
-- label: Distributed Model Tests (2 GPUs) # TBD
118
- timeout_in_minutes: 180
119
- mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
120
- agent_pool: mi250_2
121
- num_gpus: 2
122
- working_dir: "/vllm-workspace/tests"
123
- source_file_dependencies:
124
- - vllm/model_executor/model_loader/sharded_state_loader.py
125
- - vllm/model_executor/models/
126
- - vllm/model_executor/layers/
127
- - vllm/v1/attention/backends/
128
- - vllm/v1/attention/selector.py
129
- - tests/basic_correctness/
130
- - tests/model_executor/model_loader/test_sharded_state_loader.py
131
- - tests/models/
132
- - vllm/_aiter_ops.py
133
- - vllm/platforms/rocm.py
134
- commands:
135
- - TARGET_TEST_SUITE=MI250 pytest basic_correctness/ -v -s -m 'distributed(num_gpus=2)'
136
- - CUDA_VISIBLE_DEVICES=0,1 pytest -v -s model_executor/model_loader/test_sharded_state_loader.py -m '(not slow_test)'
137
- - pytest models/test_transformers.py -v -s -m 'distributed(num_gpus=2)'
138
- - pytest models/language -v -s -m 'distributed(num_gpus=2)'
139
- - pytest models/multimodal -v -s -m 'distributed(num_gpus=2)' --ignore models/multimodal/generation/test_whisper.py
140
- - VLLM_WORKER_MULTIPROC_METHOD=spawn pytest models/multimodal/generation/test_whisper.py -v -s -m 'distributed(num_gpus=2)'
141
-
115
#---------------------------------------------------------- mi250 · compile ----------------------------------------------------------#
116
117
- label: PyTorch Compilation Unit Tests # TBD
@@ -179,48 +152,8 @@ steps:
152
commands:
153
- "find compile/fullgraph/ -name 'test_*.py' -not -name 'test_full_graph.py' -exec pytest -s -v {} \\\\;"
154
182
-- label: Distributed Compile + RPC Tests (2 GPUs) # TBD
183
- timeout_in_minutes: 180
184
- mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
185
- agent_pool: mi250_2
186
- num_gpus: 2
187
- optional: true
188
- working_dir: "/vllm-workspace/tests"
189
- source_file_dependencies:
190
- - vllm/compilation/
191
- - vllm/distributed/
192
- - vllm/engine/
193
- - vllm/executor/
194
- - vllm/worker/worker_base.py
195
- - vllm/v1/engine/
196
- - vllm/v1/worker/
197
- - tests/compile/fullgraph/test_basic_correctness.py
198
- - tests/compile/test_wrapper.py
199
- - tests/entrypoints/llm/test_collective_rpc.py
200
- - vllm/platforms/rocm.py
201
- commands:
202
- - pytest -v -s entrypoints/llm/test_collective_rpc.py
203
- - pytest -v -s ./compile/fullgraph/test_basic_correctness.py
204
- - pytest -v -s ./compile/test_wrapper.py
205
-
155
#-------------------------------------------------------- mi250 · distributed --------------------------------------------------------#
156
208
-- label: Distributed Comm Ops # TBD
209
- timeout_in_minutes: 180
210
- mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
211
- agent_pool: mi250_2
212
- num_gpus: 2
213
- working_dir: "/vllm-workspace/tests"
214
- source_file_dependencies:
215
- - vllm/distributed
216
- - tests/distributed
217
- - vllm/platforms/rocm.py
218
- commands:
219
- - pytest -v -s distributed/test_comm_ops.py
220
- - pytest -v -s distributed/test_shm_broadcast.py
221
- - pytest -v -s distributed/test_shm_buffer.py
222
- - pytest -v -s distributed/test_shm_storage.py
223
-
157
- label: Pipeline + Context Parallelism (4 GPUs) # TBD
158
timeout_in_minutes: 180
159
mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
@@ -330,54 +263,6 @@ steps:
263
- pytest -v -s models/multimodal/generation/test_common.py -m core_model -k "not qwen2 and not qwen3 and not gemma"
264
- pytest -v -s models/multimodal/generation/test_qwen2_vl.py -m core_model
265
333
-#---------------------------------------------------------- mi250 · plugins ----------------------------------------------------------#
334
-
335
-- label: Plugin Tests (2 GPUs) # TBD
336
- timeout_in_minutes: 180
337
- mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
338
- agent_pool: mi250_2
339
- num_gpus: 2
340
- optional: true
341
- working_dir: "/vllm-workspace/tests"
342
- source_file_dependencies:
343
- - vllm/plugins/
344
- - tests/plugins/
345
- - vllm/platforms/rocm.py
346
- commands:
347
- # BEGIN: platform plugin and general plugin tests, all the code in-between runs on dummy platform
348
- - pip install -e ./plugins/vllm_add_dummy_platform
349
- - pytest -v -s plugins_tests/test_platform_plugins.py
350
- - pip uninstall vllm_add_dummy_platform -y
351
- # END: platform plugin tests
352
- # BEGIN: `io_processor` plugins test, all the code in between uses the `prithvi_io_processor` plugin
353
- - pip install -e ./plugins/prithvi_io_processor_plugin
354
- - pytest -v -s plugins_tests/test_io_processor_plugins.py
355
- - pytest -v -s plugins_tests/test_terratorch_io_processor_plugins.py
356
- - pip uninstall prithvi_io_processor_plugin -y
357
- # END: `io_processor` plugins test
358
- # BEGIN: `bge_m3_sparse io_processor` test
359
- - pip install -e ./plugins/bge_m3_sparse_plugin
360
- - pytest -v -s plugins_tests/test_bge_m3_sparse_io_processor_plugins.py
361
- - pip uninstall bge_m3_sparse_plugin -y
362
- # END: `bge_m3_sparse io_processor` test
363
- # BEGIN: `colbert_query io_processor` test
364
- - pip install -e ./plugins/colbert_query_plugin
365
- - pytest -v -s plugins_tests/test_colbert_query_io_processor_plugins.py
366
- - pip uninstall colbert_query_plugin -y
367
- # END: `colbert_query io_processor` test
368
- # BEGIN: `stat_logger` plugins test
369
- - pip install -e ./plugins/vllm_add_dummy_stat_logger
370
- - pytest -v -s plugins_tests/test_stats_logger_plugins.py
371
- - pip uninstall dummy_stat_logger -y
372
- # END: `stat_logger` plugins test
373
- # BEGIN: other tests
374
- - pytest -v -s plugins_tests/test_scheduler_plugins.py
375
- - pip install -e ./plugins/vllm_add_dummy_model
376
- - pytest -v -s distributed/test_distributed_oot.py
377
- - pytest -v -s plugins_tests/test_oot_registration_online.py # it needs a clean process
378
- - pytest -v -s plugins_tests/test_oot_registration_offline.py # it needs a clean process
379
- - pytest -v -s plugins_tests/lora_resolvers # unit tests for in-tree lora resolver plugins
380
-
266
#------------------------------------------------------------ mi250 · v1 -------------------------------------------------------------#
267
268
- label: Batch Invariance (H100-MI250) # TBD
@@ -505,41 +390,6 @@ steps:
390
commands:
391
- pytest -v -s v1/attention
392
508
-- label: Distributed DP Tests (2 GPUs) # TBD
509
- timeout_in_minutes: 180
510
- mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
511
- agent_pool: mi250_2
512
- num_gpus: 2
513
- optional: true
514
- working_dir: "/vllm-workspace/tests"
515
- source_file_dependencies:
516
- - vllm/distributed/
517
- - vllm/engine/
518
- - vllm/executor/
519
- - vllm/worker/worker_base.py
520
- - vllm/v1/engine/
521
- - vllm/v1/worker/
522
- - tests/v1/distributed
523
- - tests/entrypoints/openai/test_multi_api_servers.py
524
- - vllm/platforms/rocm.py
525
- commands:
526
- - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_async_llm_dp.py
527
- - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_eagle_dp.py
528
- - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_external_lb_dp.py
529
- - DP_SIZE=2 pytest -v -s entrypoints/openai/test_multi_api_servers.py
530
-
531
-- label: V1 e2e (2 GPUs) # TBD
532
- timeout_in_minutes: 180
533
- mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
534
- agent_pool: mi250_2
535
- optional: true
536
- working_dir: "/vllm-workspace/tests"
537
- source_file_dependencies:
538
- - vllm/
539
- - tests/v1/e2e
540
- commands:
541
- - pytest -v -s v1/e2e/spec_decode/test_spec_decode.py -k "tensor_parallelism"
542
-
393
#------------------------------------------------------------- mi250 · misc ------------------------------------------------------------#
394
395
- label: Async Engine, Inputs, Utils, Worker, Config (CPU) # TBD
@@ -782,6 +632,22 @@ steps:
632
633
#-------------------------------------------------------- mi300 · distributed --------------------------------------------------------#
634
635
+- label: Distributed Comm Ops # TBD
636
+ timeout_in_minutes: 180
637
+ mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
638
+ agent_pool: mi300_2
639
+ num_gpus: 2
640
+ working_dir: "/vllm-workspace/tests"
641
+ source_file_dependencies:
642
+ - vllm/distributed
643
+ - tests/distributed
644
+ - vllm/platforms/rocm.py
645
+ commands:
646
+ - pytest -v -s distributed/test_comm_ops.py
647
+ - pytest -v -s distributed/test_shm_broadcast.py
648
+ - pytest -v -s distributed/test_shm_buffer.py
649
+ - pytest -v -s distributed/test_shm_storage.py
650
+
651
- label: EPLB Algorithm # TBD
652
timeout_in_minutes: 180
653
mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
@@ -1806,6 +1672,54 @@ steps:
1672
- PYTORCH_NVML_BASED_CUDA_CHECK=1 python3 examples/generate/multimodal/vision_language_offline.py --model-type qwen2_5_vl
1673
- VLLM_WORKER_MULTIPROC_METHOD=spawn python3 examples/generate/multimodal/audio_language_offline.py --model-type whisper
1674
1675
+#---------------------------------------------------------- mi300 · plugins ----------------------------------------------------------#
1676
+
1677
+- label: Plugin Tests (2 GPUs) # TBD
1678
+ timeout_in_minutes: 180
1679
+ mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1680
+ agent_pool: mi300_2
1681
+ num_gpus: 2
1682
+ optional: true
1683
+ working_dir: "/vllm-workspace/tests"
1684
+ source_file_dependencies:
1685
+ - vllm/plugins/
1686
+ - tests/plugins/
1687
+ - vllm/platforms/rocm.py
1688
+ commands:
1689
+ # BEGIN: platform plugin and general plugin tests, all the code in-between runs on dummy platform
1690
+ - pip install -e ./plugins/vllm_add_dummy_platform
1691
+ - pytest -v -s plugins_tests/test_platform_plugins.py
1692
+ - pip uninstall vllm_add_dummy_platform -y
1693
+ # END: platform plugin tests
1694
+ # BEGIN: `io_processor` plugins test, all the code in between uses the `prithvi_io_processor` plugin
1695
+ - pip install -e ./plugins/prithvi_io_processor_plugin
1696
+ - pytest -v -s plugins_tests/test_io_processor_plugins.py
1697
+ - pytest -v -s plugins_tests/test_terratorch_io_processor_plugins.py
1698
+ - pip uninstall prithvi_io_processor_plugin -y
1699
+ # END: `io_processor` plugins test
1700
+ # BEGIN: `bge_m3_sparse io_processor` test
1701
+ - pip install -e ./plugins/bge_m3_sparse_plugin
1702
+ - pytest -v -s plugins_tests/test_bge_m3_sparse_io_processor_plugins.py
1703
+ - pip uninstall bge_m3_sparse_plugin -y
1704
+ # END: `bge_m3_sparse io_processor` test
1705
+ # BEGIN: `colbert_query io_processor` test
1706
+ - pip install -e ./plugins/colbert_query_plugin
1707
+ - pytest -v -s plugins_tests/test_colbert_query_io_processor_plugins.py
1708
+ - pip uninstall colbert_query_plugin -y
1709
+ # END: `colbert_query io_processor` test
1710
+ # BEGIN: `stat_logger` plugins test
1711
+ - pip install -e ./plugins/vllm_add_dummy_stat_logger
1712
+ - pytest -v -s plugins_tests/test_stats_logger_plugins.py
1713
+ - pip uninstall dummy_stat_logger -y
1714
+ # END: `stat_logger` plugins test
1715
+ # BEGIN: other tests
1716
+ - pytest -v -s plugins_tests/test_scheduler_plugins.py
1717
+ - pip install -e ./plugins/vllm_add_dummy_model
1718
+ - pytest -v -s distributed/test_distributed_oot.py
1719
+ - pytest -v -s plugins_tests/test_oot_registration_online.py # it needs a clean process
1720
+ - pytest -v -s plugins_tests/test_oot_registration_offline.py # it needs a clean process
1721
+ - pytest -v -s plugins_tests/lora_resolvers # unit tests for in-tree lora resolver plugins
1722
+
1723
#------------------------------------------------------- mi300 · quantization --------------------------------------------------------#
1724
1725
- label: Quantization # TBD