AttributeError: 'NoneType' object has no attribute 'train'

[miandai]

There is some code may be wrong in lavis/runners/runner_base.py：

def model(self):
        if self._model.device != self.device:
            self._model = self._model.to(self.device)
            # distributed training wrapper
            if self.use_distributed:
                if self._wrapped_model is None:
                    self._wrapped_model = DDP(
                        self._model, device_ids=[self.config.run_cfg.gpu]
                    )
            else:
                self._wrapped_model = self._model

        return self._wrapped_model

should be：

def model(self):
        # move model to device
        if self._model.device != self.device:
            self._model = self._model.to(self.device)
        # distributed training wrapper
        if self.use_distributed:
            if self._wrapped_model is None:
                self._wrapped_model = DDP(
                    self._model, device_ids=[self.config.run_cfg.gpu]
                )
        else:
            self._wrapped_model = self._model

        return self._wrapped_model

Or it will be error：

File "/home/Hamburg/github/LAVIS/train.py", line 103, in main() File "/home/Hamburg/github/LAVIS/train.py", line 99, in main runner.train() File "/home/Hamburg/github/LAVIS/lavis/runners/runner_base.py", line 369, in train train_stats = self.train_epoch(cur_epoch) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/Hamburg/github/LAVIS/lavis/runners/runner_base.py", line 426, in train_epoch self.model.train() ^^^^^^^^^^^^^^^^ AttributeError: 'NoneType' object has no attribute 'train'

[dxli94]

@miandai I can't see the difference. Can you elaborate?

[miandai]

Sorry to my poor English.

Pay attention to the tab num. The first two "if" condition should be serial, not inclusion.

"if self.use_distributed: " should not in "if self._model.device != self.device:"

[dxli94]

Could you please maybe explain in more detail, in Chinese if you like? I'd be really curious to know if this is a bug, that we have to fix it.

[miandai]

def model(self):
        if self._model.device != self.device:
            self._model = self._model.to(self.device)
            # distributed training wrapper
            if self.use_distributed:
                if self._wrapped_model is None:
                    self._wrapped_model = DDP(
                        self._model, device_ids=[self.config.run_cfg.gpu]
                    )
            else:
                self._wrapped_model = self._model

        return self._wrapped_model

以上是当前 master 分支 lavis/runners/runner_base.py 代码。如果没有进入 if self._model.device != self.device 内，会直接返回 self._wrapped_model，即 None，这会引起后续调用 self.model.train() 时报错：AttributeError: 'NoneType' object has no attribute 'train'

个人推测，这里如果没有进入 if self._model.device != self.device 内，是否应该继续执行 if self.use_distributed 条件的判断，继续给 self._wrapped_model 赋值，然后再返回 self._wrapped_model。

更正后的代码如下，跟以上代码的区别在于从 if self.use_distributed: 到 self._wrapped_model = self._model 这几行都往前缩进了一个 tab。

def model(self):
        # move model to device
        if self._model.device != self.device:
            self._model = self._model.to(self.device)
        # distributed training wrapper
        if self.use_distributed:
            if self._wrapped_model is None:
                self._wrapped_model = DDP(
                    self._model, device_ids=[self.config.run_cfg.gpu]
                )
        else:
            self._wrapped_model = self._model

        return self._wrapped_model

对项目不熟，不知猜测是否正确。

[jaycheney]

It works for me. THX.