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 16, 22:00 UTC
[ci] Add grouped tests & mark tests to run by default for fastcheck pipeline (#6365)
Signed-off-by: kevin <kevin@anyscale.com>
Kevin H. Luu committed
Jul 12, 2024 at 09:58 UTC
b75bce1008fba05ae6e0dcb8060a61015a3c0129
1 file changed
+30
.buildkite/test-pipeline.yaml
+30
@@ -7,8 +7,33 @@
7
8
9
steps:
10
+- label: Async Engine, Inputs, Utils, Worker Test
11
+ fast_check: true
12
+ fast_check_only: true
13
+ commands:
14
+ - pytest -v -s async_engine # Async Engine
15
+ - bash ../.buildkite/download-images.sh # Inputs
16
+ - pytest -v -s test_inputs.py
17
+ - pytest -v -s multimodal
18
+ - pytest -v -s test_utils.py # Utils
19
+ - pytest -v -s worker # Worker
20
+
21
+- label: Tensorizer, Metrics, Tracing Test
22
+ fast_check: true
23
+ fast_check_only: true
24
+ commands:
25
+ - apt-get install curl libsodium23 && pytest -v -s tensorizer_loader # Tensorizer
26
+ - pytest -v -s metrics # Metrics
27
+ - "pip install \
28
+ opentelemetry-sdk \
29
+ opentelemetry-api \
30
+ opentelemetry-exporter-otlp \
31
+ opentelemetry-semantic-conventions-ai" # Tracing
32
+ - pytest -v -s tracing
33
+
34
- label: Regression Test
35
mirror_hardwares: [amd]
36
+ fast_check: true
37
command: pytest -v -s test_regression.py
38
working_dir: "/vllm-workspace/tests" # optional
39
@@ -18,6 +43,7 @@ steps:
43
44
- label: Basic Correctness Test
45
mirror_hardwares: [amd]
46
+ fast_check: true
47
commands:
48
- pip install https://github.com/flashinfer-ai/flashinfer/releases/download/v0.0.8/flashinfer-0.0.8+cu121torch2.3-cp310-cp310-linux_x86_64.whl
49
- VLLM_ATTENTION_BACKEND=XFORMERS pytest -v -s basic_correctness/test_basic_correctness.py
@@ -29,6 +55,7 @@ steps:
55
56
- label: Core Test
57
mirror_hardwares: [amd]
58
+ fast_check: true
59
commands:
60
- pytest -v -s core
61
- pytest -v -s distributed/test_parallel_state.py
@@ -68,6 +95,7 @@ steps:
95
#mirror_hardwares: [amd]
96
working_dir: "/vllm-workspace/tests"
97
num_gpus: 4
98
+ fast_check: true
99
commands:
100
- pytest -v -s distributed/test_pynccl.py
101
# We want to test that models which use 2 GPUs work with 4 GPUs, which is why we duplicate them here.
@@ -95,6 +123,7 @@ steps:
123
- pytest -v -s tokenization
124
125
- label: Entrypoints Test
126
+ fast_check: true
127
mirror_hardwares: [amd]
128
129
commands:
@@ -228,6 +257,7 @@ steps:
257
258
- label: Documentation Build
259
working_dir: "/vllm-workspace/test_docs/docs"
260
+ fast_check: true
261
no_gpu: True
262
commands:
263
- pip install -r requirements-docs.txt