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 ] Re-enable Large Model LM Eval (#6031)
Robert Shaw committed
Jul 1, 2024 at 12:40 UTC
d76084c12f0400a8f8364883a841405c903b61f6
1 file changed
+9
.buildkite/test-pipeline.yaml
+9
@@ -198,6 +198,15 @@ steps:
198
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
199
- bash ./run-tests.sh -c configs/models-small.txt -t 1
200
201
+- label: LM Eval Large Models
202
+ gpu: a100
203
+ num_gpus: 4
204
+ working_dir: "/vllm-workspace/.buildkite/lm-eval-harness"
205
+ commands:
206
+ - pip install lm-eval
207
+ - export VLLM_WORKER_MULTIPROC_METHOD=spawn
208
+ - bash ./run-tests.sh -c configs/models-large.txt -t 4
209
+
210
- label: Documentation Build
211
working_dir: "/vllm-workspace/test_docs/docs"
212
no_gpu: True