A high-throughput and memory-efficient inference and serving engine for LLMs
[ROCm][CI] Moving MI300 tests to MI325 until cluster is stabilized (#45170)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Andreas Karatzas committed
Jun 10, 2026 at 15:26 UTC
16282a9c4ee754bedd67f72b01ac76ae8568dbdd
7 files changed
+17
-17
.buildkite/test_areas/attention.yaml
+1
-1
@@ -15,7 +15,7 @@ steps:
15
- pytest -v -s v1/attention
16
mirror:
17
amd:
18
- device: mi300_1
18
+ device: mi325_1
19
timeout_in_minutes: 70
20
depends_on:
21
- image-build-amd
.buildkite/test_areas/engine.yaml
+2
-2
@@ -28,7 +28,7 @@ steps:
28
- pytest -v -s engine test_sequence.py test_config.py test_logger.py test_vllm_port.py test_jit_monitor.py
29
mirror:
30
amd:
31
- device: mi300_1
31
+ device: mi325_1
32
timeout_in_minutes: 60
33
depends_on:
34
- image-build-amd
@@ -44,7 +44,7 @@ steps:
44
- pytest -v -s v1/engine --ignore v1/engine/test_preprocess_error_handling.py
45
mirror:
46
amd:
47
- device: mi300_1
47
+ device: mi325_1
48
timeout_in_minutes: 40
49
depends_on:
50
- image-build-amd
.buildkite/test_areas/entrypoints.yaml
+5
-5
@@ -46,7 +46,7 @@ steps:
46
- PYTHONPATH=/vllm-workspace pytest -v -s entrypoints/serve/dev/rpc
47
mirror:
48
amd:
49
- device: mi300_1
49
+ device: mi325_1
50
depends_on:
51
- image-build-amd
52
@@ -63,7 +63,7 @@ steps:
63
- pytest -v -s entrypoints/openai --ignore=entrypoints/openai/completion --ignore=entrypoints/openai/chat_completion --ignore=entrypoints/openai/responses --ignore=entrypoints/openai/correctness
64
mirror:
65
amd:
66
- device: mi300_1
66
+ device: mi325_1
67
timeout_in_minutes: 80
68
depends_on:
69
- image-build-amd
@@ -82,7 +82,7 @@ steps:
82
- pytest -v -s entrypoints/openai/completion --ignore=entrypoints/openai/completion/test_tensorizer_entrypoint.py
83
mirror:
84
amd:
85
- device: mi300_1
85
+ device: mi325_1
86
timeout_in_minutes: 80
87
depends_on:
88
- image-build-amd
@@ -104,7 +104,7 @@ steps:
104
- pytest -v -s entrypoints/anthropic
105
mirror:
106
amd:
107
- device: mi300_1
107
+ device: mi325_1
108
timeout_in_minutes: 60
109
depends_on:
110
- image-build-amd
@@ -165,7 +165,7 @@ steps:
165
- pytest -s entrypoints/openai/correctness/
166
mirror:
167
amd:
168
- device: mi300_1
168
+ device: mi325_1
169
depends_on:
170
- image-build-amd
171
source_file_dependencies:
.buildkite/test_areas/kernels.yaml
+1
-1
@@ -87,7 +87,7 @@ steps:
87
parallelism: 2
88
mirror:
89
amd:
90
- device: mi300_1
90
+ device: mi325_1
91
source_file_dependencies:
92
- csrc/quantization/
93
- vllm/model_executor/layers/quantization
.buildkite/test_areas/lm_eval.yaml
+1
-1
@@ -14,7 +14,7 @@ steps:
14
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-small.txt
15
mirror:
16
amd:
17
- device: mi300_1
17
+ device: mi325_1
18
timeout_in_minutes: 55
19
depends_on:
20
- image-build-amd
.buildkite/test_areas/models_multimodal.yaml
+4
-4
@@ -15,7 +15,7 @@ steps:
15
- pytest -v -s models/multimodal/generation/test_ultravox.py -m core_model
16
mirror:
17
amd:
18
- device: mi300_1
18
+ device: mi325_1
19
depends_on:
20
- image-build-amd
21
@@ -33,7 +33,7 @@ steps:
33
- pytest -v -s models/multimodal/generation/test_vit_cudagraph.py -m core_model
34
mirror:
35
amd:
36
- device: mi300_1
36
+ device: mi325_1
37
depends_on:
38
- image-build-amd
39
@@ -50,7 +50,7 @@ steps:
50
- pytest -v -s models/multimodal/generation/test_qwen2_vl.py -m core_model
51
mirror:
52
amd:
53
- device: mi300_1
53
+ device: mi325_1
54
depends_on:
55
- image-build-amd
56
@@ -155,7 +155,7 @@ steps:
155
- pytest -v -s models/multimodal/pooling -m 'not core_model'
156
mirror:
157
amd:
158
- device: mi300_1
158
+ device: mi325_1
159
timeout_in_minutes: 60
160
depends_on:
161
- image-build-amd
.buildkite/test_areas/spec_decode.yaml
+3
-3
@@ -39,7 +39,7 @@ steps:
39
- pytest -v -s v1/e2e/spec_decode -k "speculators or mtp_correctness"
40
mirror:
41
amd:
42
- device: mi300_1
42
+ device: mi325_1
43
timeout_in_minutes: 65
44
depends_on:
45
- image-build-amd
@@ -78,7 +78,7 @@ steps:
78
- pytest -v -s v1/e2e/spec_decode -k "ngram or suffix"
79
mirror:
80
amd:
81
- device: mi300_1
81
+ device: mi325_1
82
timeout_in_minutes: 65
83
depends_on:
84
- image-build-amd
@@ -103,7 +103,7 @@ steps:
103
- pytest -v -s v1/e2e/spec_decode -k "draft_model or no_sync or batch_inference"
104
mirror:
105
amd:
106
- device: mi300_1
106
+ device: mi325_1
107
timeout_in_minutes: 50
108
depends_on:
109
- image-build-amd