PyTorch Lightning TypeError：__init__()收到意外关键字参数'checkpoint_callback'

错误信息 #

TypeError                                 Traceback (most recent call last)
<ipython-input-41-2892cdd4e738> in <module>()
      5   max_epochs=N_EPOCHS,
      6   gpus=1, #GPU
----> 7   progress_bar_refresh_rate=30
      8 )

/usr/local/lib/python3.7/dist-packages/pytorch_lightning/utilities/argparse.py in insert_env_defaults(self, *args, **kwargs)
    343 
    344         # all args were already moved to kwargs
--> 345         return fn(self, **kwargs)
    346 
    347     return cast(_T, insert_env_defaults)

TypeError: __init__() got an unexpected keyword argument 'checkpoint_callback'

运行的代码片段 #

trainer = pl.Trainer(
  logger=logger, 
  checkpoint_callback=checkpoint_callback,
  callbacks=[early_stopping_callback],
  max_epochs=N_EPOCHS,
  gpus=1, #GPU
  progress_bar_refresh_rate=30
)

checkpoint_callback定义 #

checkpoint_callback = ModelCheckpoint(
  dirpath="checkpoints",
  filename="best-checkpoint",
  save_top_k=1,
  verbose=True,
  monitor="val_loss",
  mode="min"
)

解决方案 #

错误原因是PyTorch Lightning 1.5及以上版本中，checkpoint_callback不再是pl.Trainer的独立参数，所有回调类（包括模型检查点、早停）都需要统一放入callbacks列表传递。

修改后的Trainer初始化代码：

trainer = pl.Trainer(
  logger=logger, 
  callbacks=[early_stopping_callback, checkpoint_callback],
  max_epochs=N_EPOCHS,
  gpus=1, #GPU
  progress_bar_refresh_rate=30
)

只需移除checkpoint_callback=checkpoint_callback这一行，将checkpoint_callback实例添加到callbacks列表中即可解决该错误。

内容的提问来源于stack exchange，提问作者Quantizer