[ROCm][CI] Move LM Eval Large Models (8 GPUs) to mi300 pool (#47094)
Signed-off-by: pei.zhang <pei.zhang@amd.com> Co-authored-by: Andreas Karatzas <akaratza@amd.com> Co-authored-by: Claude <noreply@anthropic.com>
peizhang56 committed
Jun 29, 2026 at 18:59 UTC
f2b5fabb23a62c03dfb809bcd4b678ea2e87ce5a
1 file changed
+21
-21
.buildkite/test-amd.yaml
+21
-21
@@ -1198,6 +1198,27 @@ steps:
1198
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
1199
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-mi3xx.txt
1200
1201
+- label: ROCm LM Eval Large Models (8 GPUs) # TBD
1202
+ timeout_in_minutes: 180
1203
+ mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
1204
+ agent_pool: mi300_8
1205
+ optional: true
1206
+ num_gpus: 8
1207
+ working_dir: "/vllm-workspace/.buildkite/lm-eval-harness"
1208
+ source_file_dependencies:
1209
+ - vllm/model_executor/models/
1210
+ - vllm/model_executor/model_loader/
1211
+ - vllm/model_executor/layers/quantization/
1212
+ - vllm/v1/attention/backends/
1213
+ - vllm/v1/attention/selector.py
1214
+ - vllm/model_executor/layers/layernorm.py
1215
+ - csrc/
1216
+ - vllm/_aiter_ops.py
1217
+ - vllm/platforms/rocm.py
1218
+ commands:
1219
+ - export VLLM_WORKER_MULTIPROC_METHOD=spawn
1220
+ - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large-rocm.txt --tp-size=8
1221
+
1222
#--------------------------------------------------------- mi300 · examples ----------------------------------------------------------#
1223
1224
- label: Examples # TBD
@@ -2392,27 +2413,6 @@ steps:
2413
- export VLLM_USE_DEEP_GEMM=0
2414
- pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large-rocm-fp8.txt --tp-size=4
2415
2395
-- label: ROCm LM Eval Large Models (8 GPUs) # TBD
2396
- timeout_in_minutes: 180
2397
- mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi325]
2398
- agent_pool: mi325_8
2399
- optional: true
2400
- num_gpus: 8
2401
- working_dir: "/vllm-workspace/.buildkite/lm-eval-harness"
2402
- source_file_dependencies:
2403
- - vllm/model_executor/models/
2404
- - vllm/model_executor/model_loader/
2405
- - vllm/model_executor/layers/quantization/
2406
- - vllm/v1/attention/backends/
2407
- - vllm/v1/attention/selector.py
2408
- - vllm/model_executor/layers/layernorm.py
2409
- - csrc/
2410
- - vllm/_aiter_ops.py
2411
- - vllm/platforms/rocm.py
2412
- commands:
2413
- - export VLLM_WORKER_MULTIPROC_METHOD=spawn
2414
- - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large-rocm.txt --tp-size=8
2415
-
2416
#----------------------------------------------------- mi325 · models / language -----------------------------------------------------#
2417
2418
- label: Language Models Test (Extended Generation) # TBD