[CI] Mark L4 GPU test steps with device: l4 for EKS migration (#54326)
Signed-off-by: khluu <khluu000@gmail.com>
Kevin H. Luu committed
Aug 30, 2026 at 04:40 UTC
5e71a11eb2b595ede15ecc39c7dddca38e03deb5
13 files changed
+35
-2
.buildkite/test_areas/crcr_report.yaml
+2
-1
@@ -1,6 +1,6 @@
1
group: CRCR Report
2
depends_on:
3
- - image-build
3
+ - image-build-cpu
4
steps:
5
- label: ":satellite: Report nightly results to CRCR"
6
key: crcr-report
@@ -8,5 +8,6 @@ steps:
8
soft_fail: true
9
allow_dependency_failure: true
10
timeout_in_minutes: 15
11
+ device: cpu-small
12
commands:
13
- bash /vllm-workspace/.buildkite/scripts/crcr-report.sh
.buildkite/test_areas/disaggregated.yaml
+10
@@ -6,6 +6,7 @@ steps:
6
key: distributed-nixlconnector-pd-accuracy-4-gpus
7
timeout_in_minutes: 55
8
working_dir: "/vllm-workspace/tests"
9
+ device: l4
10
num_devices: 4
11
source_file_dependencies:
12
- vllm/distributed/kv_transfer/kv_connector/v1/nixl/
@@ -33,6 +34,7 @@ steps:
34
key: distributed-flashinfer-nixlconnector-pd-accuracy-4-gpus
35
timeout_in_minutes: 55
36
working_dir: "/vllm-workspace/tests"
37
+ device: l4
38
num_devices: 4
39
source_file_dependencies:
40
- vllm/distributed/kv_transfer/kv_connector/v1/nixl/
@@ -45,6 +47,7 @@ steps:
47
key: push-nixlconnector-pp-prefill-pd-accuracy-4-gpus
48
timeout_in_minutes: 30
49
working_dir: "/vllm-workspace/tests"
50
+ device: l4
51
num_devices: 4
52
source_file_dependencies:
53
- vllm/distributed/kv_transfer/kv_connector/v1/nixl/
@@ -58,6 +61,7 @@ steps:
61
key: dp-ep-distributed-nixlconnector-pd-accuracy-tests-4-gpus
62
timeout_in_minutes: 30
63
working_dir: "/vllm-workspace/tests"
64
+ device: l4
65
num_devices: 4
66
source_file_dependencies:
67
- vllm/distributed/kv_transfer/kv_connector/v1/nixl/
@@ -85,6 +89,7 @@ steps:
89
key: crosslayer-kv-layout-distributed-nixlconnector-pd-accuracy-tests-4-gpus
90
timeout_in_minutes: 55
91
working_dir: "/vllm-workspace/tests"
92
+ device: l4
93
num_devices: 4
94
source_file_dependencies:
95
- vllm/distributed/kv_transfer/kv_connector/v1/nixl/
@@ -112,6 +117,7 @@ steps:
117
key: hybrid-ssm-nixlconnector-pd-accuracy-tests-4-gpus
118
timeout_in_minutes: 60
119
working_dir: "/vllm-workspace/tests"
120
+ device: l4
121
num_devices: 4
122
source_file_dependencies:
123
- vllm/distributed/kv_transfer/kv_connector/v1/nixl/
@@ -139,6 +145,7 @@ steps:
145
key: nixlconnector-pd-edge-cases-2-gpus
146
timeout_in_minutes: 40
147
working_dir: "/vllm-workspace/tests"
148
+ device: l4
149
num_devices: 2
150
source_file_dependencies:
151
- vllm/distributed/kv_transfer/kv_connector/v1/nixl/
@@ -155,6 +162,7 @@ steps:
162
key: hybrid-ssm-nixlconnector-pd-prefix-cache-2-gpus
163
timeout_in_minutes: 25
164
working_dir: "/vllm-workspace/tests"
165
+ device: l4
166
num_devices: 2
167
source_file_dependencies:
168
- vllm/distributed/kv_transfer/kv_connector/v1/nixl/
@@ -169,6 +177,7 @@ steps:
177
key: multiconnector-nixl-offloading-pd-accuracy-2-gpus
178
timeout_in_minutes: 40
179
working_dir: "/vllm-workspace/tests"
180
+ device: l4
181
num_devices: 2
182
source_file_dependencies:
183
- vllm/distributed/kv_transfer/kv_connector/v1/nixl/
@@ -214,6 +223,7 @@ steps:
223
key: multiconnector-nixl-offloading-pd-edge-cases-2-gpus
224
timeout_in_minutes: 25
225
working_dir: "/vllm-workspace/tests"
226
+ device: l4
227
num_devices: 2
228
source_file_dependencies:
229
- vllm/distributed/kv_transfer/kv_connector/v1/nixl/
.buildkite/test_areas/distributed.yaml
+9
@@ -6,6 +6,7 @@ steps:
6
key: distributed-comm-ops
7
timeout_in_minutes: 25
8
working_dir: "/vllm-workspace/tests"
9
+ device: l4
10
num_devices: 2
11
source_file_dependencies:
12
- vllm/distributed
@@ -36,6 +37,7 @@ steps:
37
key: distributed-dp-tests-2-gpus
38
timeout_in_minutes: 35
39
working_dir: "/vllm-workspace/tests"
40
+ device: l4
41
num_devices: 2
42
source_file_dependencies:
43
- vllm/distributed/
@@ -78,6 +80,7 @@ steps:
80
key: distributed-compile-rpc-tests-2-gpus
81
timeout_in_minutes: 65
82
working_dir: "/vllm-workspace/tests"
83
+ device: l4
84
num_devices: 2
85
source_file_dependencies:
86
- vllm/compilation/
@@ -100,6 +103,7 @@ steps:
103
key: distributed-torchrun-shutdown-tests-2-gpus
104
timeout_in_minutes: 30
105
working_dir: "/vllm-workspace/tests"
106
+ device: l4
107
num_devices: 2
108
source_file_dependencies:
109
- vllm/distributed/
@@ -124,6 +128,7 @@ steps:
128
key: distributed-torchrun-examples-4-gpus
129
timeout_in_minutes: 45
130
working_dir: "/vllm-workspace"
131
+ device: l4
132
num_devices: 4
133
source_file_dependencies:
134
- vllm/distributed/
@@ -166,6 +171,7 @@ steps:
171
key: distributed-dp-tests-4-gpus
172
timeout_in_minutes: 45
173
working_dir: "/vllm-workspace/tests"
174
+ device: l4
175
num_devices: 4
176
source_file_dependencies:
177
- vllm/distributed/
@@ -189,6 +195,7 @@ steps:
195
key: distributed-compile-comm-4-gpus
196
timeout_in_minutes: 70
197
working_dir: "/vllm-workspace/tests"
198
+ device: l4
199
num_devices: 4
200
source_file_dependencies:
201
- vllm/distributed/
@@ -307,6 +314,7 @@ steps:
314
key: pipeline-context-parallelism-4-gpus
315
timeout_in_minutes: 55
316
working_dir: "/vllm-workspace/tests"
317
+ device: l4
318
num_devices: 4
319
source_file_dependencies:
320
- vllm/distributed/
@@ -323,6 +331,7 @@ steps:
331
key: rayexecutorv2-4-gpus
332
timeout_in_minutes: 45
333
working_dir: "/vllm-workspace/tests"
334
+ device: l4
335
num_devices: 4
336
source_file_dependencies:
337
- vllm/v1/executor/ray_executor_v2.py
.buildkite/test_areas/engine.yaml
+1
@@ -97,6 +97,7 @@ steps:
97
key: v1-e2e-2-gpus
98
timeout_in_minutes: 25 # TODO: Fix timeout after we have more confidence in the test stability
99
optional: true
100
+ device: l4
101
num_devices: 2
102
source_file_dependencies:
103
- vllm/compilation/
.buildkite/test_areas/expert_parallelism.yaml
+1
@@ -32,6 +32,7 @@ steps:
32
key: eplb-execution
33
timeout_in_minutes: 25
34
working_dir: "/vllm-workspace/tests"
35
+ device: l4
36
num_devices: 4
37
source_file_dependencies:
38
- vllm/distributed/eplb
.buildkite/test_areas/lora.yaml
+1
@@ -30,6 +30,7 @@ steps:
30
- label: ":nvidia: (L4) LoRA TP (Distributed)"
31
key: lora-tp-distributed
32
timeout_in_minutes: 60
33
+ device: l4
34
num_devices: 4
35
source_file_dependencies:
36
- vllm/lora
.buildkite/test_areas/misc.yaml
+2
@@ -150,6 +150,7 @@ steps:
150
- label: ":nvidia: (L4) Extract Hidden States Integration"
151
key: extract-hidden-states-integration-2-gpus
152
timeout_in_minutes: 20
153
+ device: l4
154
num_devices: 2
155
working_dir: "/vllm-workspace"
156
source_file_dependencies:
@@ -234,6 +235,7 @@ steps:
235
- label: ":nvidia: (L4) Metrics, Tracing"
236
key: metrics-tracing-2-gpus
237
timeout_in_minutes: 25
238
+ device: l4
239
num_devices: 2
240
source_file_dependencies:
241
- vllm/config/
.buildkite/test_areas/model_runner_v2.yaml
+2
@@ -65,6 +65,7 @@ steps:
65
key: model-runner-v2-distributed-2-gpus
66
timeout_in_minutes: 40
67
working_dir: "/vllm-workspace/tests"
68
+ device: l4
69
num_devices: 2
70
source_file_dependencies:
71
- vllm/v1/worker/gpu/
@@ -112,6 +113,7 @@ steps:
113
key: model-runner-v2-pipeline-parallelism-4-gpus
114
timeout_in_minutes: 50
115
working_dir: "/vllm-workspace/tests"
116
+ device: l4
117
num_devices: 4
118
source_file_dependencies:
119
- vllm/v1/worker/gpu/
.buildkite/test_areas/models_distributed.yaml
+1
@@ -6,6 +6,7 @@ steps:
6
key: distributed-model-tests-2-gpus
7
timeout_in_minutes: 60
8
working_dir: "/vllm-workspace/tests"
9
+ device: l4
10
num_devices: 2
11
source_file_dependencies:
12
- vllm/model_executor/model_loader/sharded_state_loader.py
.buildkite/test_areas/plugins.yaml
+2
@@ -6,6 +6,7 @@ steps:
6
key: plugin-tests-2-gpus
7
timeout_in_minutes: 35
8
working_dir: "/vllm-workspace/tests"
9
+ device: l4
10
num_devices: 2
11
source_file_dependencies:
12
- vllm/plugins/
@@ -96,6 +97,7 @@ steps:
97
key: bitsandbytes-plugin-2-gpus
98
timeout_in_minutes: 15
99
working_dir: "/vllm-workspace/tests"
100
+ device: l4
101
num_devices: 2
102
soft_fail: true
103
optional: true
.buildkite/test_areas/rust_frontend.yaml
+1
@@ -140,6 +140,7 @@ steps:
140
141
- label: ":nvidia: (L4) Rust Frontend Distributed"
142
timeout_in_minutes: 25
143
+ device: l4
144
num_devices: 4
145
working_dir: "/vllm-workspace/tests"
146
source_file_dependencies:
.buildkite/test_areas/torch_abi.yaml
+2
-1
@@ -1,10 +1,11 @@
1
group: Torch ABI
2
depends_on:
3
- - image-build
3
+ - image-build-cpu
4
steps:
5
- label: ":nvidia: (L4) Torch Stable ABI Audit"
6
key: torch-stable-abi-audit
7
timeout_in_minutes: 5
8
+ device: cpu-small
9
source_file_dependencies:
10
- .buildkite/check-torch-abi.py
11
- csrc/
.buildkite/test_areas/weight_loading.yaml
+1
@@ -6,6 +6,7 @@ steps:
6
key: weight-loading-multiple-gpu
7
timeout_in_minutes: 50
8
working_dir: "/vllm-workspace/tests"
9
+ device: l4
10
num_devices: 2
11
optional: true
12
source_file_dependencies: