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, 07:00 UTC
For ppc64le, disabled tests for now and addressed space issues (#10538)
Nishidha committed
Nov 23, 2024 at 15:03 UTC
651f6c31ac86f29aa72fa682ef6c34349bcc75db
1 file changed
+3
-41
.buildkite/run-cpu-test-ppc64le.sh
+3
-41
@@ -4,49 +4,11 @@
4
# It serves a sanity check for compilation and basic model usage.
5
set -ex
6
7
-# Try building the docker image
8
-docker build -t cpu-test -f Dockerfile.ppc64le .
9
-
7
# Setup cleanup
11
-remove_docker_container() { docker rm -f cpu-test || true; }
8
+remove_docker_container() { docker rm -f cpu-test || true; docker system prune -f; }
9
trap remove_docker_container EXIT
10
remove_docker_container
11
15
-# Run the image, setting --shm-size=4g for tensor parallel.
16
-source /etc/environment
17
-#docker run -itd --entrypoint /bin/bash -v ~/.cache/huggingface:/root/.cache/huggingface --privileged=true --network host -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=4 --shm-size=4g --name cpu-test cpu-test
18
-docker run -itd --entrypoint /bin/bash -v ~/.cache/huggingface:/root/.cache/huggingface --privileged=true --network host -e HF_TOKEN="$HF_TOKEN" --name cpu-test cpu-test
19
-
20
-function cpu_tests() {
21
- set -e
22
-
23
- # Run basic model test
24
- docker exec cpu-test bash -c "
25
- set -e
26
- pip install pytest pytest-asyncio \
27
- decord einops librosa peft Pillow sentence-transformers soundfile \
28
- transformers_stream_generator matplotlib datamodel_code_generator
29
- pip install torchvision --index-url https://download.pytorch.org/whl/cpu
30
- pytest -v -s tests/models/decoder_only/language -m cpu_model
31
- pytest -v -s tests/models/embedding/language -m cpu_model
32
- pytest -v -s tests/models/encoder_decoder/language -m cpu_model
33
- pytest -v -s tests/models/decoder_only/audio_language -m cpu_model
34
- pytest -v -s tests/models/decoder_only/vision_language -m cpu_model"
35
-
36
- # online inference
37
- docker exec cpu-test bash -c "
38
- set -e
39
- python3 -m vllm.entrypoints.openai.api_server --model facebook/opt-125m &
40
- timeout 600 bash -c 'until curl localhost:8000/v1/models; do sleep 1; done' || exit 1
41
- python3 benchmarks/benchmark_serving.py \
42
- --backend vllm \
43
- --dataset-name random \
44
- --model facebook/opt-125m \
45
- --num-prompts 20 \
46
- --endpoint /v1/completions \
47
- --tokenizer facebook/opt-125m"
48
-}
12
+# Try building the docker image
13
+docker build -t cpu-test -f Dockerfile.ppc64le .
14
50
-# All of CPU tests are expected to be finished less than 25 mins.
51
-export -f cpu_tests
52
-timeout 25m bash -c "cpu_tests"