From eb623bfe4acd670c0b079482f6b1a35598f2d7d9 Mon Sep 17 00:00:00 2001 From: qijie Date: Sat, 22 Feb 2025 10:57:48 +0800 Subject: [PATCH 1/4] fix attr error for pt<1.3 --- .../msprobe/pytorch/monitor/distributed/wrap_distributed.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/debug/accuracy_tools/msprobe/pytorch/monitor/distributed/wrap_distributed.py b/debug/accuracy_tools/msprobe/pytorch/monitor/distributed/wrap_distributed.py index b2fa26a58e..eee79d39da 100644 --- a/debug/accuracy_tools/msprobe/pytorch/monitor/distributed/wrap_distributed.py +++ b/debug/accuracy_tools/msprobe/pytorch/monitor/distributed/wrap_distributed.py @@ -20,6 +20,7 @@ import re import torch import torch.distributed as dist import torch.nn as nn +from torch._C._distributed_c10d import Work from msprobe.core.common.const import MonitorConst from msprobe.core.common.file_utils import load_yaml @@ -42,7 +43,7 @@ distributed_func = {} for f in dir(dist): distributed_func[f] = getattr(dist, f) -ORIGIN_WAIT = getattr(dist.Work, 'wait') +ORIGIN_WAIT = getattr(Work, 'wait') PENDING_ASYNC_CC_BY_HANDLE = {} -- Gitee From 1c5e756143905d5ae9a51887db5b7eb92cbf81e3 Mon Sep 17 00:00:00 2001 From: qijie Date: Sat, 22 Feb 2025 15:08:30 +0800 Subject: [PATCH 2/4] remove version check --- debug/accuracy_tools/msprobe/pytorch/monitor/module_hook.py | 4 ---- 1 file changed, 4 deletions(-) diff --git a/debug/accuracy_tools/msprobe/pytorch/monitor/module_hook.py b/debug/accuracy_tools/msprobe/pytorch/monitor/module_hook.py index eea2bdbc2d..99148e853b 100644 --- a/debug/accuracy_tools/msprobe/pytorch/monitor/module_hook.py +++ b/debug/accuracy_tools/msprobe/pytorch/monitor/module_hook.py @@ -42,10 +42,6 @@ from msprobe.pytorch.monitor.utils import get_param_struct, validate_config, val get_output_base_dir, get_target_output_dir from msprobe.pytorch.monitor.visualizer import HeatmapVisualizer -torch_version_above_or_equal_2 = torch.__version__.split('+')[0] >= '2.0' -if not torch_version_above_or_equal_2: - raise ValueError("monitor require torch>=2.0") - FORMAT_MAPPING = { MonitorConst.TENSORBOARD: SummaryWriterWithAD, -- Gitee From 99c8941d68e8096f90a97b8b8d4a6655fb147a83 Mon Sep 17 00:00:00 2001 From: qijie Date: Sat, 22 Feb 2025 15:42:21 +0800 Subject: [PATCH 3/4] update doc --- debug/accuracy_tools/msprobe/docs/19.monitor.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/debug/accuracy_tools/msprobe/docs/19.monitor.md b/debug/accuracy_tools/msprobe/docs/19.monitor.md index 03eb294a7a..41ccb15523 100644 --- a/debug/accuracy_tools/msprobe/docs/19.monitor.md +++ b/debug/accuracy_tools/msprobe/docs/19.monitor.md @@ -9,7 +9,7 @@ 要求: -- PyTorch场景:torch不低于**2.0** +- PyTorch场景:torch不低于**1.11** - MindSpore场景:mindspore不低于**2.4**,仅支持**MindSpore动态图** ## 功能介绍 -- Gitee From e57af910838d252eb01aa681d644a05564d8ad9a Mon Sep 17 00:00:00 2001 From: qijie Date: Mon, 24 Feb 2025 10:51:22 +0800 Subject: [PATCH 4/4] remove version check in init --- debug/accuracy_tools/msprobe/pytorch/__init__.py | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/debug/accuracy_tools/msprobe/pytorch/__init__.py b/debug/accuracy_tools/msprobe/pytorch/__init__.py index ce84e6b35b..da783bcfed 100644 --- a/debug/accuracy_tools/msprobe/pytorch/__init__.py +++ b/debug/accuracy_tools/msprobe/pytorch/__init__.py @@ -19,6 +19,4 @@ from .compare.pt_compare import compare from .common.utils import seed_all from .debugger.precision_debugger import PrecisionDebugger, module_dump, module_dump_end -torch_version_above_or_equal_2 = torch.__version__.split('+')[0] >= '2.0' -if torch_version_above_or_equal_2: - from msprobe.pytorch.monitor.module_hook import TrainerMon +from msprobe.pytorch.monitor.module_hook import TrainerMon -- Gitee