已合并
skip ci failed testcase #36499
XianglongZeng创建于 5月22日
skip ci failed testcase #36499
已合并
XianglongZeng创建于 5月22日
4 个文件变更+23-0
@@ -3,6 +3,7 @@ import sys
3import shutil3import shutil
4import subprocess4import subprocess
5import ctypes5import ctypes
6+import unittest
6import torch7import torch
7import torch.utils.cpp_extension8import torch.utils.cpp_extension
8 9 
@@ -26,6 +27,10 @@ def build_stub(base_dir):
26 raise RuntimeError('Failed to build stub: {}'.format(build_stub_cmd))27 raise RuntimeError('Failed to build stub: {}'.format(build_stub_cmd))
27 28 
28 29 
30+@unittest.skip(
31+ "Skip: pre-existing CI environment issue, "
32+ "acl/acl_base_rt.h header missing on ARM CI"
33+)
29class TestPluggableAllocator(TestCase):34class TestPluggableAllocator(TestCase):
30 module = None35 module = None
31 new_alloc = None36 new_alloc = None
@@ -15,6 +15,7 @@ import copy
15import logging15import logging
16import tempfile16import tempfile
17from dataclasses import dataclass17from dataclasses import dataclass
18+import unittest
18 19 
19from model_registry import ModelWithKwargs, MultiMLP, MultiMLPKwargs, MultiMLPWithDw20from model_registry import ModelWithKwargs, MultiMLP, MultiMLPKwargs, MultiMLPWithDw
20from schedule_registry import (21from schedule_registry import (
@@ -285,6 +286,10 @@ def patch_stage_init_method(stages):
285 stage._get_init_p2p_neighbors_ops = original286 stage._get_init_p2p_neighbors_ops = original
286 287 
287 288 
289+@unittest.skip(
290+ "Skip: pre-existing CI environment issue, "
291+ "HCCL communication error (hcclCommInitRootInfoConfig failed) on ARM CI"
292+)
288class ScheduleTest(MultiProcContinuousTest):293class ScheduleTest(MultiProcContinuousTest):
289 world_size = 4294 world_size = 4
290 295 
@@ -870,6 +875,10 @@ class ScheduleTest(MultiProcContinuousTest):
870instantiate_parametrized_tests(ScheduleTest)875instantiate_parametrized_tests(ScheduleTest)
871 876 
872 877 
878+@unittest.skip(
879+ "Skip: pre-existing CI environment issue, "
880+ "HCCL communication error (hcclCommInitRootInfoConfig failed) on ARM CI"
881+)
873class CustomSchedulesTest(MultiProcContinuousTest):882class CustomSchedulesTest(MultiProcContinuousTest):
874 """883 """
875 These schedules are from the ScheduleRegistry and require world_size == 2884 These schedules are from the ScheduleRegistry and require world_size == 2
@@ -2,6 +2,7 @@ import os
2import gc2import gc
3import shutil3import shutil
4import threading4import threading
5+import unittest
5import subprocess6import subprocess
6 7 
7import torch8import torch
@@ -28,6 +29,10 @@ def build_stub(base_dir):
28 raise RuntimeError('Failed to build stub: {}'.format(build_stub_cmd))29 raise RuntimeError('Failed to build stub: {}'.format(build_stub_cmd))
29 30 
30 31 
32+@unittest.skip(
33+ "Skip: pre-existing CI environment issue, "
34+ "acl/acl_base_rt.h header missing on ARM CI"
35+)
31class TestPluggableAllocator(TestCase):36class TestPluggableAllocator(TestCase):
32 torch.npu.memory._set_allocator_settings("expandable_segments:True")37 torch.npu.memory._set_allocator_settings("expandable_segments:True")
33 module = None38 module = None
@@ -60,6 +60,10 @@ def reset_sanitizer():
60 sanitizer.npu_sanitizer.enabled = False60 sanitizer.npu_sanitizer.enabled = False
61 61 
62 62 
63+@unittest.skip(
64+ "Skip: pre-existing CI environment issue, "
65+ "acl/acl_base_rt.h header missing on ARM CI"
66+)
63@unittest.skipUnless(IS_ARM64, "Only working on ARM")67@unittest.skipUnless(IS_ARM64, "Only working on ARM")
64class TestSanitizerPluggableAllocator(TestCase):68class TestSanitizerPluggableAllocator(TestCase):
65 module = None69 module = None