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 11, 11:15 UTC
[Refactor] [3/N] Move tool parser tests and run on CPU (#30693)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
Dec 15, 2025 at 21:45 UTC
ed586e7724fdf91b391abcf6f3e473be641ff5d6
16 files changed
+14
-53
.buildkite/test-amd.yaml
+5
-15
@@ -61,8 +61,8 @@ steps:
61
- pytest -v -s -m 'not cpu_test' multimodal
62
- pytest -v -s utils_
63
64
-- label: Async Engine, Inputs, Utils, Worker, Config Test (CPU) # 15min
65
- timeout_in_minutes: 20
64
+- label: Async Engine, Inputs, Utils, Worker, Config Test (CPU) # 20min
65
+ timeout_in_minutes: 30
66
mirror_hardwares: [amdexperimental, amdproduction, amdtentative]
67
agent_pool: mi325_1
68
grade: Blocking
@@ -73,6 +73,7 @@ steps:
73
- tests/multimodal
74
- tests/standalone_tests/lazy_imports.py
75
- tests/tokenizers_
76
+ - tests/tool_parsers
77
- tests/transformers_utils
78
- tests/config
79
no_gpu: true
@@ -82,6 +83,7 @@ steps:
83
- pytest -v -s test_outputs.py
84
- pytest -v -s -m 'cpu_test' multimodal
85
- pytest -v -s tokenizers_
86
+ - pytest -v -s tool_parsers
87
- pytest -v -s transformers_utils
88
- pytest -v -s config
89
@@ -759,19 +761,7 @@ steps:
761
- vllm/
762
- tests/tool_use
763
commands:
762
- - pytest -v -s -m 'not cpu_test' tool_use
763
-
764
-- label: OpenAI-Compatible Tool Use (CPU) # 5 mins
765
- mirror_hardwares: [amdexperimental, amdproduction]
766
- agent_pool: mi325_1
767
- # grade: Blocking
768
- timeout_in_minutes: 10
769
- source_file_dependencies:
770
- - vllm/
771
- - tests/tool_use
772
- no_gpu: true
773
- commands:
774
- - pytest -v -s -m 'cpu_test' tool_use
764
+ - pytest -v -s tool_use
765
766
##### models test #####
767
.buildkite/test-pipeline.yaml
+5
-12
@@ -57,8 +57,8 @@ steps:
57
- pytest -v -s -m 'not cpu_test' multimodal
58
- pytest -v -s utils_
59
60
-- label: Async Engine, Inputs, Utils, Worker, Config Test (CPU) # 15min
61
- timeout_in_minutes: 20
60
+- label: Async Engine, Inputs, Utils, Worker, Config Test (CPU) # 20min
61
+ timeout_in_minutes: 30
62
source_file_dependencies:
63
- vllm/
64
- tests/test_inputs.py
@@ -66,6 +66,7 @@ steps:
66
- tests/multimodal
67
- tests/standalone_tests/lazy_imports.py
68
- tests/tokenizers_
69
+ - tests/tool_parsers
70
- tests/transformers_utils
71
- tests/config
72
no_gpu: true
@@ -75,6 +76,7 @@ steps:
76
- pytest -v -s test_outputs.py
77
- pytest -v -s -m 'cpu_test' multimodal
78
- pytest -v -s tokenizers_
79
+ - pytest -v -s tool_parsers
80
- pytest -v -s transformers_utils
81
- pytest -v -s config
82
@@ -672,16 +674,7 @@ steps:
674
- vllm/
675
- tests/tool_use
676
commands:
675
- - pytest -v -s -m 'not cpu_test' tool_use
676
-
677
-- label: OpenAI-Compatible Tool Use (CPU) # 5 mins
678
- timeout_in_minutes: 10
679
- source_file_dependencies:
680
- - vllm/
681
- - tests/tool_use
682
- no_gpu: true
683
- commands:
684
- - pytest -v -s -m 'cpu_test' tool_use
677
+ - pytest -v -s tool_use
678
679
##### models test #####
680
.buildkite/test_areas/misc.yaml
+3
-1
@@ -115,7 +115,7 @@ steps:
115
116
- label: Async Engine, Inputs, Utils, Worker, Config (CPU)
117
depends_on: ~
118
- timeout_in_minutes: 20
118
+ timeout_in_minutes: 30
119
source_file_dependencies:
120
- vllm/
121
- tests/test_inputs.py
@@ -123,6 +123,7 @@ steps:
123
- tests/multimodal
124
- tests/standalone_tests/lazy_imports.py
125
- tests/tokenizers_
126
+ - tests/tool_parsers
127
- tests/transformers_utils
128
- tests/config
129
no_gpu: true
@@ -132,6 +133,7 @@ steps:
133
- pytest -v -s test_outputs.py
134
- pytest -v -s -m 'cpu_test' multimodal
135
- pytest -v -s tokenizers_
136
+ - pytest -v -s tool_parsers
137
- pytest -v -s transformers_utils
138
- pytest -v -s config
139
.buildkite/test_areas/tool_use.yaml
+1
-11
@@ -10,14 +10,4 @@ steps:
10
- vllm/
11
- tests/tool_use
12
commands:
13
- - pytest -v -s -m 'not cpu_test' tool_use
14
-
15
-- label: OpenAI-Compatible Tool Use (CPU)
16
- depends_on: ~
17
- timeout_in_minutes: 10
18
- source_file_dependencies:
19
- - vllm/
20
- - tests/tool_use
21
- no_gpu: true
22
- commands:
23
- - pytest -v -s -m 'cpu_test' tool_use
13
+ - pytest -v -s tool_use
tests/tool_parsers/__init__.py
tests/tool_parsers/test_deepseekv31_tool_parser.py
renamed
tests/tool_parsers/test_ernie45_moe_tool_parser.py
renamed
tests/tool_parsers/test_glm4_moe_tool_parser.py
renamed
-2
@@ -12,8 +12,6 @@ from vllm.tool_parsers.glm4_moe_tool_parser import (
12
Glm4MoeModelToolParser,
13
)
14
15
-pytestmark = pytest.mark.cpu_test
16
-
15
pytest.skip("skip glm4_moe parser test", allow_module_level=True)
16
# Use a common model that is likely to be available
17
MODEL = "zai-org/GLM-4.5"
tests/tool_parsers/test_jamba_tool_parser.py
renamed
-2
@@ -13,8 +13,6 @@ from vllm.tokenizers import TokenizerLike, get_tokenizer
13
from vllm.tokenizers.detokenizer_utils import detokenize_incrementally
14
from vllm.tool_parsers.jamba_tool_parser import JambaToolParser
15
16
-pytestmark = pytest.mark.cpu_test
17
-
16
MODEL = "ai21labs/Jamba-tiny-dev"
17
18
tests/tool_parsers/test_kimi_k2_tool_parser.py
renamed
-2
@@ -10,8 +10,6 @@ from vllm.entrypoints.openai.protocol import FunctionCall, ToolCall
10
from vllm.tokenizers import get_tokenizer
11
from vllm.tool_parsers.kimi_k2_tool_parser import KimiK2ToolParser
12
13
-pytestmark = pytest.mark.cpu_test
14
-
13
# Use a common model that is likely to be available
14
MODEL = "moonshotai/Kimi-K2-Instruct"
15
tests/tool_parsers/test_minimax_tool_parser.py
renamed
-2
@@ -15,8 +15,6 @@ from vllm.entrypoints.openai.protocol import (
15
from vllm.tokenizers import get_tokenizer
16
from vllm.tool_parsers.minimax_tool_parser import MinimaxToolParser
17
18
-pytestmark = pytest.mark.cpu_test
19
-
18
# Use a common model that is likely to be available
19
MODEL = "MiniMaxAi/MiniMax-M1-40k"
20
tests/tool_parsers/test_mistral_tool_parser.py
renamed
tests/tool_parsers/test_openai_tool_parser.py
renamed
tests/tool_parsers/test_qwen3coder_tool_parser.py
renamed
-2
@@ -20,8 +20,6 @@ from vllm.tool_parsers.qwen3coder_tool_parser import (
20
)
21
from vllm.tool_parsers.qwen3xml_tool_parser import Qwen3XMLToolParser
22
23
-pytestmark = pytest.mark.cpu_test
24
-
23
MODEL = "Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8"
24
25
tests/tool_parsers/test_seed_oss_tool_parser.py
renamed
-2
@@ -18,8 +18,6 @@ from vllm.tokenizers import TokenizerLike, get_tokenizer
18
from vllm.tokenizers.detokenizer_utils import detokenize_incrementally
19
from vllm.tool_parsers.seed_oss_tool_parser import SeedOssToolParser
20
21
-pytestmark = pytest.mark.cpu_test
22
-
21
# Use a common model that is likely to be available
22
MODEL = "ByteDance-Seed/Seed-OSS-36B-Instruct"
23
tests/tool_parsers/test_xlam_tool_parser.py
renamed
-2
@@ -16,8 +16,6 @@ from vllm.tokenizers import TokenizerLike, get_tokenizer
16
from vllm.tokenizers.detokenizer_utils import detokenize_incrementally
17
from vllm.tool_parsers.xlam_tool_parser import xLAMToolParser
18
19
-pytestmark = pytest.mark.cpu_test
20
-
19
# Use a common model that is likely to be available
20
MODEL = "Salesforce/Llama-xLAM-2-8B-fc-r"
21