A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 5, 06:00 UTC
[ROCm][CI] Removed stale tests and extended acceptance test (#39651)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Andreas Karatzas committed
Apr 12, 2026 at 21:40 UTC
4e4ad41d11f3baba8e84e15560c01fc7e52733ab
1 file changed
+1
-67
.buildkite/test-amd.yaml
+1
-67
@@ -532,28 +532,6 @@ steps:
532
- pytest -v -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine
533
534
535
-- label: V1 Speculative Decoding (slow) # TBD
536
- timeout_in_minutes: 180
537
- mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
538
- agent_pool: mi250_1
539
- working_dir: "/vllm-workspace/tests"
540
- source_file_dependencies:
541
- - vllm/v1/spec_decode/
542
- - vllm/model_executor/models/
543
- - vllm/v1/attention/
544
- - vllm/model_executor/layers/
545
- - tests/v1/spec_decode/
546
- - vllm/platforms/rocm.py
547
- commands:
548
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_eagle.py
549
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_extract_hidden_states.py
550
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_max_len.py
551
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_mtp.py
552
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_ngram.py
553
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_speculators_eagle3.py
554
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_tree_attention.py
555
-
556
-
535
- label: V1 attention (H100-MI250) # TBD
536
timeout_in_minutes: 180
537
mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
@@ -1879,28 +1857,6 @@ steps:
1857
- pytest -v -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine
1858
1859
1882
-- label: V1 Speculative Decoding (slow) # TBD
1883
- timeout_in_minutes: 180
1884
- mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi325]
1885
- agent_pool: mi325_1
1886
- optional: true
1887
- working_dir: "/vllm-workspace/tests"
1888
- source_file_dependencies:
1889
- - vllm/v1/spec_decode/
1890
- - vllm/model_executor/models/
1891
- - vllm/v1/attention/
1892
- - vllm/model_executor/layers/
1893
- - tests/v1/spec_decode/
1894
- - vllm/platforms/rocm.py
1895
- commands:
1896
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_eagle.py
1897
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_extract_hidden_states.py
1898
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_max_len.py
1899
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_mtp.py
1900
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_ngram.py
1901
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_speculators_eagle3.py
1902
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_tree_attention.py
1903
-
1860
1861
- label: Acceptance Length Test (Large Models) # TBD
1862
timeout_in_minutes: 180
@@ -1915,7 +1871,7 @@ steps:
1871
- vllm/platforms/rocm.py
1872
commands:
1873
- export VLLM_ALLOW_INSECURE_SERIALIZATION=1
1918
- - pytest -v -s v1/spec_decode/test_acceptance_length.py -m slow_test
1874
+ - pytest -v -s v1/spec_decode/test_acceptance_length.py
1875
1876
1877
- label: V1 attention (H100-MI325) # 14.5m
@@ -3188,28 +3144,6 @@ steps:
3144
- pytest -v -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine
3145
3146
3191
-- label: V1 Speculative Decoding (slow) # TBD
3192
- timeout_in_minutes: 180
3193
- mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]
3194
- agent_pool: mi355_1
3195
- working_dir: "/vllm-workspace/tests"
3196
- source_file_dependencies:
3197
- - vllm/v1/spec_decode/
3198
- - vllm/model_executor/models/
3199
- - vllm/v1/attention/
3200
- - vllm/model_executor/layers/
3201
- - tests/v1/spec_decode/
3202
- - vllm/platforms/rocm.py
3203
- commands:
3204
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_eagle.py
3205
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_extract_hidden_states.py
3206
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_max_len.py
3207
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_mtp.py
3208
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_ngram.py
3209
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_speculators_eagle3.py
3210
- - pytest -v -s -m 'slow_test' v1/spec_decode/test_tree_attention.py
3211
-
3212
-
3147
- label: V1 attention (B200-MI355) # TBD
3148
timeout_in_minutes: 180
3149
mirror_hardwares: [amdexperimental, amdproduction, amdgfx950nightly, amdmi355]