apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[CI] Allow occasional distributed worker exit_code (#5341)

Browse Source
This commit is contained in:
YuBaoku
2025-12-03 10:56:59 +08:00
committed by GitHub
parent 0eb799a324
commit dfeabee123
5 changed files with 5 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
scripts/run_pre_ce.sh
View File

@@ -47,7 +47,7 @@ for subdir in "$run_path"*/; do
if [ "$exit_code" -eq 1 ] || [ "$exit_code" -eq 124 ]; then if [ "$exit_code" -eq 1 ] || [ "$exit_code" -eq 124 ]; then
echo "[ERROR] $file 起服务或执行异常，exit_code=$exit_code" echo "[ERROR] $file 起服务或执行异常，exit_code=$exit_code"
if [ "$exit_code" -eq 124 ]; then if [ "$exit_code" -eq 124 ]; then
echo "[TIMEOUT] $file 脚本执行超过 6 分钟, 任务超时退出！" echo "[TIMEOUT] $file 脚本执行超过 10 分钟, 任务超时退出！"
fi fi
fi fi

2
tests/ci_use/GLM-45-AIR/test_rollout_model.py
View File

@@ -63,4 +63,4 @@ def test_rollout_model_with_distributed_launch():
print("\n" + "=" * 50 + " STDERR " + "=" * 50) print("\n" + "=" * 50 + " STDERR " + "=" * 50)
print(stderr) print(stderr)
assert return_code != 1, f"Process exited with code {return_code}" assert return_code in (0, 250), f"Process exited with code {return_code}"

2
tests/distributed/test_chunked_moe.py
View File

@@ -43,7 +43,7 @@ def test_fused_moe_launch():
stdout, stderr = process.communicate() stdout, stderr = process.communicate()
return_code = -1 return_code = -1
print(f"std_out: {stdout}") print(f"std_out: {stdout}")
assert return_code != 1, f"Process exited with code {return_code}, stdout: {stdout}, stderr: {stderr}" assert return_code in (0, 250), f"Process exited with code {return_code}, stdout: {stdout}, stderr: {stderr}"
test_fused_moe_launch() test_fused_moe_launch()

2
tests/distributed/test_custom_all_reduce.py
View File

@@ -42,7 +42,7 @@ def test_custom_all_reduce_launch():
process.kill() process.kill()
stdout, stderr = process.communicate() stdout, stderr = process.communicate()
return_code = -1 return_code = -1
assert return_code == 0, f"Process exited with code {return_code}" assert return_code in (0, 250), f"Process exited with code {return_code}"
test_custom_all_reduce_launch() test_custom_all_reduce_launch()

2
tests/e2e/EB_VL_Lite/test_rollout_model.py
View File

@@ -65,4 +65,4 @@ def test_rollout_model_with_distributed_launch():
print("\n" + "=" * 50 + " STDERR " + "=" * 50) print("\n" + "=" * 50 + " STDERR " + "=" * 50)
print(stderr) print(stderr)
assert return_code != 1, f"Process exited with code {return_code}" assert return_code in (0, 250), f"Process exited with code {return_code}"