@tduncan / Nvidia-Optimization-Stack-VLLM / commits / 4f66bc3e0d

[CI][ROCm] Standardize AMD test job labels by device (#52976)

Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com> Signed-off-by: Andreas Karatzas <akaratza@amd.com>

Andreas Karatzas committed Aug 19, 2026 at 22:33 UTC 4f66bc3e0d92aca0e8d1a29e22425327a07c916e
4 files changed +384 -340
.buildkite/test-amd.yaml
+376 -332
@@ -13,7 +13,7 @@
13 # soft_fail(bool): allow this step to fail without failing the entire pipeline (useful for flaky or experimental tests).
14 # command(str): the single command to run for tests. incompatible with commands.
15 # commands(list): the list of commands to run for the test. incompatible with command.
16 -# mirror_hardwares(list): the list of hardware to run the test on as well. currently only supports [amdexperimental]
16 +# mirror_hardwares(list): selector tags for AMD pipelines that should include the test.
17 # dind(bool): when false, run the job directly in the AMD Kubernetes pod.
18 # When true or omitted, use the legacy Docker-in-Docker path.
19 # num_gpus(int): override the number of GPUs for the test. defaults to 1 GPU. currently supports 2,4.
@@ -105,6 +105,156 @@
105
106 steps:
107
108 +#########################################################################################################################################
109 +# #
110 +# AMD CPU tests #
111 +# #
112 +#########################################################################################################################################
113 +
114 +- label: ":computer: (CPU) Basic Models Other" # TBD
115 + timeout_in_minutes: 180
116 + mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
117 + agent_pool: mi250_1
118 + no_gpu: true
119 + optional: true
120 + working_dir: "/vllm-workspace/tests"
121 + source_file_dependencies:
122 + - vllm/
123 + - tests/models/test_utils.py
124 + - tests/models/test_vision.py
125 + - tests/models/test_adapters.py
126 + - tests/models/transformers/fusers/
127 + commands:
128 + - pytest -v -s models/test_utils.py models/test_vision.py models/test_adapters.py models/transformers/fusers/
129 +
130 +- label: ":computer: (CPU) Multimodal Processor Shard %N" # TBD
131 + timeout_in_minutes: 180
132 + mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
133 + agent_pool: mi250_1
134 + no_gpu: true
135 + parallelism: 6
136 + optional: true
137 + working_dir: "/vllm-workspace/tests"
138 + source_file_dependencies:
139 + - vllm/
140 + - tests/models/multimodal
141 + - tests/models/registry.py
142 + commands:
143 + - pytest -v -s models/multimodal/processing --ignore models/multimodal/processing/test_tensor_schema.py --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
144 +
145 +- label: ":computer: (CPU) V1 Others" # TBD
146 + timeout_in_minutes: 180
147 + mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
148 + agent_pool: mi250_1
149 + no_gpu: true
150 + optional: true
151 + working_dir: "/vllm-workspace/tests"
152 + source_file_dependencies:
153 + - vllm/
154 + - tests/v1
155 + commands:
156 + - pytest -v -s -m 'cpu_test' v1/core
157 + - pytest -v -s v1/structured_output
158 + - pytest -v -s v1/test_serial_utils.py
159 + - pytest -v -s v1/test_kv_cache_spec_registry.py
160 + - pytest -v -s v1/cudagraph/test_cudagraph_manager.py
161 + - pytest -v -s -m 'cpu_test' v1/kv_connector/unit
162 + - pytest -v -s -m 'cpu_test' v1/ec_connector/unit
163 + - pytest -v -s -m 'cpu_test' v1/metrics
164 +
165 +- label: ":computer: (CPU) Async Engine, Inputs, Utils, Worker, Config" # TBD
166 + timeout_in_minutes: 180
167 + mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
168 + agent_pool: mi250_1
169 + no_gpu: true
170 + optional: true
171 + working_dir: "/vllm-workspace/tests"
172 + source_file_dependencies:
173 + - vllm/
174 + - tests/test_envs.py
175 + - tests/test_inputs.py
176 + - tests/test_outputs.py
177 + - tests/test_pooling_params.py
178 + - tests/test_ray_env.py
179 + - tests/test_sampling_params.py
180 + - tests/multimodal
181 + - tests/renderers
182 + - tests/standalone_tests/lazy_imports.py
183 + - tests/tokenizers_
184 + - tests/reasoning
185 + - tests/tool_parsers
186 + - tests/parser
187 + - tests/transformers_utils
188 + - tests/config
189 + commands:
190 + - python3 standalone_tests/lazy_imports.py
191 + - pytest -v -s test_envs.py
192 + - pytest -v -s test_inputs.py
193 + - pytest -v -s test_outputs.py
194 + - pytest -v -s test_pooling_params.py
195 + - pytest -v -s test_ray_env.py
196 + - pytest -v -s test_sampling_params.py
197 + - pytest -v -s -m 'cpu_test' multimodal
198 + - pytest -v -s renderers
199 + - pytest -v -s tokenizers_
200 + - pytest -v -s reasoning
201 + - pytest -v -s tool_parsers
202 + - pytest -v -s parser
203 + - pytest -v -s transformers_utils
204 + - pytest -v -s config --ignore=config/test_mp_reducer.py
205 +
206 +- label: ":computer: (CPU) Rust Frontend Cargo Style + Clippy" # TBD
207 + timeout_in_minutes: 180
208 + mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
209 + agent_pool: mi250_1
210 + no_gpu: true
211 + optional: true
212 + working_dir: "/vllm-workspace"
213 + source_file_dependencies:
214 + - rust/
215 + - rust-toolchain.toml
216 + - .buildkite/test_areas/rust_frontend_cargo.yaml
217 + - .buildkite/scripts/run-rust-frontend-cargo-ci.sh
218 + commands:
219 + - bash .buildkite/scripts/run-rust-frontend-cargo-ci.sh style-clippy
220 +
221 +- label: ":computer: (CPU) Rust Frontend Cargo" # TBD
222 + timeout_in_minutes: 180
223 + mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
224 + agent_pool: mi250_1
225 + no_gpu: true
226 + optional: true
227 + working_dir: "/vllm-workspace"
228 + source_file_dependencies:
229 + - rust/
230 + - rust-toolchain.toml
231 + - .buildkite/test_areas/rust_frontend_cargo.yaml
232 + - .buildkite/scripts/run-rust-frontend-cargo-ci.sh
233 + commands:
234 + - bash .buildkite/scripts/run-rust-frontend-cargo-ci.sh test
235 +
236 +- label: ":computer: (CPU) Docker Build Metadata" # TBD
237 + timeout_in_minutes: 180
238 + mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
239 + agent_pool: mi250_1
240 + no_gpu: true
241 + optional: true
242 + working_dir: "/vllm-workspace/tests"
243 + source_file_dependencies:
244 + - .buildkite/scripts/docker-build-metadata-args.sh
245 + - .buildkite/scripts/ci-bake-rocm.sh
246 + - .buildkite/release-pipeline.yaml
247 + - docker/Dockerfile
248 + - docker/Dockerfile.cpu
249 + - docker/Dockerfile.rocm
250 + - docker/Dockerfile.rocm_base
251 + - docker/ci-rocm.hcl
252 + - docker/docker-bake.hcl
253 + - docker/docker-bake-rocm.hcl
254 + - tests/tools/test_docker_build_metadata_args.py
255 + commands:
256 + - pytest -v -s tools/test_docker_build_metadata_args.py
257 +
258 #########################################################################################################################################
259 # #
260 # MI250 (gfx90a) tests #
@@ -113,7 +263,7 @@ steps:
263
264 #---------------------------------------------------------- mi250 · compile ----------------------------------------------------------#
265
116 -- label: PyTorch Fullgraph Smoke Test # TBD
266 +- label: ":amd: (MI250) PyTorch Fullgraph Smoke" # TBD
267 timeout_in_minutes: 180
268 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
269 agent_pool: mi250_1
@@ -132,7 +282,7 @@ steps:
282
283 #-------------------------------------------------------- mi250 · distributed --------------------------------------------------------#
284
135 -- label: Pipeline + Context Parallelism (4 GPUs) # TBD
285 +- label: ":amd: (MI250) Pipeline + Context Parallelism" # TBD
286 timeout_in_minutes: 180
287 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
288 agent_pool: mi250_4
@@ -155,7 +305,7 @@ steps:
305
306 #---------------------------------------------------------- mi250 · kernels ----------------------------------------------------------#
307
158 -- label: Kernels Helion Test # TBD
308 +- label: ":amd: (MI250) Helion Kernels" # TBD
309 timeout_in_minutes: 180
310 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
311 agent_pool: mi250_1
@@ -170,25 +320,9 @@ steps:
320 - pip install helion==1.4.0
321 - pytest -v -s kernels/helion/
322
173 -#------------------------------------------------------ mi250 · models / basic -------------------------------------------------------#
174 -
175 -- label: Basic Models Test (Other CPU) # TBD
176 - timeout_in_minutes: 180
177 - mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
178 - agent_pool: mi250_1
179 - no_gpu: true
180 - optional: true
181 - working_dir: "/vllm-workspace/tests"
182 - source_file_dependencies:
183 - - vllm/
184 - - tests/models/test_utils.py
185 - - tests/models/test_vision.py
186 - commands:
187 - - pytest -v -s models/test_utils.py models/test_vision.py
188 -
323 #----------------------------------------------------- mi250 · models / language -----------------------------------------------------#
324
191 -- label: Language Models Test (MTEB) # TBD
325 +- label: ":amd: (MI250) Language Models (MTEB)" # TBD
326 timeout_in_minutes: 180
327 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
328 agent_pool: mi250_1
@@ -200,7 +334,7 @@ steps:
334 commands:
335 - pytest -v -s models/language/pooling_mteb_test
336
203 -- label: Language Models Test (PPL) # TBD
337 +- label: ":amd: (MI250) Language Models (PPL)" # TBD
338 timeout_in_minutes: 180
339 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
340 agent_pool: mi250_1
@@ -213,7 +347,7 @@ steps:
347
348 #---------------------------------------------------- mi250 · models / multimodal ----------------------------------------------------#
349
216 -- label: "Multi-Modal Models (Standard) 3: llava + qwen2_vl" # TBD
350 +- label: ":amd: (MI250) Multimodal Models (Standard) 3: llava + qwen2_vl" # TBD
351 timeout_in_minutes: 180
352 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
353 agent_pool: mi250_1
@@ -226,23 +360,9 @@ steps:
360 - pytest -v -s models/multimodal/generation/test_common.py -m core_model -k "not qwen2 and not qwen3 and not gemma"
361 - pytest -v -s models/multimodal/generation/test_qwen2_vl.py -m core_model
362
229 -- label: Multi-Modal Processor (CPU) %N # TBD
230 - timeout_in_minutes: 180
231 - mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
232 - agent_pool: mi250_1
233 - parallelism: 6
234 - optional: true
235 - working_dir: "/vllm-workspace/tests"
236 - source_file_dependencies:
237 - - vllm/
238 - - tests/models/multimodal
239 - - tests/models/registry.py
240 - commands:
241 - - pytest -v -s models/multimodal/processing --ignore models/multimodal/processing/test_tensor_schema.py --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
242 -
363 #------------------------------------------------------------ mi250 · v1 -------------------------------------------------------------#
364
245 -- label: Batch Invariance (H100-MI250) # TBD
365 +- label: ":amd: (MI250) Batch Invariance" # TBD
366 timeout_in_minutes: 180
367 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
368 agent_pool: mi250_1
@@ -260,7 +380,7 @@ steps:
380 - pytest -v -s v1/determinism/test_batch_invariance.py
381 - pytest -v -s v1/determinism/test_rms_norm_batch_invariant.py
382
263 -- label: Cudagraph # TBD
383 +- label: ":amd: (MI250) CUDAGraph" # TBD
384 timeout_in_minutes: 180
385 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
386 agent_pool: mi250_1
@@ -276,7 +396,7 @@ steps:
396 - pytest -v -s v1/cudagraph/test_cudagraph_dispatch.py
397 - pytest -v -s v1/cudagraph/test_cudagraph_mode.py
398
279 -- label: e2e Core (1 GPU) # TBD
399 +- label: ":amd: (MI250) E2E Core" # TBD
400 timeout_in_minutes: 180
401 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
402 agent_pool: mi250_1
@@ -289,7 +409,7 @@ steps:
409 commands:
410 - pytest -v -s v1/e2e/general --ignore v1/e2e/general/test_async_scheduling.py
411
292 -- label: e2e Scheduling (1 GPU) # TBD
412 +- label: ":amd: (MI250) E2E Scheduling" # TBD
413 timeout_in_minutes: 180
414 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
415 agent_pool: mi250_1
@@ -302,7 +422,7 @@ steps:
422 commands:
423 - pytest -v -s v1/e2e/general/test_async_scheduling.py
424
305 -- label: Engine (1 GPU) # TBD
425 +- label: ":amd: (MI250) V1 Engine" # TBD
426 timeout_in_minutes: 180
427 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
428 agent_pool: mi250_1
@@ -311,12 +431,19 @@ steps:
431 source_file_dependencies:
432 - vllm/v1/
433 - tests/v1/engine/
434 + - tests/v1/test_tensor_ipc_queue.py
435 + - tests/config/test_mp_reducer.py
436 + - vllm/config/
437 + - vllm/engine/arg_utils.py
438 + - vllm/transformers_utils/config.py
439 - vllm/platforms/rocm.py
440 commands:
441 - pytest -v -s v1/engine/test_preprocess_error_handling.py
442 - pytest -v -s v1/engine --ignore v1/engine/test_preprocess_error_handling.py
443 + - pytest -v -s v1/test_tensor_ipc_queue.py
444 + - pytest -v -s config/test_mp_reducer.py
445
319 -- label: Spec Decode Draft Model # TBD
446 +- label: ":amd: (MI250) Spec Decode Draft Model" # TBD
447 timeout_in_minutes: 180
448 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
449 agent_pool: mi250_1
@@ -333,7 +460,7 @@ steps:
460 commands:
461 - pytest -v -s v1/e2e/spec_decode/draft_model/
462
336 -- label: Spec Decode Speculators + MTP # TBD
463 +- label: ":amd: (MI250) Spec Decode Speculators + MTP" # TBD
464 timeout_in_minutes: 180
465 mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
466 agent_pool: mi250_1
@@ -352,64 +479,9 @@ steps:
479 - pytest -v -s v1/e2e/spec_decode/speculators/
480 - pytest -v -s v1/e2e/spec_decode/mtp/
481
355 -- label: V1 others (CPU) # TBD
356 - timeout_in_minutes: 180
357 - mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
358 - agent_pool: mi250_1
359 - optional: true
360 - working_dir: "/vllm-workspace/tests"
361 - source_file_dependencies:
362 - - vllm/
363 - - tests/v1
364 - commands:
365 - - pytest -v -s -m 'cpu_test' v1/core
366 - - pytest -v -s v1/structured_output
367 - - pytest -v -s v1/test_serial_utils.py
368 - - pytest -v -s -m 'cpu_test' v1/kv_connector/unit
369 - - pytest -v -s -m 'cpu_test' v1/metrics
370 -
482 #------------------------------------------------------------- mi250 · misc ------------------------------------------------------------#
483
373 -- label: Async Engine, Inputs, Utils, Worker, Config (CPU) # TBD
374 - timeout_in_minutes: 180
375 - mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
376 - agent_pool: mi250_1
377 - optional: true
378 - no_gpu: true
379 - working_dir: "/vllm-workspace/tests"
380 - source_file_dependencies:
381 - - vllm/
382 - - tests/test_inputs.py
383 - - tests/test_outputs.py
384 - - tests/test_pooling_params.py
385 - - tests/test_ray_env.py
386 - - tests/test_sampling_params.py
387 - - tests/multimodal
388 - - tests/renderers
389 - - tests/standalone_tests/lazy_imports.py
390 - - tests/tokenizers_
391 - - tests/reasoning
392 - - tests/tool_parsers
393 - - tests/parser
394 - - tests/transformers_utils
395 - - tests/config
396 - commands:
397 - - python3 standalone_tests/lazy_imports.py
398 - - pytest -v -s test_inputs.py
399 - - pytest -v -s test_outputs.py
400 - - pytest -v -s test_pooling_params.py
401 - - pytest -v -s test_ray_env.py
402 - - pytest -v -s test_sampling_params.py
403 - - pytest -v -s -m 'cpu_test' multimodal
404 - - pytest -v -s renderers
405 - - pytest -v -s tokenizers_
406 - - pytest -v -s reasoning
407 - - pytest -v -s tool_parsers
408 - - pytest -v -s parser
409 - - pytest -v -s transformers_utils
410 - - pytest -v -s config
411 -
412 -- label: Python-only Installation # TBD
484 +- label: ":amd: (MI300) Python-only Installation" # TBD
485 timeout_in_minutes: 180
486 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
487 dind: false
@@ -423,59 +495,6 @@ steps:
495 commands:
496 - bash standalone_tests/python_only_compile.sh
497
426 -#------------------------------------------------------------ mi250 · rust -----------------------------------------------------------#
427 -
428 -- label: Rust Frontend Cargo Style + Clippy # TBD
429 - timeout_in_minutes: 180
430 - mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
431 - agent_pool: mi250_1
432 - no_gpu: true
433 - working_dir: "/vllm-workspace"
434 - source_file_dependencies:
435 - - rust/
436 - - rust-toolchain.toml
437 - - .buildkite/test_areas/rust_frontend_cargo.yaml
438 - - .buildkite/scripts/run-rust-frontend-cargo-ci.sh
439 - commands:
440 - - bash .buildkite/scripts/run-rust-frontend-cargo-ci.sh style-clippy
441 -
442 -- label: Rust Frontend Cargo Tests # TBD
443 - timeout_in_minutes: 180
444 - mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
445 - agent_pool: mi250_1
446 - no_gpu: true
447 - working_dir: "/vllm-workspace"
448 - source_file_dependencies:
449 - - rust/
450 - - rust-toolchain.toml
451 - - .buildkite/test_areas/rust_frontend_cargo.yaml
452 - - .buildkite/scripts/run-rust-frontend-cargo-ci.sh
453 - commands:
454 - - bash .buildkite/scripts/run-rust-frontend-cargo-ci.sh test
455 -
456 -#----------------------------------------------------------- mi250 · docker ----------------------------------------------------------#
457 -
458 -- label: Docker Build Metadata (ROCm) # TBD
459 - timeout_in_minutes: 180
460 - mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
461 - agent_pool: mi250_1
462 - no_gpu: true
463 - optional: true
464 - working_dir: "/vllm-workspace/tests"
465 - source_file_dependencies:
466 - - .buildkite/scripts/docker-build-metadata-args.sh
467 - - .buildkite/scripts/ci-bake-rocm.sh
468 - - docker/Dockerfile
469 - - docker/Dockerfile.cpu
470 - - docker/Dockerfile.rocm
471 - - docker/Dockerfile.rocm_base
472 - - docker/ci-rocm.hcl
473 - - docker/docker-bake.hcl
474 - - docker/docker-bake-rocm.hcl
475 - - tests/tools/test_docker_build_metadata_args.py
476 - commands:
477 - - pytest -v -s tools/test_docker_build_metadata_args.py
478 -
498 #########################################################################################################################################
499 # #
500 # MI300 (gfx942) tests #
@@ -484,7 +503,7 @@ steps:
503
504 #----------------------------------------------------- mi300 · basic_correctness -----------------------------------------------------#
505
487 -- label: Basic Correctness # TBD
506 +- label: ":amd: (MI300) Basic Correctness" # TBD
507 timeout_in_minutes: 180
508 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
509 dind: false
@@ -502,7 +521,7 @@ steps:
521 - VLLM_TARGET_TEST_SUITE=MI300 pytest -v -s basic_correctness/test_basic_correctness.py
522 - pytest -v -s basic_correctness/test_cpu_offload.py
523
505 -- label: Distributed Model Tests (2 GPUs) # TBD
524 +- label: ":amd: (MI300) Distributed Models" # TBD
525 timeout_in_minutes: 180
526 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
527 dind: false
@@ -529,10 +548,11 @@ steps:
548 - pytest models/multimodal -v -s -m 'distributed(num_gpus=2)' --ignore models/multimodal/generation/test_whisper.py --ignore models/multimodal/generation/test_phi4siglip.py
549 - pytest models/multimodal/generation/test_phi4siglip.py -v -s -m 'distributed(num_gpus=2)'
550 - VLLM_WORKER_MULTIPROC_METHOD=spawn pytest models/multimodal/generation/test_whisper.py -v -s -m 'distributed(num_gpus=2)'
551 + - pytest -v -s models/test_vision.py -m 'distributed(num_gpus=2)'
552
553 #-------------------------------------------------------- mi300 · benchmarks ---------------------------------------------------------#
554
535 -- label: Benchmarks CLI Test # TBD
555 +- label: ":amd: (MI300) Benchmarks CLI" # TBD
556 timeout_in_minutes: 180
557 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
558 dind: false
@@ -547,7 +567,7 @@ steps:
567
568 #---------------------------------------------------------- mi300 · compile ----------------------------------------------------------#
569
550 -- label: PyTorch Compilation Unit Tests # TBD
570 +- label: ":amd: (MI300) PyTorch Compilation" # TBD
571 timeout_in_minutes: 180
572 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
573 dind: false
@@ -567,7 +587,7 @@ steps:
587 commands:
588 - "find compile/ -maxdepth 1 -name 'test_*.py' -print0 | xargs -0 -n1 -I{} pytest -s -v '{}'"
589
570 -- label: Fusion E2E Config Sweep (H100-MI300) # TBD
590 +- label: ":amd: (MI300) Fusion E2E Config Sweep" # TBD
591 timeout_in_minutes: 180
592 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
593 dind: false
@@ -588,7 +608,7 @@ steps:
608 - rocm-smi
609 - pytest -v -s tests/compile/fusions_e2e/test_tp1_quant.py -k "llama-3"
610
591 -- label: Fusion E2E Quick (H100-MI300) # TBD
611 +- label: ":amd: (MI300) Fusion E2E Quick" # TBD
612 timeout_in_minutes: 180
613 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
614 dind: false
@@ -610,7 +630,7 @@ steps:
630 # Different from CUDA, Qwen requires +rms_norm and +quant_fp8 as rms+quant fusion is only supported on AITER
631 - "pytest -v -s tests/compile/fusions_e2e/test_tp1_quant.py -k 'inductor_partition and +rms_norm and +quant_fp8 and qwen3'"
632
613 -- label: PyTorch Compilation Passes Unit Tests # TBD
633 +- label: ":amd: (MI300) PyTorch Compilation Passes" # TBD
634 timeout_in_minutes: 180
635 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
636 dind: false
@@ -622,7 +642,7 @@ steps:
642 commands:
643 - pytest -s -v compile/passes --ignore compile/passes/distributed
644
625 -- label: PyTorch Fullgraph # TBD
645 +- label: ":amd: (MI300) PyTorch Fullgraph" # TBD
646 timeout_in_minutes: 180
647 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
648 dind: false
@@ -640,7 +660,7 @@ steps:
660 commands:
661 - pytest -v -s compile/fullgraph/test_full_graph.py
662
643 -- label: Pytorch Nightly Dependency Override Check # TBD
663 +- label: ":amd: (MI300) PyTorch Nightly Dependency Override Check" # TBD
664 timeout_in_minutes: 180
665 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
666 dind: false
@@ -654,7 +674,7 @@ steps:
674 commands:
675 - bash standalone_tests/pytorch_nightly_dependency.sh
676
657 -- label: Distributed Compile Unit Tests (2xH100-2xMI300) # TBD
677 +- label: ":amd: (MI300) Distributed Compile" # TBD
678 timeout_in_minutes: 180
679 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
680 dind: false
@@ -673,7 +693,7 @@ steps:
693 - VLLM_TEST_CLEAN_GPU_MEMORY=1 pytest -v -s tests/compile/passes/distributed/test_async_tp.py
694 - pytest -v -s tests/compile/fusions_e2e/test_tp2_ar_rms.py::test_tp2_ar_rms_fusions
695
676 -- label: Distributed Compile + RPC Tests (2 GPUs) # TBD
696 +- label: ":amd: (MI300) Distributed Compile + RPC" # TBD
697 timeout_in_minutes: 180
698 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
699 dind: false
@@ -699,7 +719,7 @@ steps:
719
720 #----------------------------------------------------------- mi300 · cuda ------------------------------------------------------------#
721
702 -- label: Platform Tests # TBD
722 +- label: ":amd: (MI300) CUDA Platform" # TBD
723 timeout_in_minutes: 180
724 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
725 dind: false
@@ -714,7 +734,7 @@ steps:
734
735 #-------------------------------------------------------- mi300 · detokenizer --------------------------------------------------------#
736
717 -- label: Async Engine, Inputs, Utils, Worker # TBD
737 +- label: ":amd: (MI300) Async Engine, Inputs, Utils, Worker" # TBD
738 timeout_in_minutes: 180
739 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
740 dind: false
@@ -729,11 +749,12 @@ steps:
749 commands:
750 - pytest -v -s detokenizer
751 - pytest -v -s -m 'not cpu_test' multimodal
752 + - pytest -v -s multimodal/test_cache.py::test_sleep_wake_preserves_mm_cache_consistency
753 - pytest -v -s utils_
754
755 #-------------------------------------------------------- mi300 · distributed --------------------------------------------------------#
756
736 -- label: Distributed Comm Ops # TBD
757 +- label: ":amd: (MI300) Distributed Comm Ops" # TBD
758 timeout_in_minutes: 180
759 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
760 dind: false
@@ -750,7 +771,7 @@ steps:
771 - pytest -v -s distributed/test_shm_buffer.py
772 - pytest -v -s distributed/test_shm_storage.py
773
753 -- label: EPLB Algorithm # TBD
774 +- label: ":amd: (MI300) EPLB Algorithm" # TBD
775 timeout_in_minutes: 180
776 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
777 dind: false
@@ -766,7 +787,7 @@ steps:
787 - pytest -v -s distributed/test_eplb_algo.py
788 - pytest -v -s distributed/test_eplb_utils.py
789
769 -- label: EPLB Execution # TBD
790 +- label: ":amd: (MI300) EPLB Execution" # TBD
791 timeout_in_minutes: 180
792 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
793 dind: false
@@ -782,7 +803,7 @@ steps:
803 - pytest -v -s distributed/test_eplb_execute.py
804 - pytest -v -s distributed/test_eplb_spec_decode.py
805
785 -- label: Distributed Tests (2xH100-2xMI300) # TBD
806 +- label: ":amd: (MI300) Distributed Features" # TBD
807 timeout_in_minutes: 180
808 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
809 dind: false
@@ -809,7 +830,7 @@ steps:
830 - VLLM_ALLOW_INSECURE_SERIALIZATION=1 pytest -v -s tests/distributed/test_weight_transfer.py
831 - pytest -v -s tests/distributed/test_packed_tensor.py
832
812 -- label: Distributed Tests (4xA100-4xMI300) # TBD
833 +- label: ":amd: (MI300) Distributed" # TBD
834 timeout_in_minutes: 180
835 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
836 dind: false
@@ -819,13 +840,15 @@ steps:
840 working_dir: "/vllm-workspace/tests"
841 source_file_dependencies:
842 - vllm/
843 + - tests/models/test_vision.py
844 commands:
845 - pytest -v -s distributed/test_custom_all_reduce.py
846 - torchrun --nproc_per_node=2 distributed/test_ca_buffer_sharing.py
847 - TARGET_TEST_SUITE=MI300 pytest basic_correctness/ -v -s -m 'distributed(num_gpus=2)'
848 - pytest -v -s -x lora/test_mixtral.py
849 + - pytest -v -s models/test_vision.py -m 'distributed(num_gpus=4)'
850
828 -- label: Distributed Torchrun + Examples (4 GPUs) # TBD
851 +- label: ":amd: (MI300) Distributed Torchrun + Examples" # TBD
852 timeout_in_minutes: 180
853 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
854 dind: false
@@ -851,7 +874,7 @@ steps:
874 - VLLM_ALLOW_INSECURE_SERIALIZATION=1 python3 ../examples/rl/rlhf_http_nccl.py
875 - VLLM_ALLOW_INSECURE_SERIALIZATION=1 python3 ../examples/rl/rlhf_http_ipc.py
876
854 -- label: Elastic EP Scaling Test # TBD
877 +- label: ":amd: (MI300) Elastic EP Scaling" # TBD
878 timeout_in_minutes: 180
879 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
880 dind: false
@@ -869,7 +892,7 @@ steps:
892 commands:
893 - pytest -v -s distributed/test_elastic_ep.py
894
872 -- label: RayExecutorV2 (4 GPUs) # TBD
895 +- label: ":amd: (MI300) RayExecutorV2" # TBD
896 timeout_in_minutes: 180
897 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
898 dind: false
@@ -892,7 +915,7 @@ steps:
915 - pytest -v -s distributed/test_pipeline_parallel.py -k "ray"
916 - TARGET_TEST_SUITE=L4 pytest -v -s basic_correctness/test_basic_correctness.py -k "ray"
917
895 -- label: Distributed Tests (8xH100-8xMI300) # TBD
918 +- label: ":amd: (MI300) Distributed DP + EP" # TBD
919 timeout_in_minutes: 180
920 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
921 dind: false
@@ -911,7 +934,7 @@ steps:
934 commands:
935 - torchrun --nproc-per-node=8 ../examples/features/torchrun/torchrun_dp_example_offline.py --tp-size=2 --pp-size=1 --dp-size=4 --enable-ep
936
914 -- label: Distributed Torchrun + Shutdown Tests (2 GPUs) # TBD
937 +- label: ":amd: (MI300) Distributed Torchrun + Shutdown" # TBD
938 timeout_in_minutes: 180
939 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
940 dind: false
@@ -935,7 +958,7 @@ steps:
958 - HIP_VISIBLE_DEVICES=0,1 pytest -v -s v1/shutdown
959 - pytest -v -s v1/worker/test_worker_memory_snapshot.py
960
938 -- label: Distributed Compile + Comm (4 GPUs) # TBD
961 +- label: ":amd: (MI300) Distributed Compile + Comm" # TBD
962 timeout_in_minutes: 180
963 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
964 dind: false
@@ -959,7 +982,7 @@ steps:
982
983 #---------------------------------------------------------- mi300 · engine -----------------------------------------------------------#
984
962 -- label: Engine # TBD
985 +- label: ":amd: (MI300) Engine Core" # TBD
986 timeout_in_minutes: 180
987 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
988 dind: false
@@ -973,12 +996,14 @@ steps:
996 - tests/test_config
997 - tests/test_logger
998 - tests/test_vllm_port
999 + - tests/jit_monitor/test_hooks.py
1000 + - tests/jit_monitor/test_hooks_gpu.py
1001 commands:
1002 - pytest -v -s engine test_sequence.py test_config.py test_logger.py test_vllm_port.py jit_monitor/test_hooks.py jit_monitor/test_hooks_gpu.py
1003
1004 #-------------------------------------------------------- mi300 · entrypoints --------------------------------------------------------#
1005
981 -- label: Entrypoints Unit Tests # TBD
1006 +- label: ":amd: (MI300) Entrypoints Unit" # TBD
1007 timeout_in_minutes: 180
1008 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1009 dind: false
@@ -997,7 +1022,7 @@ steps:
1022 - pytest -v -s entrypoints/weight_transfer
1023 - pytest -v -s entrypoints/launchers
1024
1000 -- label: Entrypoints Integration (LLM) # TBD
1025 +- label: ":amd: (MI300) Entrypoints Integration (LLM)" # TBD
1026 timeout_in_minutes: 180
1027 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1028 dind: false
@@ -1014,7 +1039,7 @@ steps:
1039 - pytest -v -s entrypoints/llm/test_generate.py # it needs a clean process
1040 - pytest -v -s entrypoints/llm/offline_mode # Needs to avoid interference with other tests
1041
1017 -- label: Entrypoints Integration (API Server) # TBD
1042 +- label: ":amd: (MI300) Entrypoints Integration (API Server)" # TBD
1043 timeout_in_minutes: 180
1044 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1045 dind: false
@@ -1032,7 +1057,7 @@ steps:
1057 - PYTHONPATH=/vllm-workspace pytest -v -s entrypoints/serve/dev/rpc
1058 - pytest -v -s entrypoints/scale_out
1059
1035 -- label: Entrypoints Integration (API Server OpenAI - Part 1) # TBD
1060 +- label: ":amd: (MI300) Entrypoints Integration (API Server OpenAI - Part 1)" # TBD
1061 timeout_in_minutes: 180
1062 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1063 dind: false
@@ -1048,7 +1073,7 @@ steps:
1073 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
1074 - pytest -v -s entrypoints/openai/ --ignore=entrypoints/openai/completion --ignore=entrypoints/openai/chat_completion --ignore=entrypoints/openai/responses --ignore=entrypoints/openai/correctness
1075
1051 -- label: Entrypoints Integration (API Server OpenAI - Part 2) # TBD
1076 +- label: ":amd: (MI300) Entrypoints Integration (API Server OpenAI - Part 2)" # TBD
1077 timeout_in_minutes: 180
1078 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1079 dind: false
@@ -1065,7 +1090,7 @@ steps:
1090 - pytest -v -s entrypoints/openai/chat_completion
1091 - pytest -v -s entrypoints/openai/completion --ignore=entrypoints/openai/completion/test_tensorizer_entrypoint.py
1092
1068 -- label: Entrypoints Integration (API Server Generate) # TBD
1093 +- label: ":amd: (MI300) Entrypoints Integration (API Server Generate)" # TBD
1094 timeout_in_minutes: 180
1095 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1096 dind: false
@@ -1086,7 +1111,7 @@ steps:
1111 - pytest -v -s entrypoints/generate
1112 - pytest -v -s entrypoints/anthropic
1113
1089 -- label: Entrypoints Integration (Responses API) # TBD
1114 +- label: ":amd: (MI300) Entrypoints Integration (Responses API)" # TBD
1115 timeout_in_minutes: 180
1116 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1117 dind: false
@@ -1101,7 +1126,7 @@ steps:
1126 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
1127 - pytest -v -s entrypoints/openai/responses
1128
1104 -- label: Entrypoints Integration (Speech to Text) # TBD
1129 +- label: ":amd: (MI300) Entrypoints Integration (Speech to Text)" # TBD
1130 timeout_in_minutes: 180
1131 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1132 dind: false
@@ -1115,7 +1140,7 @@ steps:
1140 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
1141 - pytest -v -s entrypoints/speech_to_text
1142
1118 -- label: Entrypoints Integration (Multimodal)
1143 +- label: ":amd: (MI300) Entrypoints Integration (Multimodal)"
1144 timeout_in_minutes: 180
1145 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1146 dind: false
@@ -1130,7 +1155,7 @@ steps:
1155 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
1156 - pytest -v -s entrypoints/multimodal
1157
1133 -- label: Entrypoints Integration (Pooling) # TBD
1158 +- label: ":amd: (MI300) Entrypoints Integration (Pooling)" # TBD
1159 timeout_in_minutes: 180
1160 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1161 dind: false
@@ -1145,7 +1170,7 @@ steps:
1170 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
1171 - pytest -v -s entrypoints/pooling
1172
1148 -- label: OpenAI API correctness # TBD
1173 +- label: ":amd: (MI300) OpenAI API Correctness" # TBD
1174 timeout_in_minutes: 180
1175 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1176 dind: false
@@ -1167,7 +1192,7 @@ steps:
1192
1193 #----------------------------------------------------------- mi300 · evals -----------------------------------------------------------#
1194
1170 -- label: DeepSeek V2-Lite Prefetch Offload Accuracy (H100-MI300) # TBD
1195 +- label: ":amd: (MI300) DeepSeek V2-Lite Prefetch Offload Accuracy" # TBD
1196 timeout_in_minutes: 180
1197 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1198 dind: false
@@ -1188,7 +1213,7 @@ steps:
1213 commands:
1214 - bash .buildkite/scripts/scheduled_integration_test/deepseek_v2_lite_prefetch_offload.sh 0.25 200 8030
1215
1191 -- label: LM Eval Small Models # TBD
1216 +- label: ":amd: (MI300) LM Eval Small Models" # TBD
1217 timeout_in_minutes: 180
1218 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1219 dind: false
@@ -1206,7 +1231,7 @@ steps:
1231 commands:
1232 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-small.txt
1233
1209 -- label: MRCR Eval Small Models # TBD
1234 +- label: ":amd: (MI300) MRCR Eval Small Models" # TBD
1235 timeout_in_minutes: 180
1236 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1237 dind: false
@@ -1224,7 +1249,7 @@ steps:
1249 commands:
1250 - pytest -s -v evals/mrcr/test_mrcr_correctness.py --config-list-file=evals/mrcr/configs/models-small.txt
1251
1227 -- label: LM Eval Small Models (MI300) # TBD
1252 +- label: ":amd: (MI300) LM Eval Small Models Harness" # TBD
1253 timeout_in_minutes: 180
1254 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1255 dind: false
@@ -1243,7 +1268,7 @@ steps:
1268 commands:
1269 - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-small-rocm.txt
1270
1246 -- label: Multi-Modal Accuracy Eval (Small Models) # TBD
1271 +- label: ":amd: (MI300) Multimodal Accuracy Eval (Small Models)" # TBD
1272 timeout_in_minutes: 180
1273 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1274 dind: false
@@ -1259,7 +1284,7 @@ steps:
1284 commands:
1285 - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-mm-small.txt --tp-size=1
1286
1262 -- label: GPQA Eval (GPT-OSS) (2xH100-2xMI300) # TBD
1287 +- label: ":amd: (MI300) GPQA Eval (GPT-OSS)" # TBD
1288 timeout_in_minutes: 180
1289 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1290 dind: false
@@ -1282,7 +1307,7 @@ steps:
1307 - uv pip install --system 'gpt-oss[eval]==0.0.5'
1308 - pytest -s -v evals/gpt_oss/test_gpqa_correctness.py --config-list-file=configs/models-gfx942.txt
1309
1285 -- label: LM Eval Small Models (2xB200-2xMI300) # TBD
1310 +- label: ":amd: (MI300) LM Eval Small Models FP8 + Mixed" # TBD
1311 timeout_in_minutes: 180
1312 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1313 dind: false
@@ -1301,7 +1326,7 @@ steps:
1326 commands:
1327 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-mi3xx-fp8-and-mixed.txt
1328
1304 -- label: DeepSeek V2-Lite Sync EPLB Accuracy (4xH100-4xMI300) # TBD
1329 +- label: ":amd: (MI300) DeepSeek V2-Lite Sync EPLB Accuracy" # TBD
1330 timeout_in_minutes: 180
1331 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1332 dind: false
@@ -1324,7 +1349,7 @@ steps:
1349 commands:
1350 - bash .buildkite/scripts/scheduled_integration_test/deepseek_v2_lite_ep_eplb.sh 0.25 200 8010
1351
1327 -- label: LM Eval Large Models (4xA100-4xMI300) # TBD
1352 +- label: ":amd: (MI300) LM Eval Large Models Harness" # TBD
1353 timeout_in_minutes: 180
1354 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1355 dind: false
@@ -1345,7 +1370,7 @@ steps:
1370 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
1371 - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large.txt --tp-size=4
1372
1348 -- label: Qwen3-30B-A3B-FP8-block Sync EPLB Accuracy (4xH100-4xMI300) # TBD
1373 +- label: ":amd: (MI300) Qwen3-30B-A3B-FP8-block Sync EPLB Accuracy" # TBD
1374 timeout_in_minutes: 180
1375 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1376 dind: false
@@ -1367,7 +1392,7 @@ steps:
1392 commands:
1393 - bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_block_ep_eplb.sh 0.8 200 8020
1394
1370 -- label: Qwen3-30B-A3B-FP8 DP4 Async EPLB Accuracy (4xH100-4xMI300) # TBD
1395 +- label: ":amd: (MI300) Qwen3-30B-A3B-FP8 DP4 Async EPLB Accuracy" # TBD
1396 timeout_in_minutes: 180
1397 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1398 dind: false
@@ -1389,7 +1414,7 @@ steps:
1414 commands:
1415 - bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_dp4_async_eplb.sh 0.8 200 8050
1416
1392 -- label: Qwen3-Next-80B-A3B-Instruct MTP Async EPLB Accuracy # TBD
1417 +- label: ":amd: (MI300) Qwen3-Next-80B-A3B-Instruct MTP Async EPLB Accuracy" # TBD
1418 timeout_in_minutes: 180
1419 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1420 dind: false
@@ -1412,7 +1437,7 @@ steps:
1437 commands:
1438 - bash .buildkite/scripts/scheduled_integration_test/qwen3_next_mtp_async_eplb.sh 0.8 1319 8040
1439
1415 -- label: LM Eval Large Models (8xH200-8xMI300) # TBD
1440 +- label: ":amd: (MI300) LM Eval Large Models" # TBD
1441 timeout_in_minutes: 180
1442 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1443 dind: false
@@ -1433,9 +1458,10 @@ steps:
1458 - tests/evals/
1459 commands:
1460 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
1461 + - export PYTORCH_ROCM_ARCH=gfx942
1462 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-mi3xx.txt
1463
1438 -- label: ROCm LM Eval Large Models (8 GPUs) # TBD
1464 +- label: ":amd: (MI300) LM Eval Large Models ROCm Harness" # TBD
1465 timeout_in_minutes: 180
1466 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1467 dind: false
@@ -1457,7 +1483,7 @@ steps:
1483 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
1484 - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large-rocm.txt --tp-size=8
1485
1460 -- label: LM Eval Large Models (4xH100-4xMI300) # TBD
1486 +- label: ":amd: (MI300) LM Eval Large Models FP8" # TBD
1487 timeout_in_minutes: 180
1488 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1489 dind: false
@@ -1480,7 +1506,7 @@ steps:
1506
1507 #--------------------------------------------------------- mi300 · examples ----------------------------------------------------------#
1508
1483 -- label: Examples # TBD
1509 +- label: ":amd: (MI300) Examples" # TBD
1510 timeout_in_minutes: 180
1511 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1512 dind: false
@@ -1517,7 +1543,7 @@ steps:
1543
1544 #---------------------------------------------------------- mi300 · kernels ----------------------------------------------------------#
1545
1520 -- label: vLLM IR Tests # TBD
1546 +- label: ":amd: (MI300) vLLM IR" # TBD
1547 timeout_in_minutes: 180
1548 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1549 dind: false
@@ -1533,7 +1559,7 @@ steps:
1559 - pytest -v -s tests/ir
1560 - pytest -v -s tests/kernels/ir
1561
1536 -- label: Kernels MLA (MI300) # TBD
1562 +- label: ":amd: (MI300) MLA Kernels" # TBD
1563 timeout_in_minutes: 180
1564 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1565 dind: false
@@ -1555,7 +1581,7 @@ steps:
1581 - pytest -v -s tests/kernels/attention/test_rocm_aiter_mla_fp8_support.py
1582 - pytest -v -s tests/kernels/attention/test_rocm_aiter_mla_op_registration.py
1583
1558 -- label: Kernels Attention Test %N # TBD
1584 +- label: ":amd: (MI300) Attention Kernels Shard %N" # TBD
1585 timeout_in_minutes: 180
1586 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1587 dind: false
@@ -1574,7 +1600,7 @@ steps:
1600 commands:
1601 - pytest -v -s kernels/attention --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
1602
1577 -- label: Kernels Core Operation Test %N # TBD
1603 +- label: ":amd: (MI300) Core Operation Kernels Shard %N" # TBD
1604 timeout_in_minutes: 180
1605 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1606 dind: false
@@ -1592,7 +1618,7 @@ steps:
1618 commands:
1619 - pytest -v -s kernels/core --ignore=kernels/core/test_minimax_reduce_rms.py kernels/test_concat_mla_q.py kernels/test_top_k_per_row.py --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
1620
1595 -- label: Kernels KDA Test # TBD
1621 +- label: ":amd: (MI300) KDA Kernels" # TBD
1622 timeout_in_minutes: 180
1623 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1624 dind: false
@@ -1612,7 +1638,7 @@ steps:
1638 commands:
1639 - pytest -v -s models/kimi_k3/test_kda.py models/kimi_k3/test_kda_metadata.py
1640
1615 -- label: Kernels Mamba Test # TBD
1641 +- label: ":amd: (MI300) Mamba Kernels" # TBD
1642 timeout_in_minutes: 180
1643 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1644 dind: false
@@ -1627,7 +1653,7 @@ steps:
1653 commands:
1654 - pytest -v -s kernels/mamba
1655
1630 -- label: Kernels MoE Test %N # TBD
1656 +- label: ":amd: (MI300) MoE Kernels Shard %N" # TBD
1657 timeout_in_minutes: 180
1658 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1659 dind: false
@@ -1653,7 +1679,7 @@ steps:
1679 - pytest -v -s kernels/moe --ignore=kernels/moe/test_modular_oai_triton_moe.py --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
1680 - pytest -v -s kernels/moe/test_modular_oai_triton_moe.py --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
1681
1656 -- label: Kernels Quantization Test %N # TBD
1682 +- label: ":amd: (MI300) Quantization Kernels" # TBD
1683 timeout_in_minutes: 180
1684 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1685 dind: false
@@ -1677,7 +1703,7 @@ steps:
1703 commands:
1704 - pytest -v -s kernels/quantization
1705
1680 -- label: Kernels FP8 MoE Test (2xH100-2xMI300) # TBD
1706 +- label: ":amd: (MI300) DeepEP FP8 MoE Kernels" # TBD
1707 timeout_in_minutes: 180
1708 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1709 dind: false
@@ -1698,7 +1724,7 @@ steps:
1724
1725 #----------------------------------------------------------- mi300 · lora ------------------------------------------------------------#
1726
1701 -- label: LoRA %N # TBD
1727 +- label: ":amd: (MI300) LoRA Shard %N" # TBD
1728 timeout_in_minutes: 180
1729 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1730 dind: false
@@ -1713,7 +1739,7 @@ steps:
1739 commands:
1740 - pytest -v -s lora --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --ignore=lora/test_chatglm3_tp.py --ignore=lora/test_llama_tp.py --ignore=lora/test_qwen3_with_multi_loras.py --ignore=lora/test_olmoe_tp.py --ignore=lora/test_deepseekv2_tp.py --ignore=lora/test_gptoss_tp.py --ignore=lora/test_qwen3moe_tp.py --ignore=lora/test_qwen35_densemodel_lora.py
1741
1716 -- label: LoRA TP (Distributed) # TBD
1742 +- label: ":amd: (MI300) LoRA TP (Distributed)" # TBD
1743 timeout_in_minutes: 180
1744 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1745 dind: false
@@ -1722,6 +1748,7 @@ steps:
1748 working_dir: "/vllm-workspace/tests"
1749 source_file_dependencies:
1750 - vllm/lora
1751 + - vllm/model_executor/layers/fused_moe/
1752 - tests/lora
1753 - vllm/platforms/rocm.py
1754 commands:
@@ -1731,10 +1758,11 @@ steps:
1758 - pytest -v -s -x lora/test_olmoe_tp.py
1759 - pytest -v -s -x lora/test_gptoss_tp.py
1760 - pytest -v -s -x lora/test_qwen35_densemodel_lora.py
1761 + - pytest -v -s -x lora/test_gemma4_tp.py
1762
1763 #------------------------------------------------------ mi300 · model_executor -------------------------------------------------------#
1764
1737 -- label: Model Executor # TBD
1765 +- label: ":amd: (MI300) Model Executor" # TBD
1766 timeout_in_minutes: 180
1767 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1768 dind: false
@@ -1757,7 +1785,7 @@ steps:
1785
1786 #---------------------------------------------------- mi300 · model_runner_v2 -------------------------------------------------------#
1787
1760 -- label: Model Runner V2 Core Tests # TBD
1788 +- label: ":amd: (MI300) Model Runner V2 Core" # TBD
1789 timeout_in_minutes: 180
1790 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1791 dind: false
@@ -1782,7 +1810,7 @@ steps:
1810 - pytest -v -s v1/e2e/general/test_min_tokens.py
1811 - pytest -v -s entrypoints/llm/test_struct_output_generate.py -k "xgrammar and not speculative_config6 and not speculative_config7 and not speculative_config8 and not speculative_config0"
1812
1785 -- label: Model Runner V2 Examples # TBD
1813 +- label: ":amd: (MI300) Model Runner V2 Examples" # TBD
1814 timeout_in_minutes: 180
1815 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1816 dind: false
@@ -1817,7 +1845,7 @@ steps:
1845 - python3 features/speculative_decoding/spec_decode_offline.py --test --method eagle --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 2048
1846 - python3 features/speculative_decoding/spec_decode_offline.py --test --method eagle3 --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 1536
1847
1820 -- label: Model Runner V2 Distributed (2 GPUs) # TBD
1848 +- label: ":amd: (MI300) Model Runner V2 Distributed" # TBD
1849 timeout_in_minutes: 180
1850 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1851 dind: false
@@ -1839,7 +1867,7 @@ steps:
1867 - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_async_llm_dp.py -k "not ray"
1868 - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_eagle_dp.py
1869
1842 -- label: Model Runner V2 Pipeline Parallelism (4 GPUs) # TBD
1870 +- label: ":amd: (MI300) Model Runner V2 Pipeline Parallelism" # TBD
1871 timeout_in_minutes: 180
1872 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1873 dind: false
@@ -1859,7 +1887,7 @@ steps:
1887 - pytest -v -s distributed/test_pipeline_parallel.py -k "not ray"
1888 - pytest -v -s distributed/test_pp_cudagraph.py -k "not ray"
1889
1862 -- label: Model Runner V2 Spec Decode # TBD
1890 +- label: ":amd: (MI300) Model Runner V2 Spec Decode" # TBD
1891 timeout_in_minutes: 180
1892 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1893 dind: false
@@ -1886,7 +1914,7 @@ steps:
1914
1915 #------------------------------------------------------ mi300 · models / basic -------------------------------------------------------#
1916
1889 -- label: Basic Models Tests (Extra Initialization) %N # TBD
1917 +- label: ":amd: (MI300) Basic Models (Extra Initialization) Shard %N" # TBD
1918 timeout_in_minutes: 180
1919 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1920 dind: false
@@ -1903,7 +1931,7 @@ steps:
1931 commands:
1932 - pytest -v -s models/test_initialization.py -k 'not test_can_initialize_small_subset' --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
1933
1906 -- label: Basic Models Tests (Initialization) # TBD
1934 +- label: ":amd: (MI300) Basic Models (Initialization)" # TBD
1935 timeout_in_minutes: 180
1936 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1937 dind: false
@@ -1916,7 +1944,7 @@ steps:
1944 commands:
1945 - pytest -v -s models/test_initialization.py::test_can_initialize_small_subset
1946
1919 -- label: Basic Models Tests (Other) # TBD
1947 +- label: ":amd: (MI300) Basic Models (Other)" # TBD
1948 timeout_in_minutes: 180
1949 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1950 dind: false
@@ -1928,12 +1956,14 @@ steps:
1956 - tests/models/test_terratorch.py
1957 - tests/models/transformers/test_backend.py
1958 - tests/models/test_registry.py
1959 + - tests/models/test_vision.py
1960 commands:
1961 - pytest -v -s models/test_terratorch.py models/transformers/test_backend.py models/test_registry.py
1962 + - pytest -v -s models/test_vision.py::test_simple_mrope_vision_model_spatial_merge
1963
1964 #----------------------------------------------------- mi300 · models / language -----------------------------------------------------#
1965
1936 -- label: Language Models Test (Extended Pooling) # TBD
1966 +- label: ":amd: (MI300) Language Models (Extended Pooling)" # TBD
1967 timeout_in_minutes: 180
1968 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1969 dind: false
@@ -1946,7 +1976,7 @@ steps:
1976 commands:
1977 - pytest -v -s models/language/pooling -m 'not core_model'
1978
1949 -- label: Language Models Tests (Standard) # TBD
1979 +- label: ":amd: (MI300) Language Models (Standard)" # TBD
1980 timeout_in_minutes: 180
1981 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1982 dind: false
@@ -1960,7 +1990,7 @@ steps:
1990 - pip freeze | grep -E 'torch'
1991 - pytest -v -s models/language -m 'core_model and (not slow_test)'
1992
1963 -- label: Language Models Tests (Extra Standard) %N # TBD
1993 +- label: ":amd: (MI300) Language Models (Extra Standard) Shard %N" # TBD
1994 timeout_in_minutes: 180
1995 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1996 dind: false
@@ -1983,7 +2013,7 @@ steps:
2013 - pip freeze | grep -E 'torch'
2014 - pytest -v -s models/language -m 'core_model and slow_test' --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
2015
1986 -- label: Language Models Test (Extended Generation) # TBD
2016 +- label: ":amd: (MI300) Language Models (Extended Generation)" # TBD
2017 timeout_in_minutes: 180
2018 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2019 dind: false
@@ -1998,7 +2028,7 @@ steps:
2028 - CAUSAL_CONV1D_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
2029 - pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
2030
2001 -- label: Language Models Tests (Hybrid) %N # TBD
2031 +- label: ":amd: (MI300) Language Models (Hybrid) Shard %N" # TBD
2032 timeout_in_minutes: 180
2033 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2034 dind: false
@@ -2016,7 +2046,7 @@ steps:
2046
2047 #---------------------------------------------------- mi300 · models / multimodal ----------------------------------------------------#
2048
2019 -- label: Multi-Modal Models (Extended Generation 1) # TBD
2049 +- label: ":amd: (MI300) Multimodal Models (Extended Generation 1)" # TBD
2050 timeout_in_minutes: 180
2051 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2052 dind: false
@@ -2031,7 +2061,7 @@ steps:
2061 - pytest -v -s models/multimodal/generation -m 'not core_model' --ignore models/multimodal/generation/test_common.py
2062 - pytest -v -s models/multimodal/test_mapping.py
2063
2034 -- label: Multi-Modal Models (Extended Generation 2) # TBD
2064 +- label: ":amd: (MI300) Multimodal Models (Extended Generation 2)" # TBD
2065 timeout_in_minutes: 180
2066 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2067 dind: false
@@ -2045,7 +2075,7 @@ steps:
2075 - pytest -v -s models/multimodal/generation/test_common.py -m 'split(group=0) and not core_model'
2076
2077
2048 -- label: Multi-Modal Models (Extended Generation 3) # TBD
2078 +- label: ":amd: (MI300) Multimodal Models (Extended Generation 3)" # TBD
2079 timeout_in_minutes: 180
2080 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2081 dind: false
@@ -2058,7 +2088,7 @@ steps:
2088 commands:
2089 - pytest -v -s models/multimodal/generation/test_common.py -m 'split(group=1) and not core_model'
2090
2061 -- label: "Multi-Modal Models (Standard) 1: qwen2" # TBD
2091 +- label: ":amd: (MI300) Multimodal Models (Standard) 1: qwen2" # TBD
2092 timeout_in_minutes: 180
2093 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2094 dind: false
@@ -2072,7 +2102,7 @@ steps:
2102 - pytest -v -s models/multimodal/generation/test_common.py -m core_model -k "qwen2"
2103 - pytest -v -s models/multimodal/generation/test_ultravox.py -m core_model
2104
2075 -- label: "Multi-Modal Models (Standard) 3: llava + qwen2_vl" # TBD
2105 +- label: ":amd: (MI300) Multimodal Models (Standard) 3: llava + qwen2_vl" # TBD
2106 timeout_in_minutes: 180
2107 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2108 dind: false
@@ -2087,7 +2117,7 @@ steps:
2117 - pytest -v -s models/multimodal/generation/test_common.py -m core_model -k "not qwen2 and not qwen3 and not gemma"
2118 - pytest -v -s models/multimodal/generation/test_qwen2_vl.py -m core_model
2119
2090 -- label: "Multi-Modal Models (Standard) 4: other + whisper" # TBD
2120 +- label: ":amd: (MI300) Multimodal Models (Standard) 4: other + whisper" # TBD
2121 timeout_in_minutes: 180
2122 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2123 dind: false
@@ -2095,14 +2125,14 @@ steps:
2125 working_dir: "/vllm-workspace/tests"
2126 source_file_dependencies:
2127 - vllm/
2098 - - tests/models/multimodal/generation
2099 - - tests/models/multimodal/test_mapping.py
2128 + - tests/models/multimodal
2129 commands:
2101 - - pytest -v -s models/multimodal -m core_model --ignore models/multimodal/generation/test_common.py --ignore models/multimodal/generation/test_ultravox.py --ignore models/multimodal/generation/test_qwen2_5_vl.py --ignore models/multimodal/generation/test_qwen2_vl.py --ignore models/multimodal/generation/test_whisper.py --ignore models/multimodal/generation/test_memory_leak.py --ignore models/multimodal/generation/test_vit_cudagraph.py --ignore models/multimodal/processing
2130 + - pytest -v -s models/multimodal -m core_model --ignore models/multimodal/generation/test_common.py --ignore models/multimodal/generation/test_ultravox.py --ignore models/multimodal/generation/test_qwen2_5_vl.py --ignore models/multimodal/generation/test_qwen2_vl.py --ignore models/multimodal/generation/test_whisper.py --ignore models/multimodal/generation/test_mm_prefix_lm.py --ignore models/multimodal/generation/test_memory_leak.py --ignore models/multimodal/generation/test_vit_cudagraph.py --ignore models/multimodal/processing
2131 + - pytest -v -s models/multimodal/generation/test_vit_cudagraph.py -m core_model
2132 - pytest -v -s models/multimodal/generation/test_memory_leak.py -m core_model
2133 - cd .. && VLLM_WORKER_MULTIPROC_METHOD=spawn pytest -v -s tests/models/multimodal/generation/test_whisper.py -m core_model
2134
2105 -- label: Multi-Modal Processor # 1h 42m
2135 +- label: ":amd: (MI300) Multimodal Processor" # 1h 42m
2136 timeout_in_minutes: 180
2137 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2138 dind: false
@@ -2116,7 +2146,7 @@ steps:
2146 commands:
2147 - pytest -v -s models/multimodal/processing/test_tensor_schema.py
2148
2119 -- label: Multi-Modal Models (Extended Pooling) # TBD
2149 +- label: ":amd: (MI300) Multimodal Models (Extended Pooling)" # TBD
2150 timeout_in_minutes: 180
2151 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2152 dind: false
@@ -2129,7 +2159,7 @@ steps:
2159 commands:
2160 - pytest -v -s models/multimodal/pooling -m 'not core_model'
2161
2132 -- label: "Multi-Modal Models (Standard) 2: qwen3 + gemma" # TBD
2162 +- label: ":amd: (MI300) Multimodal Models (Standard) 2: qwen3 + gemma" # TBD
2163 timeout_in_minutes: 180
2164 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2165 dind: false
@@ -2141,11 +2171,12 @@ steps:
2171 - tests/models/multimodal
2172 commands:
2173 - pytest -v -s models/multimodal/generation/test_common.py -m core_model -k "qwen3 or gemma"
2174 + - pytest -v -s models/multimodal/generation/test_mm_prefix_lm.py -m core_model
2175 - pytest -v -s models/multimodal/generation/test_qwen2_5_vl.py -m core_model
2176
2177 #----------------------------------------------------- mi300 · models / quantized -----------------------------------------------------#
2178
2148 -- label: Quantized Models Test # TBD
2179 +- label: ":amd: (MI300) Quantized Models" # TBD
2180 timeout_in_minutes: 180
2181 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2182 dind: false
@@ -2162,7 +2193,7 @@ steps:
2193
2194 #-------------------------------------------------- mi300 · models / transformers ---------------------------------------------------#
2195
2165 -- label: Transformers Nightly Models (Initialization) %N # TBD
2196 +- label: ":amd: (MI300) Transformers Nightly Models (Initialization) Shard %N" # TBD
2197 timeout_in_minutes: 180
2198 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2199 dind: false
@@ -2184,7 +2215,7 @@ steps:
2215 - pip install --upgrade git+https://github.com/huggingface/transformers
2216 - pytest -v -s tests/models/test_initialization.py --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
2217
2187 -- label: Transformers Nightly Models (Processing) %N # TBD
2218 +- label: ":amd: (MI300) Transformers Nightly Models (Processing) Shard %N" # TBD
2219 timeout_in_minutes: 180
2220 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2221 dind: false
@@ -2206,7 +2237,7 @@ steps:
2237 - pip install --upgrade git+https://github.com/huggingface/transformers
2238 - pytest -v -s tests/models/multimodal/processing/ --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
2239
2209 -- label: Transformers Nightly Models (Single) # TBD
2240 +- label: ":amd: (MI300) Transformers Nightly Models (Single)" # TBD
2241 timeout_in_minutes: 180
2242 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2243 dind: false
@@ -2234,7 +2265,7 @@ steps:
2265
2266 #---------------------------------------------------------- mi300 · plugins ----------------------------------------------------------#
2267
2237 -- label: Plugin Tests (2 GPUs) # TBD
2268 +- label: ":amd: (MI300) Plugin Integration" # TBD
2269 timeout_in_minutes: 180
2270 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2271 dind: false
@@ -2286,7 +2317,7 @@ steps:
2317 - pytest -v -s plugins_tests/test_oot_registration_offline.py # it needs a clean process
2318 - pytest -v -s plugins_tests/lora_resolvers # unit tests for in-tree lora resolver plugins
2319
2289 -- label: GGUF Plugin # TBD
2320 +- label: ":amd: (MI300) GGUF Plugin" # TBD
2321 timeout_in_minutes: 180
2322 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2323 dind: false
@@ -2305,7 +2336,7 @@ steps:
2336
2337 #------------------------------------------------------- mi300 · rust_frontend -------------------------------------------------------#
2338
2308 -- label: Rust Frontend OpenAI Coverage # TBD
2339 +- label: ":amd: (MI300) Rust Frontend OpenAI Coverage" # TBD
2340 timeout_in_minutes: 180
2341 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2342 dind: false
@@ -2338,7 +2369,7 @@ steps:
2369 - pytest -v -s entrypoints/serve/instrumentator/test_uds.py
2370 - pytest -v -s v1/sample/test_logprobs_e2e.py -k "test_prompt_logprobs_e2e_server"
2371
2341 -- label: Rust Frontend Serve Admin Coverage # TBD
2372 +- label: ":amd: (MI300) Rust Frontend Serve + Admin Coverage" # TBD
2373 timeout_in_minutes: 180
2374 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2375 dind: false
@@ -2366,7 +2397,7 @@ steps:
2397 - pytest -v -s entrypoints/serve/instrumentator/test_metrics.py -k "text and not show and not run_batch and not test_metrics_counts and not test_metrics_exist"
2398 - pytest -v -s entrypoints/serve/tokenize/test_tokenization.py -k "not tokenizer_info"
2399
2369 -- label: Rust Frontend Core Correctness # TBD
2400 +- label: ":amd: (MI300) Rust Frontend Core Correctness" # TBD
2401 timeout_in_minutes: 180
2402 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2403 dind: false
@@ -2384,7 +2415,7 @@ steps:
2415 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
2416 - pytest -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine
2417
2387 -- label: Rust Frontend Tool Use # TBD
2418 +- label: ":amd: (MI300) Rust Frontend Tool Use" # TBD
2419 timeout_in_minutes: 180
2420 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2421 dind: false
@@ -2403,7 +2434,7 @@ steps:
2434 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
2435 - pytest -v -s tool_use --ignore=tool_use/mistral --models llama3.2 -k "not test_response_format_with_tool_choice_required and not test_parallel_tool_calls_false and not test_tool_call_and_choice"
2436
2406 -- label: Rust Frontend Distributed # TBD
2437 +- label: ":amd: (MI300) Rust Frontend Distributed" # TBD
2438 timeout_in_minutes: 180
2439 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2440 dind: false
@@ -2434,7 +2465,7 @@ steps:
2465
2466 #------------------------------------------------------- mi300 · quantization --------------------------------------------------------#
2467
2437 -- label: Quantization # TBD
2468 +- label: ":amd: (MI300) Quantization" # TBD
2469 timeout_in_minutes: 180
2470 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2471 dind: false
@@ -2461,7 +2492,7 @@ steps:
2492
2493 #--------------------------------------------------------- mi300 · samplers ----------------------------------------------------------#
2494
2464 -- label: Samplers Test # TBD
2495 +- label: ":amd: (MI300) Samplers" # TBD
2496 timeout_in_minutes: 180
2497 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2498 dind: false
@@ -2482,7 +2513,7 @@ steps:
2513
2514 #------------------------------------------------------------ mi300 · misc ------------------------------------------------------------#
2515
2485 -- label: Regression # TBD
2516 +- label: ":amd: (MI300) Regression" # TBD
2517 timeout_in_minutes: 180
2518 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2519 dind: false
@@ -2497,7 +2528,7 @@ steps:
2528
2529 #--------------------------------------------------------- mi300 · ray_compat ---------------------------------------------------------#
2530
2500 -- label: Ray Dependency Compatibility Check # TBD
2531 +- label: ":amd: (MI300) Ray Dependency Compatibility Check" # TBD
2532 timeout_in_minutes: 180
2533 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2534 dind: false
@@ -2512,7 +2543,7 @@ steps:
2543
2544 #------------------------------------------------------------ mi300 · v1 -------------------------------------------------------------#
2545
2515 -- label: Acceptance Length Test (Large Models) # TBD
2546 +- label: ":amd: (MI300) Acceptance Length (Large Models)" # TBD
2547 timeout_in_minutes: 180
2548 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2549 dind: false
@@ -2528,7 +2559,7 @@ steps:
2559 - export VLLM_ALLOW_INSECURE_SERIALIZATION=1
2560 - pytest -v -s v1/spec_decode/test_acceptance_length.py -m slow_test
2561
2531 -- label: e2e Scheduling (1 GPU) # TBD
2562 +- label: ":amd: (MI300) E2E Scheduling" # TBD
2563 timeout_in_minutes: 180
2564 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2565 dind: false
@@ -2542,7 +2573,7 @@ steps:
2573 commands:
2574 - pytest -v -s v1/e2e/general/test_async_scheduling.py
2575
2545 -- label: Engine (1 GPU) # TBD
2576 +- label: ":amd: (MI300) V1 Engine" # TBD
2577 timeout_in_minutes: 180
2578 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2579 dind: false
@@ -2557,7 +2588,7 @@ steps:
2588 - pytest -v -s v1/engine/test_preprocess_error_handling.py
2589 - pytest -v -s v1/engine --ignore v1/engine/test_preprocess_error_handling.py
2590
2560 -- label: Spec Decode Draft Model # TBD
2591 +- label: ":amd: (MI300) Spec Decode Draft Model" # TBD
2592 timeout_in_minutes: 180
2593 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2594 dind: false
@@ -2575,7 +2606,7 @@ steps:
2606 commands:
2607 - pytest -v -s v1/e2e/spec_decode/draft_model/
2608
2578 -- label: Spec Decode Eagle # TBD
2609 +- label: ":amd: (MI300) Spec Decode Eagle" # TBD
2610 timeout_in_minutes: 180
2611 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2612 dind: false
@@ -2593,7 +2624,7 @@ steps:
2624 commands:
2625 - pytest -v -s v1/e2e/spec_decode/eagle/
2626
2596 -- label: Spec Decode Ngram + Suffix # TBD
2627 +- label: ":amd: (MI300) Spec Decode N-Gram + Suffix" # TBD
2628 timeout_in_minutes: 180
2629 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2630 dind: false
@@ -2607,11 +2638,13 @@ steps:
2638 - vllm/v1/sample/
2639 - vllm/model_executor/layers/
2640 - tests/v1/e2e/spec_decode/
2641 + - tests/spec_decode/
2642 - vllm/platforms/rocm.py
2643 commands:
2644 - pytest -v -s v1/e2e/spec_decode/ngram_suffix/
2645 + - python3 spec_decode/test_custom_proposer.py
2646
2614 -- label: Spec Decode Speculators + MTP # TBD
2647 +- label: ":amd: (MI300) Spec Decode Speculators + MTP" # TBD
2648 timeout_in_minutes: 180
2649 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2650 dind: false
@@ -2631,7 +2664,7 @@ steps:
2664 - pytest -v -s v1/e2e/spec_decode/speculators/
2665 - pytest -v -s v1/e2e/spec_decode/mtp/
2666
2634 -- label: Speculators Correctness # TBD
2667 +- label: ":amd: (MI300) Speculators Correctness" # TBD
2668 timeout_in_minutes: 180
2669 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2670 dind: false
@@ -2663,7 +2696,7 @@ steps:
2696 - export VLLM_ALLOW_INSECURE_SERIALIZATION=1
2697 - pytest -v -s v1/spec_decode/test_speculators_correctness.py -m slow_test
2698
2666 -- label: V1 Spec Decode # TBD
2699 +- label: ":amd: (MI300) V1 Spec Decode" # TBD
2700 timeout_in_minutes: 180
2701 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2702 dind: false
@@ -2673,9 +2706,10 @@ steps:
2706 - vllm/
2707 - tests/v1/spec_decode
2708 commands:
2709 + - export VLLM_WORKER_MULTIPROC_METHOD=spawn
2710 - pytest -v -s -m 'not slow_test' v1/spec_decode
2711
2678 -- label: Extract Hidden States Integration # TBD
2712 +- label: ":amd: (MI300) Extract Hidden States Integration" # TBD
2713 timeout_in_minutes: 180
2714 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2715 dind: false
@@ -2707,7 +2741,7 @@ steps:
2741 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
2742 - pytest -v -s v1/kv_connector/extract_hidden_states_integration
2743
2710 -- label: V1 attention (H100-MI300) %N # TBD
2744 +- label: ":amd: (MI300) V1 Attention Shard %N" # TBD
2745 timeout_in_minutes: 180
2746 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2747 dind: false
@@ -2726,7 +2760,7 @@ steps:
2760 commands:
2761 - pytest -v -s v1/attention --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
2762
2729 -- label: V1 Core + KV + Metrics # TBD
2763 +- label: ":amd: (MI300) V1 Core + KV + Metrics" # TBD
2764 timeout_in_minutes: 180
2765 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2766 dind: false
@@ -2737,23 +2771,30 @@ steps:
2771 - tests/v1/core
2772 - tests/v1/executor
2773 - tests/v1/kv_offload
2774 + - tests/v1/simple_kv_offload
2775 - tests/v1/worker
2776 + - tests/v1/streaming_input
2777 - tests/v1/kv_connector/unit
2778 + - tests/v1/ec_connector/unit
2779 - tests/v1/metrics
2780 - tests/entrypoints/openai/correctness/test_lmeval.py
2781 commands:
2782 - uv pip install --system -r /vllm-workspace/requirements/kv_connectors_rocm.txt
2783 + - export VLLM_WORKER_MULTIPROC_METHOD=spawn
2784 - pytest -v -s -m 'not cpu_test' v1/core
2785 - pytest -v -s v1/executor
2786 - pytest -v -s v1/kv_offload
2787 + - pytest -v -s v1/simple_kv_offload
2788 - pytest -v -s v1/worker
2789 + - pytest -v -s v1/streaming_input
2790 - pytest -v -s -m 'not cpu_test' v1/kv_connector/unit
2791 + - pytest -v -s -m 'not cpu_test' v1/ec_connector/unit
2792 - pytest -v -s -m 'not cpu_test' v1/metrics
2793 - pip install -U git+https://github.com/vllm-project/lm-evaluation-harness.git@streaming-api
2794 # - export HSA_NO_SCRATCH_RECLAIM=1
2795 - pytest -v -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine
2796
2756 -- label: V1 Sample + Logits # TBD
2797 +- label: ":amd: (MI300) V1 Sample + Logits" # TBD
2798 timeout_in_minutes: 180
2799 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2800 dind: false
@@ -2768,13 +2809,14 @@ steps:
2809 - tests/v1/test_request.py
2810 - tests/v1/test_outputs.py
2811 commands:
2812 + - export VLLM_WORKER_MULTIPROC_METHOD=spawn
2813 - pytest -v -s v1/sample
2814 - pytest -v -s v1/logits_processors
2815 - pytest -v -s v1/test_oracle.py
2816 - pytest -v -s v1/test_request.py
2817 - pytest -v -s v1/test_outputs.py
2818
2777 -- label: Distributed DP Tests (2 GPUs) # TBD
2819 +- label: ":amd: (MI300) Distributed DP Basic" # TBD
2820 timeout_in_minutes: 180
2821 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2822 dind: false
@@ -2790,16 +2832,16 @@ steps:
2832 - vllm/v1/engine/
2833 - vllm/v1/worker/
2834 - tests/v1/distributed
2793 - - tests/entrypoints/openai/test_multi_api_servers.py
2835 + - tests/entrypoints/launchers/api_server/test_multi_api_servers.py
2836 - vllm/platforms/rocm.py
2837 commands:
2838 - export NCCL_CUMEM_HOST_ENABLE=0
2839 - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_async_llm_dp.py
2840 - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_eagle_dp.py
2841 - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_external_lb_dp.py
2800 - - DP_SIZE=2 pytest -v -s entrypoints/openai/test_multi_api_servers.py
2842 + - DP_SIZE=2 pytest -v -s entrypoints/launchers/api_server/test_multi_api_servers.py
2843
2802 -- label: Metrics, Tracing (2 GPUs) # TBD
2844 +- label: ":amd: (MI300) Metrics, Tracing" # TBD
2845 timeout_in_minutes: 180
2846 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2847 dind: false
@@ -2810,6 +2852,7 @@ steps:
2852 source_file_dependencies:
2853 - vllm/
2854 - tests/v1/tracing
2855 + - tests/tracing/
2856 commands:
2857 - "pip install \
2858 'opentelemetry-sdk>=1.26.0' \
@@ -2817,8 +2860,9 @@ steps:
2860 'opentelemetry-exporter-otlp>=1.26.0' \
2861 'opentelemetry-semantic-conventions-ai>=0.4.1'"
2862 - pytest -v -s v1/tracing
2863 + - pytest -v -s tracing
2864
2821 -- label: V1 e2e (2 GPUs) # TBD
2865 +- label: ":amd: (MI300) V1 E2E" # TBD
2866 timeout_in_minutes: 180
2867 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2868 dind: false
@@ -2835,7 +2879,7 @@ steps:
2879 v1/e2e/spec_decode/draft_model/test_draft_model.py::test_draft_model_tensor_parallelism
2880 v1/e2e/spec_decode/draft_model/test_draft_model.py::test_draft_model_engine_args_tensor_parallelism
2881
2838 -- label: NixlConnector PD + Spec Decode acceptance (2 GPUs) # TBD
2882 +- label: ":amd: (MI300) NixlConnector PD + Spec Decode acceptance" # TBD
2883 timeout_in_minutes: 180
2884 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2885 dind: false
@@ -2852,7 +2896,7 @@ steps:
2896 - uv pip install --system -r /vllm-workspace/requirements/kv_connectors_rocm.txt
2897 - KV_CACHE_MEMORY_BYTES=8G ATTENTION_BACKEND=TRITON_ATTN bash v1/kv_connector/nixl_integration/config_sweep_spec_decode_test.sh
2898
2855 -- label: CrossLayer KV layout Distributed NixlConnector PD accuracy tests (4 GPUs) # TBD
2899 +- label: ":amd: (MI300) CrossLayer KV layout Distributed NixlConnector PD accuracy" # TBD
2900 timeout_in_minutes: 180
2901 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2902 dind: false
@@ -2868,7 +2912,7 @@ steps:
2912 - uv pip install --system -r /vllm-workspace/requirements/kv_connectors_rocm.txt
2913 - CROSS_LAYERS_BLOCKS=True ATTENTION_BACKEND=TRITON_ATTN bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
2914
2871 -- label: Distributed DP Tests (4 GPUs) # TBD
2915 +- label: ":amd: (MI300) Distributed DP Extended" # TBD
2916 timeout_in_minutes: 180
2917 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2918 dind: false
@@ -2892,7 +2936,7 @@ steps:
2936 - pytest -v -s v1/engine/test_engine_core_client.py::test_kv_cache_events_dp
2937 - pytest -v -s distributed/test_utils.py
2938
2895 -- label: Distributed NixlConnector PD accuracy (4 GPUs) # TBD
2939 +- label: ":amd: (MI300) Distributed NixlConnector PD accuracy" # TBD
2940 timeout_in_minutes: 180
2941 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2942 dind: false
@@ -2908,7 +2952,7 @@ steps:
2952 - uv pip install --system -r /vllm-workspace/requirements/kv_connectors_rocm.txt
2953 - ATTENTION_BACKEND=TRITON_ATTN bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
2954
2911 -- label: DP EP Distributed NixlConnector PD accuracy tests (4 GPUs) # TBD
2955 +- label: ":amd: (MI300) DP EP Distributed NixlConnector PD accuracy" # TBD
2956 timeout_in_minutes: 180
2957 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2958 dind: false
@@ -2924,7 +2968,7 @@ steps:
2968 - uv pip install --system -r /vllm-workspace/requirements/kv_connectors_rocm.txt
2969 - DP_EP=1 ATTENTION_BACKEND=TRITON_ATTN bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
2970
2927 -- label: Hybrid SSM NixlConnector PD accuracy tests (4 GPUs) # TBD
2971 +- label: ":amd: (MI300) Hybrid SSM NixlConnector PD accuracy" # TBD
2972 timeout_in_minutes: 180
2973 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2974 dind: false
@@ -2940,7 +2984,7 @@ steps:
2984 - uv pip install --system -r /vllm-workspace/requirements/kv_connectors_rocm.txt
2985 - HYBRID_SSM=1 ATTENTION_BACKEND=TRITON_ATTN bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
2986
2943 -- label: Hybrid SSM NixlConnector PD prefix cache test (2 GPUs) # TBD
2987 +- label: ":amd: (MI300) Hybrid SSM NixlConnector PD prefix cache" # TBD
2988 timeout_in_minutes: 180
2989 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2990 dind: false
@@ -2958,7 +3002,7 @@ steps:
3002 - uv pip install --system -r /vllm-workspace/requirements/kv_connectors_rocm.txt
3003 - ATTENTION_BACKEND=TRITON_ATTN bash v1/kv_connector/nixl_integration/run_mamba_prefix_cache_test.sh
3004
2961 -- label: MultiConnector (Nixl+Offloading) PD accuracy (2 GPUs) # TBD
3005 +- label: ":amd: (MI300) MultiConnector (Nixl+Offloading) PD accuracy" # TBD
3006 timeout_in_minutes: 180
3007 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
3008 dind: false
@@ -2977,7 +3021,7 @@ steps:
3021 - uv pip install --system -r /vllm-workspace/requirements/kv_connectors_rocm.txt
3022 - ATTENTION_BACKEND=TRITON_ATTN bash v1/kv_connector/nixl_integration/run_multi_connector_accuracy_test.sh
3023
2980 -- label: MultiConnector (Nixl+Offloading) PD edge cases (2 GPUs) # TBD
3024 +- label: ":amd: (MI300) MultiConnector (Nixl+Offloading) PD edge cases" # TBD
3025 timeout_in_minutes: 180
3026 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
3027 dind: false
@@ -2996,7 +3040,7 @@ steps:
3040 - uv pip install --system -r /vllm-workspace/requirements/kv_connectors_rocm.txt
3041 - ATTENTION_BACKEND=TRITON_ATTN bash v1/kv_connector/nixl_integration/run_multi_connector_edge_case_test.sh
3042
2999 -- label: V1 e2e (4xH100-4xMI300) # TBD
3043 +- label: ":amd: (MI300) V1 E2E Hybrid Chunked Prefill" # TBD
3044 timeout_in_minutes: 180
3045 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
3046 dind: false
@@ -3012,7 +3056,7 @@ steps:
3056
3057 #------------------------------------------------------ mi300 · weight_loading -------------------------------------------------------#
3058
3015 -- label: Weight Loading Multiple GPU # TBD
3059 +- label: ":amd: (MI300) Weight Loading Multi-GPU" # TBD
3060 timeout_in_minutes: 180
3061 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
3062 dind: false
@@ -3025,7 +3069,7 @@ steps:
3069 commands:
3070 - bash weight_loading/run_model_weight_loading_test.sh -c weight_loading/models-amd.txt
3071
3028 -- label: Weight Loading Multiple GPU - Large Models # TBD
3072 +- label: ":amd: (MI300) Weight Loading Multi-GPU (Large Models)" # TBD
3073 timeout_in_minutes: 180
3074 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
3075 dind: false
@@ -3047,7 +3091,7 @@ steps:
3091
3092 #-------------------------------------------------------- mi355 · benchmarks ---------------------------------------------------------#
3093
3050 -- label: Attention Benchmarks Smoke Test (B200-MI355) # TBD
3094 +- label: ":amd: (MI355) Attention Benchmark Smoke" # TBD
3095 timeout_in_minutes: 180
3096 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3097 dind: false
@@ -3064,7 +3108,7 @@ steps:
3108
3109 #-------------------------------------------------------- mi355 · distributed --------------------------------------------------------#
3110
3067 -- label: Distributed Tests (2xH100-2xMI355) # TBD
3111 +- label: ":amd: (MI355) Distributed Features" # TBD
3112 timeout_in_minutes: 180
3113 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3114 dind: false
@@ -3109,7 +3153,7 @@ steps:
3153
3154 #-------------------------------------------------------- mi355 · entrypoints --------------------------------------------------------#
3155
3112 -- label: Entrypoints Integration (API Server) # TBD
3156 +- label: ":amd: (MI355) Entrypoints Integration (API Server)" # TBD
3157 timeout_in_minutes: 180
3158 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3159 dind: false
@@ -3127,7 +3171,7 @@ steps:
3171 - PYTHONPATH=/vllm-workspace pytest -v -s entrypoints/serve/dev/rpc
3172 - pytest -v -s entrypoints/scale_out
3173
3130 -- label: Entrypoints Integration (API Server OpenAI - Part 1) # TBD
3174 +- label: ":amd: (MI355) Entrypoints Integration (API Server OpenAI - Part 1)" # TBD
3175 timeout_in_minutes: 180
3176 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3177 dind: false
@@ -3143,7 +3187,7 @@ steps:
3187 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
3188 - pytest -v -s entrypoints/openai --ignore=entrypoints/openai/completion --ignore=entrypoints/openai/chat_completion --ignore=entrypoints/openai/responses --ignore=entrypoints/openai/correctness
3189
3146 -- label: Entrypoints Integration (API Server OpenAI - Part 2) # TBD
3190 +- label: ":amd: (MI355) Entrypoints Integration (API Server OpenAI - Part 2)" # TBD
3191 timeout_in_minutes: 180
3192 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3193 dind: false
@@ -3160,7 +3204,7 @@ steps:
3204 - pytest -v -s entrypoints/openai/chat_completion
3205 - pytest -v -s entrypoints/openai/completion --ignore=entrypoints/openai/completion/test_tensorizer_entrypoint.py
3206
3163 -- label: Entrypoints Integration (API Server Generate) # TBD
3207 +- label: ":amd: (MI355) Entrypoints Integration (API Server Generate)" # TBD
3208 timeout_in_minutes: 180
3209 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3210 dind: false
@@ -3181,7 +3225,7 @@ steps:
3225 - pytest -v -s entrypoints/generate
3226 - pytest -v -s entrypoints/anthropic
3227
3184 -- label: Entrypoints Integration (Speech to Text) # TBD
3228 +- label: ":amd: (MI355) Entrypoints Integration (Speech to Text)" # TBD
3229 timeout_in_minutes: 180
3230 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi355]
3231 dind: false
@@ -3195,7 +3239,7 @@ steps:
3239 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
3240 - pytest -v -s entrypoints/speech_to_text
3241
3198 -- label: Entrypoints Integration (Multimodal)
3242 +- label: ":amd: (MI355) Entrypoints Integration (Multimodal)"
3243 timeout_in_minutes: 180
3244 mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi355]
3245 dind: false
@@ -3209,7 +3253,7 @@ steps:
3253 - export VLLM_WORKER_MULTIPROC_METHOD=spawn
3254 - pytest -v -s entrypoints/multimodal
3255
3212 -- label: Entrypoints Integration (Pooling) # TBD
3256 +- label: ":amd: (MI355) Entrypoints Integration (Pooling)" # TBD
3257 timeout_in_minutes: 180
3258 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3259 dind: false
@@ -3225,7 +3269,7 @@ steps:
3269
3270 #----------------------------------------------------------- mi355 · evals -----------------------------------------------------------#
3271
3228 -- label: GPQA Eval (GPT-OSS) (2xB200-2xMI355) # TBD
3272 +- label: ":amd: (MI355) GPQA Eval (GPT-OSS)" # TBD
3273 timeout_in_minutes: 180
3274 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3275 dind: false
@@ -3248,7 +3292,7 @@ steps:
3292 - uv pip install --system 'gpt-oss[eval]==0.0.5'
3293 - pytest -s -v evals/gpt_oss/test_gpqa_correctness.py --config-list-file=configs/models-gfx950.txt
3294
3251 -- label: LM Eval Qwen3-5 Models (B200-MI355) # TBD
3295 +- label: ":amd: (MI355) LM Eval Qwen3-5 Models" # TBD
3296 timeout_in_minutes: 180
3297 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3298 dind: false
@@ -3271,7 +3315,7 @@ steps:
3315 commands:
3316 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-qwen35-mi355.txt
3317
3274 -- label: LM Eval Small Models (2xB200-2xMI355) # TBD
3318 +- label: ":amd: (MI355) LM Eval Small Models" # TBD
3319 timeout_in_minutes: 180
3320 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3321 dind: false
@@ -3291,7 +3335,7 @@ steps:
3335 commands:
3336 - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-mi3xx-fp8-and-mixed.txt
3337
3294 -- label: Qwen3-30B-A3B-FP8-block Sync EPLB Accuracy (B200-MI355) # TBD
3338 +- label: ":amd: (MI355) Qwen3-30B-A3B-FP8-block Sync EPLB Accuracy" # TBD
3339 timeout_in_minutes: 180
3340 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3341 dind: false
@@ -3312,7 +3356,7 @@ steps:
3356 commands:
3357 - bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_block_ep_eplb.sh 0.8 200 8020 2 1
3358
3315 -- label: LM Eval Large Models (4xH100-4xMI355) # TBD
3359 +- label: ":amd: (MI355) LM Eval Large Models FP8" # TBD
3360 timeout_in_minutes: 180
3361 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3362 dind: false
@@ -3333,7 +3377,7 @@ steps:
3377 - export VLLM_USE_DEEP_GEMM=0
3378 - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large-rocm-fp8.txt --tp-size=4
3379
3336 -- label: LM Eval Large Models (8xB200-8xMI355) # TBD
3380 +- label: ":amd: (MI355) LM Eval Large Models" # TBD
3381 timeout_in_minutes: 180
3382 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3383 dind: false
@@ -3358,7 +3402,7 @@ steps:
3402
3403 #--------------------------------------------------------- mi355 · examples ----------------------------------------------------------#
3404
3361 -- label: Examples # TBD
3405 +- label: ":amd: (MI355) Examples" # TBD
3406 timeout_in_minutes: 180
3407 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3408 dind: false
@@ -3394,7 +3438,7 @@ steps:
3438
3439 #---------------------------------------------------------- mi355 · kernels ----------------------------------------------------------#
3440
3397 -- label: Kernels Core Operation Test %N # TBD
3441 +- label: ":amd: (MI355) Core Operation Kernels Shard %N" # TBD
3442 timeout_in_minutes: 180
3443 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3444 dind: false
@@ -3412,7 +3456,7 @@ steps:
3456 commands:
3457 - pytest -v -s kernels/core --ignore=kernels/core/test_minimax_reduce_rms.py kernels/test_concat_mla_q.py kernels/test_top_k_per_row.py --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
3458
3415 -- label: Kernels (B200-MI355) # TBD
3459 +- label: ":amd: (MI355) Kernels" # TBD
3460 timeout_in_minutes: 180
3461 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3462 dind: false
@@ -3438,7 +3482,7 @@ steps:
3482 - pytest -v -s tests/kernels/attention/test_attention_selector.py
3483 - pytest -v -s tests/kernels/attention/test_rocm_aiter_mla_decode_metadata.py
3484
3441 -- label: Kernels MLA (MI355) # TBD
3485 +- label: ":amd: (MI355) MLA Kernels" # TBD
3486 timeout_in_minutes: 180
3487 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3488 dind: false
@@ -3460,7 +3504,7 @@ steps:
3504 - pytest -v -s tests/kernels/attention/test_rocm_aiter_mla_fp8_support.py
3505 - pytest -v -s tests/kernels/attention/test_rocm_aiter_mla_op_registration.py
3506
3463 -- label: Kernels Attention Test %N # TBD
3507 +- label: ":amd: (MI355) Attention Kernels Shard %N" # TBD
3508 timeout_in_minutes: 180
3509 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3510 dind: false
@@ -3478,7 +3522,7 @@ steps:
3522 commands:
3523 - pytest -v -s kernels/attention --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
3524
3481 -- label: Kernels MoE Test %N # TBD
3525 +- label: ":amd: (MI355) MoE Kernels Shard %N" # TBD
3526 timeout_in_minutes: 180
3527 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3528 dind: false
@@ -3512,7 +3556,7 @@ steps:
3556 --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
3557 - pytest -v -s kernels/moe/test_modular_oai_triton_moe.py --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
3558
3515 -- label: Kernels FusedMoE Layer Test (2xB200-2xMI355) # TBD
3559 +- label: ":amd: (MI355) FusedMoE Layer Kernels" # TBD
3560 timeout_in_minutes: 180
3561 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3562 dind: false
@@ -3538,7 +3582,7 @@ steps:
3582 commands:
3583 - pytest -v -s kernels/moe/test_moe_layer.py
3584
3541 -- label: Kernels Quantization Test %N # TBD
3585 +- label: ":amd: (MI355) Quantization Kernels" # TBD
3586 timeout_in_minutes: 180
3587 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3588 dind: false
@@ -3561,7 +3605,7 @@ steps:
3605 commands:
3606 - pytest -v -s kernels/quantization
3607
3564 -- label: Kernels FP8 MoE Test (2xH100-1xMI355) # TBD
3608 +- label: ":amd: (MI355) FP8 MoE Kernels" # TBD
3609 timeout_in_minutes: 180
3610 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3611 dind: false
@@ -3582,7 +3626,7 @@ steps:
3626 - pytest -v -s kernels/moe/test_triton_moe_no_act_mul.py
3627 - pytest -v -s kernels/moe/test_triton_moe_ptpc_fp8.py
3628
3585 -- label: Kernels FP8 MoE Test (2xH100-2xMI355) # TBD
3629 +- label: ":amd: (MI355) DeepEP FP8 MoE Kernels" # TBD
3630 timeout_in_minutes: 180
3631 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3632 dind: false
@@ -3603,7 +3647,7 @@ steps:
3647
3648 #----------------------------------------------------- mi355 · models / language -----------------------------------------------------#
3649
3606 -- label: Language Models Test (Extended Generation) # TBD
3650 +- label: ":amd: (MI355) Language Models (Extended Generation)" # TBD
3651 timeout_in_minutes: 180
3652 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3653 dind: false
@@ -3617,7 +3661,7 @@ steps:
3661 - CAUSAL_CONV1D_FORCE_BUILD=TRUE uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.6.0'
3662 - pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
3663
3620 -- label: Language Models Test (Extended Pooling) # TBD
3664 +- label: ":amd: (MI355) Language Models (Extended Pooling)" # TBD
3665 timeout_in_minutes: 180
3666 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3667 dind: false
@@ -3630,7 +3674,7 @@ steps:
3674 commands:
3675 - pytest -v -s models/language/pooling -m 'not core_model'
3676
3633 -- label: Language Models Test (PPL) # TBD
3677 +- label: ":amd: (MI355) Language Models (PPL)" # TBD
3678 timeout_in_minutes: 180
3679 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3680 dind: false
@@ -3659,7 +3703,7 @@ steps:
3703 commands:
3704 - pytest -v -s models/language/generation_ppl_test
3705
3662 -- label: Language Models Tests (Standard) # TBD
3706 +- label: ":amd: (MI355) Language Models (Standard)" # TBD
3707 timeout_in_minutes: 180
3708 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3709 dind: false
@@ -3674,7 +3718,7 @@ steps:
3718
3719 #---------------------------------------------------- mi355 · models / multimodal ----------------------------------------------------#
3720
3677 -- label: Multi-Modal Models (Extended Generation 1) # TBD
3721 +- label: ":amd: (MI355) Multimodal Models (Extended Generation 1)" # TBD
3722 timeout_in_minutes: 180
3723 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3724 dind: false
@@ -3689,7 +3733,7 @@ steps:
3733 - pytest -v -s models/multimodal/generation -m 'not core_model' --ignore models/multimodal/generation/test_common.py
3734 - pytest -v -s models/multimodal/test_mapping.py
3735
3692 -- label: Multi-Modal Models (Extended Generation 3) # TBD
3736 +- label: ":amd: (MI355) Multimodal Models (Extended Generation 3)" # TBD
3737 timeout_in_minutes: 180
3738 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3739 dind: false
@@ -3702,7 +3746,7 @@ steps:
3746 commands:
3747 - pytest -v -s models/multimodal/generation/test_common.py -m 'split(group=1) and not core_model'
3748
3705 -- label: Multi-Modal Models (Extended Pooling) # TBD
3749 +- label: ":amd: (MI355) Multimodal Models (Extended Pooling)" # TBD
3750 timeout_in_minutes: 180
3751 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3752 dind: false
@@ -3715,7 +3759,7 @@ steps:
3759 commands:
3760 - pytest -v -s models/multimodal/pooling -m 'not core_model'
3761
3718 -- label: "Multi-Modal Models (Standard) 1: qwen2" # TBD
3762 +- label: ":amd: (MI355) Multimodal Models (Standard) 1: qwen2" # TBD
3763 timeout_in_minutes: 180
3764 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3765 dind: false
@@ -3729,7 +3773,7 @@ steps:
3773 - pytest -v -s models/multimodal/generation/test_common.py -m core_model -k "qwen2"
3774 - pytest -v -s models/multimodal/generation/test_ultravox.py -m core_model
3775
3732 -- label: "Multi-Modal Models (Standard) 4: other + whisper" # TBD
3776 +- label: ":amd: (MI355) Multimodal Models (Standard) 4: other + whisper" # TBD
3777 timeout_in_minutes: 180
3778 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3779 dind: false
@@ -3738,16 +3782,16 @@ steps:
3782 working_dir: "/vllm-workspace/tests"
3783 source_file_dependencies:
3784 - vllm/
3741 - - tests/models/multimodal/generation
3785 + - tests/models/multimodal
3786 commands:
3743 - - pytest -v -s models/multimodal -m core_model --ignore models/multimodal/generation/test_common.py --ignore models/multimodal/generation/test_ultravox.py --ignore models/multimodal/generation/test_qwen2_5_vl.py --ignore models/multimodal/generation/test_qwen2_vl.py --ignore models/multimodal/generation/test_whisper.py --ignore models/multimodal/generation/test_memory_leak.py --ignore models/multimodal/generation/test_vit_cudagraph.py --ignore models/multimodal/processing
3787 + - pytest -v -s models/multimodal -m core_model --ignore models/multimodal/generation/test_common.py --ignore models/multimodal/generation/test_ultravox.py --ignore models/multimodal/generation/test_qwen2_5_vl.py --ignore models/multimodal/generation/test_qwen2_vl.py --ignore models/multimodal/generation/test_whisper.py --ignore models/multimodal/generation/test_mm_prefix_lm.py --ignore models/multimodal/generation/test_memory_leak.py --ignore models/multimodal/generation/test_vit_cudagraph.py --ignore models/multimodal/processing
3788 - pytest -v -s models/multimodal/generation/test_vit_cudagraph.py -m core_model
3789 - pytest -v -s models/multimodal/generation/test_memory_leak.py -m core_model
3790 - cd .. && VLLM_WORKER_MULTIPROC_METHOD=spawn pytest -v -s tests/models/multimodal/generation/test_whisper.py -m core_model
3791
3792 #----------------------------------------------------- mi355 · models / quantized -----------------------------------------------------#
3793
3750 -- label: Quantized Models Test # TBD
3794 +- label: ":amd: (MI355) Quantized Models" # TBD
3795 timeout_in_minutes: 180
3796 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3797 dind: false
@@ -3764,7 +3808,7 @@ steps:
3808
3809 #------------------------------------------------------- mi355 · quantization --------------------------------------------------------#
3810
3767 -- label: Quantization # TBD
3811 +- label: ":amd: (MI355) Quantization" # TBD
3812 timeout_in_minutes: 180
3813 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3814 dind: false
@@ -3810,7 +3854,7 @@ steps:
3854
3855 #------------------------------------------------------------ mi355 · v1 -------------------------------------------------------------#
3856
3813 -- label: V1 attention (B200-MI355) %N # TBD
3857 +- label: ":amd: (MI355) V1 Attention Shard %N" # TBD
3858 timeout_in_minutes: 180
3859 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3860 dind: false
@@ -3828,7 +3872,7 @@ steps:
3872 commands:
3873 - pytest -v -s v1/attention --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
3874
3831 -- label: V1 Core + KV + Metrics # TBD
3875 +- label: ":amd: (MI355) V1 Core + KV + Metrics" # TBD
3876 timeout_in_minutes: 180
3877 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3878 dind: false
@@ -3857,7 +3901,7 @@ steps:
3901 - pip install -U git+https://github.com/vllm-project/lm-evaluation-harness.git@streaming-api
3902 - pytest -v -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine
3903
3860 -- label: V1 Sample + Logits # TBD
3904 +- label: ":amd: (MI355) V1 Sample + Logits" # TBD
3905 timeout_in_minutes: 180
3906 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3907 dind: false
@@ -3878,7 +3922,7 @@ steps:
3922 - pytest -v -s v1/test_request.py
3923 - pytest -v -s v1/test_outputs.py
3924
3881 -- label: V1 Spec Decode # TBD
3925 +- label: ":amd: (MI355) V1 Spec Decode" # TBD
3926 timeout_in_minutes: 180
3927 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3928 dind: false
@@ -3892,7 +3936,7 @@ steps:
3936
3937 #------------------------------------------------------ mi355 · weight_loading -------------------------------------------------------#
3938
3895 -- label: Weight Loading Multiple GPU # TBD
3939 +- label: ":amd: (MI355) Weight Loading Multi-GPU" # TBD
3940 timeout_in_minutes: 180
3941 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3942 dind: false
@@ -3905,7 +3949,7 @@ steps:
3949 commands:
3950 - bash weight_loading/run_model_weight_loading_test.sh -c weight_loading/models-amd.txt
3951
3908 -- label: Weight Loading Multiple GPU - Large Models # TBD
3952 +- label: ":amd: (MI355) Weight Loading Multi-GPU (Large Models)" # TBD
3953 timeout_in_minutes: 180
3954 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3955 dind: false
@@ -3921,7 +3965,7 @@ steps:
3965
3966 #----------------------------------------------------------- mi355 · misc ------------------------------------------------------------#
3967
3924 -- label: Regression # TBD
3968 +- label: ":amd: (MI355) Regression" # TBD
3969 timeout_in_minutes: 180
3970 mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3971 dind: false
.buildkite/test_areas/distributed.yaml
+3 -3
@@ -31,14 +31,14 @@ steps:
31 - vllm/v1/engine/
32 - vllm/v1/worker/
33 - tests/v1/distributed
34 - - tests/entrypoints/openai/test_multi_api_servers.py
34 + - tests/entrypoints/launchers/api_server/test_multi_api_servers.py
35 commands:
36 # https://github.com/NVIDIA/nccl/issues/1838
37 - export NCCL_CUMEM_HOST_ENABLE=0
38 - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_async_llm_dp.py
39 - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_eagle_dp.py
40 - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_external_lb_dp.py
41 - - DP_SIZE=2 pytest -v -s entrypoints/openai/test_multi_api_servers.py
41 + - DP_SIZE=2 pytest -v -s entrypoints/launchers/api_server/test_multi_api_servers.py
42 mirror:
43 amd:
44 label: ":amd: (MI300) Distributed DP Basic"
@@ -56,7 +56,7 @@ steps:
56 - vllm/v1/engine/
57 - vllm/v1/worker/
58 - tests/v1/distributed
59 - - tests/entrypoints/openai/test_multi_api_servers.py
59 + - tests/entrypoints/launchers/api_server/test_multi_api_servers.py
60 - vllm/platforms/rocm.py
61
62 - label: ":nvidia: (L4) Distributed Compile + RPC"
.buildkite/test_areas/models_multimodal.yaml
+1 -1
@@ -80,7 +80,7 @@ steps:
80 label: ":amd: (MI300) Multimodal Models (Standard) 4: other + whisper"
81 dind: false
82 device: mi300_1
83 - timeout_in_minutes: 50
83 + timeout_in_minutes: 70
84 depends_on:
85 - image-build-amd
86
tests/models/test_vision.py
+4 -4
@@ -102,7 +102,7 @@ def run_dp_sharded_vision_model_vs_direct(
102 # Set random seed for reproducibility
103 set_random_seed(0)
104
105 - device = f"{current_platform.device_name}:{local_rank}"
105 + device = f"{current_platform.device_type}:{local_rank}"
106 torch.accelerator.set_device_index(device)
107 torch.set_default_device(device)
108
@@ -288,7 +288,7 @@ def run_dp_sharded_mrope_vision_model_vs_direct(
288 """
289 # Set random seed for reproducibility
290 set_random_seed(0)
291 - device = f"{current_platform.device_name}:{local_rank}"
291 + device = f"{current_platform.device_type}:{local_rank}"
292 torch.accelerator.set_device_index(device)
293 torch.set_default_device(device)
294
@@ -365,7 +365,7 @@ def run_dp_sharded_mrope_vision_model_empty_input_worker(
365 ):
366 """Test run_dp_sharded_mrope_vision_model with empty input."""
367 # Set up distributed environment
368 - device = f"{current_platform.device_name}:{local_rank}"
368 + device = f"{current_platform.device_type}:{local_rank}"
369 torch.accelerator.set_device_index(device)
370 torch.set_default_device(device)
371
@@ -414,7 +414,7 @@ def run_dp_sharded_mrope_vision_model_uneven_load_worker(
414 """Test run_dp_sharded_mrope_vision_model with uneven load distribution."""
415 # Set up distributed environment
416 set_random_seed(123)
417 - device = f"{current_platform.device_name}:{local_rank}"
417 + device = f"{current_platform.device_type}:{local_rank}"
418 torch.accelerator.set_device_index(device)
419 torch.set_default_device(device)
420