@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 5844017285
Fork 0

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