| 1 | group: Attention |
| 2 | depends_on: |
| 3 | - image-build |
| 4 | steps: |
| 5 | - label: ":nvidia: (H100) V1 Attention Shard %N" |
| 6 | key: v1-attention-h100-mi300 |
| 7 | timeout_in_minutes: 85 |
| 8 | device: h100 |
| 9 | source_file_dependencies: |
| 10 | - vllm/config/attention.py |
| 11 | - vllm/model_executor/layers/attention |
| 12 | - vllm/v1/attention |
| 13 | - tests/v1/attention |
| 14 | commands: |
| 15 | - pytest -v -s v1/attention --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT |
| 16 | parallelism: 2 |
| 17 | mirror: |
| 18 | amd: |
| 19 | label: ":amd: (MI300) V1 Attention Shard %N" |
| 20 | dind: false |
| 21 | device: mi300_1 |
| 22 | timeout_in_minutes: 125 |
| 23 | depends_on: |
| 24 | - image-build-amd |
| 25 | source_file_dependencies: |
| 26 | - vllm/config/attention.py |
| 27 | - vllm/model_executor/layers/attention |
| 28 | - vllm/v1/attention |
| 29 | - tests/v1/attention |
| 30 | - vllm/_aiter_ops.py |
| 31 | - vllm/envs.py |
| 32 | - vllm/platforms/rocm.py |
| 33 | |
| 34 | - label: ":nvidia: (B200) V1 Attention Shard %N" |
| 35 | key: v1-attention-b200 |
| 36 | timeout_in_minutes: 80 |
| 37 | device: b200-k8s |
| 38 | source_file_dependencies: |
| 39 | - vllm/config/attention.py |
| 40 | - vllm/model_executor/layers/attention |
| 41 | - vllm/v1/attention |
| 42 | - tests/v1/attention |
| 43 | commands: |
| 44 | - pytest -v -s v1/attention --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT |
| 45 | parallelism: 2 |
| 46 | mirror: |
| 47 | amd: |
| 48 | label: ":amd: (MI355 DPX) V1 Attention Shard %N" |
| 49 | dind: false |
| 50 | device: mi355_dpx |
| 51 | timeout_in_minutes: 125 |
| 52 | depends_on: |
| 53 | - image-build-amd |
| 54 | source_file_dependencies: |
| 55 | - vllm/config/attention.py |
| 56 | - vllm/model_executor/layers/attention |
| 57 | - vllm/v1/attention |
| 58 | - tests/v1/attention |
| 59 | - vllm/_aiter_ops.py |
| 60 | - vllm/envs.py |
| 61 | - vllm/platforms/rocm.py |
| 62 | commands: |
| 63 | - pytest -v -s v1/attention --ignore=v1/attention/test_mla_backends.py --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT |
| 64 | - VLLM_ROCM_USE_AITER=1 pytest -v -s v1/attention/test_mla_backends.py --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT |