A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 8, 10:00 UTC
[ci] Integrate AMD tests into CI (#33626)
Signed-off-by: Kevin H. Luu <khluu000@gmail.com> Signed-off-by: khluu <khluu000@gmail.com> Co-authored-by: TJian <tunjian.tan@embeddedllm.com>
Kevin H. Luu committed
Feb 11, 2026 at 16:54 UTC
83b47f67b1dfad505606070ae4d9f83e50ad4ebd
6 files changed
+25
-10
.buildkite/hardware_tests/amd.yaml
+2
-1
@@ -1,6 +1,7 @@
1
-group: Hardware
1
+group: Hardware - AMD Build
2
steps:
3
- label: "AMD: :docker: build image"
4
+ key: image-build-amd
5
depends_on: []
6
device: amd_cpu
7
no_plugin: true
.buildkite/test_areas/basic_correctness.yaml
+5
@@ -14,3 +14,8 @@ steps:
14
- pytest -v -s basic_correctness/test_cumem.py
15
- pytest -v -s basic_correctness/test_basic_correctness.py
16
- pytest -v -s basic_correctness/test_cpu_offload.py
17
+ mirror:
18
+ amd:
19
+ device: mi325_1
20
+ depends_on:
21
+ - image-build-amd
.buildkite/test_areas/entrypoints.yaml
+5
@@ -24,6 +24,11 @@ steps:
24
- pytest -v -s entrypoints/llm --ignore=entrypoints/llm/test_generate.py --ignore=entrypoints/llm/test_collective_rpc.py
25
- pytest -v -s entrypoints/llm/test_generate.py # it needs a clean process
26
- pytest -v -s entrypoints/offline_mode # Needs to avoid interference with other tests
27
+ mirror:
28
+ amd:
29
+ device: mi325_1
30
+ depends_on:
31
+ - image-build-amd
32
33
- label: Entrypoints Integration (API Server 1)
34
timeout_in_minutes: 130
.buildkite/test_areas/models_basic.yaml
+6
-2
@@ -4,7 +4,6 @@ depends_on:
4
steps:
5
- label: Basic Models Tests (Initialization)
6
timeout_in_minutes: 45
7
- mirror_hardwares: [amdexperimental]
7
torch_nightly: true
8
source_file_dependencies:
9
- vllm/
@@ -16,7 +15,6 @@ steps:
15
16
- label: Basic Models Tests (Extra Initialization) %N
17
timeout_in_minutes: 45
19
- mirror_hardwares: [amdexperimental]
18
torch_nightly: true
19
source_file_dependencies:
20
- vllm/model_executor/models/
@@ -38,6 +36,12 @@ steps:
36
- tests/models/test_registry.py
37
commands:
38
- pytest -v -s models/test_terratorch.py models/test_transformers.py models/test_registry.py
39
+ mirror:
40
+ amd:
41
+ device: mi325_1
42
+ depends_on:
43
+ - image-build-amd
44
+
45
46
- label: Basic Models Test (Other CPU) # 5min
47
depends_on:
.buildkite/test_areas/models_language.yaml
-7
@@ -4,7 +4,6 @@ depends_on:
4
steps:
5
- label: Language Models Tests (Standard)
6
timeout_in_minutes: 25
7
- mirror_hardwares: [amdexperimental]
7
torch_nightly: true
8
source_file_dependencies:
9
- vllm/
@@ -16,7 +15,6 @@ steps:
15
16
- label: Language Models Tests (Extra Standard) %N
17
timeout_in_minutes: 45
19
- mirror_hardwares: [amdexperimental]
18
torch_nightly: true
19
source_file_dependencies:
20
- vllm/model_executor/models/
@@ -32,7 +30,6 @@ steps:
30
31
- label: Language Models Tests (Hybrid) %N
32
timeout_in_minutes: 75
35
- mirror_hardwares: [amdexperimental]
33
torch_nightly: true
34
source_file_dependencies:
35
- vllm/
@@ -48,7 +45,6 @@ steps:
45
46
- label: Language Models Test (Extended Generation) # 80min
47
timeout_in_minutes: 110
51
- mirror_hardwares: [amdexperimental]
48
optional: true
49
source_file_dependencies:
50
- vllm/
@@ -62,7 +58,6 @@ steps:
58
59
- label: Language Models Test (PPL)
60
timeout_in_minutes: 110
65
- mirror_hardwares: [amdexperimental]
61
optional: true
62
source_file_dependencies:
63
- vllm/
@@ -72,7 +67,6 @@ steps:
67
68
- label: Language Models Test (Extended Pooling) # 36min
69
timeout_in_minutes: 50
75
- mirror_hardwares: [amdexperimental]
70
optional: true
71
source_file_dependencies:
72
- vllm/
@@ -82,7 +76,6 @@ steps:
76
77
- label: Language Models Test (MTEB)
78
timeout_in_minutes: 110
85
- mirror_hardwares: [amdexperimental]
79
optional: true
80
source_file_dependencies:
81
- vllm/
.buildkite/test_areas/samplers.yaml
+7
@@ -12,3 +12,10 @@ steps:
12
commands:
13
- pytest -v -s samplers
14
- VLLM_USE_FLASHINFER_SAMPLER=1 pytest -v -s samplers
15
+ mirror:
16
+ amd:
17
+ device: mi325_1
18
+ depends_on:
19
+ - image-build-amd
20
+ commands:
21
+ - pytest -v -s -m 'not skip_v1' samplers