[BUG] Import Error: `name '_disable_dynamo_if_unsupported' is not defined`

[Thinksky5124]

Describe the bug
When torch < 2.4.0 and deepspeed >= 0.17.5，import deepspeed will lead error:

Python 3.10.15 (main, Oct  3 2024, 07:27:34) [GCC 11.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import deepspeed
[2025-08-25 10:30:30,981] [INFO] [real_accelerator.py:260:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[WARNING] ZenFlow disabled: torch internal optimizer symbols could not be imported: cannot import name '_disable_dynamo_if_unsupported' from 'torch.optim.optimizer' (/home/users/wujun.wen/miniconda3/envs/torch230_py310/lib/python3.10/site-packages/torch/optim/optimizer.py)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/users/wujun.wen/miniconda3/envs/torch230_py310/lib/python3.10/site-packages/deepspeed/__init__.py", line 25, in <module>
    from . import ops
  File "/home/users/wujun.wen/miniconda3/envs/torch230_py310/lib/python3.10/site-packages/deepspeed/ops/__init__.py", line 6, in <module>
    from . import adam
  File "/home/users/wujun.wen/miniconda3/envs/torch230_py310/lib/python3.10/site-packages/deepspeed/ops/adam/__init__.py", line 9, in <module>
    from .zenflow_torch_adam import ZenFlowSelectiveAdamW
  File "/home/users/wujun.wen/miniconda3/envs/torch230_py310/lib/python3.10/site-packages/deepspeed/ops/adam/zenflow_torch_adam.py", line 685, in <module>
    @_disable_dynamo_if_unsupported(single_tensor_fn=_single_tensor_adamw)
NameError: name '_disable_dynamo_if_unsupported' is not defined

It is becase a wrong version check for torch in deepspeed/ops/adam/zenflow_torch_adam.py L13 .

from deepspeed.utils.torch import required_torch_version

# Check if we have PyTorch >= 2.0 for ZenFlow features
_ZENFLOW_AVAILABLE = required_torch_version(min_version=2.1)

if _ZENFLOW_AVAILABLE:
    try:
        from torch.optim.optimizer import (
            _default_to_fused_or_foreach,
            _disable_dynamo_if_unsupported,
            _get_capturable_supported_devices,
            _get_value,
            _stack_if_compiling,
            _view_as_real,
            DeviceDict,
            Optimizer,
        )
    except ImportError as e:
        print(f"[WARNING] ZenFlow disabled: torch internal optimizer symbols could not be imported: {e}")
        _ZENFLOW_AVAILABLE = False
else:
    # safe disable dynamo if unsupported
    def _disable_dynamo_if_unsupported(**kwargs):

        def wrapper(fn):
            return fn

        return wrapper

    _ZENFLOW_AVAILABLE = False

Function _disable_dynamo_if_unsupported is avaliable from torch >= 2.4.0. And try except not registry _disable_dynamo_if_unsupported function when import error occur.