@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 8c6f694a79
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] refine dependency for distributed tests (#7776)

youkaichao committed Aug 22, 2024 at 00:54 UTC 8c6f694a79d7a607ef9ba1f865c610f15c3fc2a4
1 file changed +17 -8
.buildkite/test-pipeline.yaml
+17 -8
@@ -95,7 +95,8 @@ steps:
95 num_gpus: 4
96 fast_check: true
97 source_file_dependencies:
98 - - vllm/
98 + - vllm/distributed/
99 + - vllm/core/
100 - tests/distributed
101 - tests/spec_decode/e2e/test_integration_dist_tp4
102 commands:
@@ -284,8 +285,11 @@ steps:
285 num_gpus: 2
286 num_nodes: 2
287 source_file_dependencies:
287 - - vllm/
288 - - tests/distributed/test_same_node
288 + - vllm/distributed/
289 + - vllm/engine/
290 + - vllm/executor/
291 + - vllm/model_executor/models/
292 + - tests/distributed/
293 commands:
294 - # the following commands are for the first node, with ip 192.168.10.10 (ray environment already set up)
295 - VLLM_TEST_SAME_HOST=0 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_same_node.py
@@ -298,8 +302,11 @@ steps:
302 working_dir: "/vllm-workspace/tests"
303 num_gpus: 2
304 source_file_dependencies:
301 - - vllm/
302 - - tests/distributed
305 + - vllm/distributed/
306 + - vllm/engine/
307 + - vllm/executor/
308 + - vllm/model_executor/models/
309 + - tests/distributed/
310 commands:
311 - VLLM_TEST_SAME_HOST=1 torchrun --nproc-per-node=4 distributed/test_same_node.py
312 - TARGET_TEST_SUITE=L4 pytest -v -s distributed/test_basic_distributed_correctness.py
@@ -333,9 +340,11 @@ steps:
340 working_dir: "/vllm-workspace/tests"
341 num_gpus: 4
342 source_file_dependencies:
336 - - vllm/
337 - - tests/distributed/test_pp_cudagraph.py
338 - - tests/distributed/test_pipeline_parallel
343 + - vllm/distributed/
344 + - vllm/engine/
345 + - vllm/executor/
346 + - vllm/model_executor/models/
347 + - tests/distributed/
348 commands:
349 - pytest -v -s distributed/test_pp_cudagraph.py
350 - pytest -v -s distributed/test_pipeline_parallel.py