Skip to content

Commit

Permalink
[ci][distributed] fix flaky tests (vllm-project#6806)
Browse files Browse the repository at this point in the history
  • Loading branch information
youkaichao authored and cadedaniel committed Jul 27, 2024
1 parent 017a755 commit 4de3a86
Showing 1 changed file with 8 additions and 1 deletion.
9 changes: 8 additions & 1 deletion tests/distributed/test_pipeline_parallel.py
Original file line number Diff line number Diff line change
@@ -1,3 +1,10 @@
"""
WARNING: This test runs in both single-node (4 GPUs) and multi-node
(2 node with 2 GPUs each) modes. If the test only uses 2 GPUs, it is
important to set the distributed backend to "mp" to avoid Ray scheduling
all workers in a node other than the head node, which can cause the test
to fail.
"""
import os

import pytest
Expand Down Expand Up @@ -78,7 +85,7 @@ def test_pp_cudagraph(PP_SIZE, MODEL_NAME, ATTN_BACKEND):
"--pipeline-parallel-size",
str(PP_SIZE),
"--distributed-executor-backend",
"ray",
"mp",
]
os.environ["VLLM_ATTENTION_BACKEND"] = ATTN_BACKEND

Expand Down

0 comments on commit 4de3a86

Please sign in to comment.