anodyine/vllm-omni
1
0 Fork 0
Code Issues 4 Pull Requests 2 Actions Packages Projects Releases Wiki Activity

[AMD][CI] Fix AMD CI (Partial) (#6704)

Browse Source 
Signed-off-by: root <root@tw020.mia.tensorwave.lan>
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
Co-authored-by: root <root@tw020.mia.tensorwave.lan>
This commit is contained in:
TJianandroot authored and GitHub committed 2026-08-28 14:34:11 +08:00
1 parent c201fd77c6
commit 15c6ab8fa7
6 files changed
+196 -62
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

tests/helpers/clean.py
+10 -2
View File
@@ -150,8 +150,16 @@ def _print_device_processes() -> None:
elif current_omni_platform.is_npu():
_run_smi("Ascend NPU Information (npu-smi info)", ["npu-smi", "info"], 40)
elif current_omni_platform.is_rocm():
_run_smi("AMD GPU Information (amd-smi)", ["amd-smi"], 30)
_run_smi("Detailed AMD GPU Processes (amd-smi process)", ["amd-smi", "process"], 100, timeout=3)
# amd-smi can enter uninterruptible sleep in the amdgpu CPER ioctl
# (amdgpu_cper_ring_write). In that state subprocess.run's timeout
# cannot reap it, causing this diagnostic hook to hang the entire test.
# rocm-smi provides the information needed here without that ioctl.
_run_smi(
"AMD GPU Information (rocm-smi)",
["rocm-smi", "--showuse", "--showmeminfo", "vram"],
60,
)
_run_smi("Detailed AMD GPU Processes (rocm-smi)", ["rocm-smi", "--showpids"], 100, timeout=3)
elif current_omni_platform.is_xpu():
_run_smi("Intel XPU Information (xpu-smi discovery)", ["xpu-smi", "discovery"], 40)
elif current_omni_platform.is_musa():