Pytorch-lightning导入报错：无法导入_TRANSFORMERS_GREATER_EQUAL_4_4

执行以下导入代码：

import torch
from torch.nn import functional as F
from torch import nn
from torch.utils.data import TensorDataset, DataLoader

from torchvision import datasets, transforms

from pytorch_lightning.core.lightning import LightningModule
from pytorch_lightning import Trainer
from torchmetrics.functional.classification.accuracy import accuracy

出现如下导入错误：

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
Cell In[77], line 8
      4 from torch.utils.data import TensorDataset, DataLoader
      6 from torchvision import datasets, transforms
----> 8 from pytorch_lightning.core.lightning import LightningModule
      9 from pytorch_lightning import Trainer
     10 from torchmetrics.functional.classification.accuracy import accuracy

File d:\anaconda3\envs\dl\Lib\site-packages\pytorch_lightning\__init__.py:26
     23     _logger.propagate = False
     25 from lightning_fabric.utilities.seed import seed_everything  # noqa: E402
---> 26 from pytorch_lightning.callbacks import Callback  # noqa: E402
     27 from pytorch_lightning.core import LightningDataModule, LightningModule  # noqa: E402
     28 from pytorch_lightning.trainer import Trainer  # noqa: E402

File d:\anaconda3\envs\dl\Lib\site-packages\pytorch_lightning\callbacks\__init__.py:14
      1 # Copyright The Lightning AI team.
      2 #
      3 # Licensed under the Apache License, Version 2.0 (the "License");
   (...)
     12 # See the License for the specific language governing permissions and
     13 # limitations under the License.
---> 14 from pytorch_lightning.callbacks.batch_size_finder import BatchSizeFinder
     15 from pytorch_lightning.callbacks.callback import Callback
...
---> 28 from torchmetrics.utilities.imports import _TRANSFORMERS_GREATER_EQUAL_4_4
     30 if _TRANSFORMERS_GREATER_EQUAL_4_4:
     31     from torchmetrics.functional.text.bert import bert_score

ImportError: cannot import name '_TRANSFORMERS_GREATER_EQUAL_4_4' from 'torchmetrics.utilities.imports' (d:\anaconda3\envs\dl\Lib\site-packages\torchmetrics\utilities\imports.py)

尝试重装torch、pytorch-lightning及安装旧版本pytorch-lightning均未解决。将torchmetrics降级至0.7.0后，报错变为：

ImportError: cannot import name 'pit' from 'torchmetrics.functional.audio.pit' (d:\anaconda3\envs\dl\Lib\site-packages\torchmetrics\functional\audio\pit.py)

这是PyTorch Lightning和TorchMetrics版本不兼容导致的依赖问题，按以下步骤操作即可解决：

• 先卸载现有相关包：

  pip uninstall -y pytorch-lightning torchmetrics
  

• 安装经过验证的兼容版本（根据你的PyTorch版本选其一）：

  • 若使用PyTorch 1.12及以下，选这个稳定组合：

    pip install pytorch-lightning==1.9.5 torchmetrics==0.11.4
    

  • 若使用PyTorch 2.x，可选这个较新组合：

    pip install pytorch-lightning==2.0.9 torchmetrics==0.11.4
    

• 重新运行原导入代码，确认无报错即可。

注意：不要单独升级或降级某一个包，PyTorch Lightning对TorchMetrics的版本有明确要求，安装时必须保持两者版本匹配。

内容的提问来源于stack exchange，提问作者Karol