[ROCm][CI] Query HIP device memory for test GPU teardown waits. (#57450)

Signed-off-by: Sheral Kumar <shekumar@amd.com>
Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Co-authored-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Andreas Karatzas <akaratza@amd.com>
This commit is contained in:
Sheral Kumar
2026-09-20 15:20:39 -05:00
committed by GitHub
co-authored by Andreas Karatzas Codex Andreas Karatzas
parent 4868312128
commit 01f1f58f10
4 changed files with 56 additions and 53 deletions
+2 -5
View File
@@ -901,13 +901,10 @@ class HfRunner:
# tests/basic_correctness/test_basic_correctness.py::test_models_distributed
# where vllm worker processes are still alive and holding GPU
# memory when hf_runner.__exit__ is called.
from tests.utils import (
get_physical_device_indices,
record_gpu_memory_usage_stats,
)
from tests.utils import record_gpu_memory_usage_stats
if (device_count := current_platform.device_count()) > 0:
devices = get_physical_device_indices(devices=list(range(device_count)))
devices = list(range(device_count))
mem_usage_stats = record_gpu_memory_usage_stats(devices=devices)
self.threshold_ratios = {
device: 0.05 + mem_used / mem_tot