[Hardware][AMD][CI] Fix e2e core test group (#46024)
Signed-off-by: Matthew Wong <Matthew.Wong2@amd.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
Matt committed
Jun 20, 2026 at 02:04 UTC
dced2907693e3d6bf9eb7168d0a8fecf1cd22dca
3 files changed
+18
-14
.buildkite/test-amd.yaml
+1
-14
@@ -647,7 +647,7 @@ steps:
647
- pytest -v -s v1/cudagraph/test_cudagraph_mode.py
648
649
- label: e2e Core (1 GPU) # TBD
650
- timeout_in_minutes: 180
650
+ timeout_in_minutes: 35
651
mirror_hardwares: [amdexperimental, amdproduction, amdgfx90anightly, amdmi250]
652
agent_pool: mi250_1
653
optional: true
@@ -2075,19 +2075,6 @@ steps:
2075
- export VLLM_ALLOW_INSECURE_SERIALIZATION=1
2076
- pytest -v -s v1/spec_decode/test_acceptance_length.py -m slow_test
2077
2078
-- label: e2e Core (1 GPU) # TBD
2079
- timeout_in_minutes: 180
2080
- mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
2081
- agent_pool: mi300_1
2082
- optional: true
2083
- working_dir: "/vllm-workspace/tests"
2084
- source_file_dependencies:
2085
- - vllm/v1/
2086
- - tests/v1/e2e/
2087
- - vllm/platforms/rocm.py
2088
- commands:
2089
- - pytest -v -s v1/e2e/general --ignore v1/e2e/general/test_async_scheduling.py
2090
-
2078
- label: e2e Scheduling (1 GPU) # TBD
2079
timeout_in_minutes: 180
2080
mirror_hardwares: [amdexperimental, amdproduction, amdgfx942nightly, amdmi300]
.buildkite/test_areas/engine.yaml
+10
@@ -74,6 +74,16 @@ steps:
74
- tests/v1/e2e/general/
75
commands:
76
- pytest -v -s v1/e2e/general --ignore v1/e2e/general/test_async_scheduling.py
77
+ mirror:
78
+ amd:
79
+ device: mi250_1
80
+ timeout_in_minutes: 35
81
+ depends_on:
82
+ - image-build-amd
83
+ source_file_dependencies:
84
+ - vllm/v1/
85
+ - tests/v1/e2e/general/
86
+ - vllm/platforms/rocm.py
87
88
- label: V1 e2e (2 GPUs)
89
key: v1-e2e-2-gpus
tests/v1/e2e/general/test_cascade_attention.py
+7
@@ -4,9 +4,16 @@
4
import pytest
5
6
from vllm import LLM, SamplingParams
7
+from vllm.platforms import current_platform
8
9
from ....utils import create_new_process_for_each_test
10
11
+if current_platform.is_rocm():
12
+ pytest.skip(
13
+ "Cascade attention backends FLASH_ATTN and FLASHINFER are notsupported on ROCm",
14
+ allow_module_level=True,
15
+ )
16
+
17
18
@create_new_process_for_each_test()
19
@pytest.mark.parametrize("attn_backend", ["FLASH_ATTN", "FLASHINFER"])