From e6fdeddb9c3bdc15f44a49acfdef053479a4c9dc Mon Sep 17 00:00:00 2001 From: Shengliang Xu Date: Tue, 27 Oct 2026 18:42:21 +0000 Subject: [PATCH] Run the GPTQ Hessian dp-sync test on the pooled gloo workers The test spawned its own two-process gloo group next to the persistent two-rank pool that `cpu_dist_workers` keeps for the other multi-rank unit tests. It was in flight in all 13 runner stops of the `torch_28` unit job since 10-26 (3.1% of its attempts), always at 97-99% of the run, where the 16 GB runner is nearly full. With a live pool the spawn adds two processes and about 2 GiB of RSS; the test now runs on the pool like the other multi-rank tests. Signed-off-by: Shengliang Xu --- tests/unit/torch/quantization/test_gptq.py | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-) diff --git a/tests/unit/torch/quantization/test_gptq.py b/tests/unit/torch/quantization/test_gptq.py index 6e5b01190ec..1c303130299 100644 --- a/tests/unit/torch/quantization/test_gptq.py +++ b/tests/unit/torch/quantization/test_gptq.py @@ -17,11 +17,9 @@ import copy import warnings -from functools import partial import pytest import torch -from _test_utils.torch.distributed.utils import spawn_multiprocess_job import modelopt.torch.quantization as mtq from modelopt.torch.quantization.utils import calib_utils @@ -200,11 +198,9 @@ def _gptq_hessian_dp_sync_helper(rank, size, token_counts): torch.testing.assert_close(helper.hessian, ref) -def test_gptq_hessian_dp_sync(skip_on_windows): +def test_gptq_hessian_dp_sync(cpu_dist_workers): """Token-weighted merge over a gloo (CPU) group, with one rank that saw no tokens.""" - spawn_multiprocess_job( - 2, partial(_gptq_hessian_dp_sync_helper, token_counts=(5, 0)), backend="gloo" - ) + cpu_dist_workers.run(_gptq_hessian_dp_sync_helper, token_counts=(5, 0)) def test_a_calibration_run_forgets_earlier_forward_measurements(monkeypatch):