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, 18:15 UTC
[ci] [multi-step] narrow multi-step test dependency paths (#7760)
William Lin committed
Aug 21, 2024 at 15:52 UTC
5844017285acda7060ffc62e3dcedc0775eb4fe2
1 file changed
+11
-3
.buildkite/test-pipeline.yaml
+11
-3
@@ -312,12 +312,20 @@ steps:
312
- CUDA_VISIBLE_DEVICES=0,1 pytest -v -s test_sharded_state_loader.py
313
- CUDA_VISIBLE_DEVICES=0,1 pytest -v -s distributed/test_utils.py
314
315
-- label: Multi-step Tests (4 GPUs) # 10min
315
+- label: Multi-step Tests (4 GPUs) # 21min
316
working_dir: "/vllm-workspace/tests"
317
num_gpus: 4
318
source_file_dependencies:
319
- - vllm/
320
- - tests/multi_step/test_correctness.py
319
+ - vllm/model_executor/layers/sampler.py
320
+ - vllm/sequence.py
321
+ - vllm/worker/worker_base.py
322
+ - vllm/worker/worker.py
323
+ - vllm/worker/multi_step_worker.py
324
+ - vllm/worker/model_runner_base.py
325
+ - vllm/worker/model_runner.py
326
+ - vllm/worker/multi_step_model_runner.py
327
+ - vllm/engine
328
+ - tests/multi_step
329
commands:
330
- pytest -v -s multi_step/test_correctness.py
331