Pytorch Lightning Automatic Logging - AttributeError: 'NoneType' object has no attribute '_results'

Question

Unable to use Automatic Logging (self.log) when calling training_step() on Pytorch Lightning, what am I missing? Here is a minimal example:

import pytorch_lightning as pl
import torch
import torch.nn as nn
import torch.nn.functional as F

class LitModel(pl.LightningModule):
    def __init__(self):
        super().__init__()
        self.l1 = nn.Linear(100, 4)

    def forward(self, x):
        return torch.relu(self.l1(x.view(x.size(0), -1)))

    def training_step(self, batch, batch_idx):
        x, y = batch
        y_hat = self(x)
        loss = F.cross_entropy(y_hat, y.long())
        self.log("train_loss", loss) # <-- error
        return loss

    def configure_optimizers(self):
        return torch.optim.Adam(self.parameters(), lr=0.02)

pl_model = LitModel()
x = torch.rand((10,100))
y = torch.randint(0,4, size=(10,))
batch = (x,y)
loss = pl_model.training_step(batch, 0)

Error:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-34-b9419bfca30f> in <module>
     25 y = torch.randint(0,4, size=(10,))
     26 batch = (x,y)
---> 27 loss = pl_model.training_step(batch, 0)

<ipython-input-34-b9419bfca30f> in training_step(self, batch, batch_idx)
     14         y_hat = self(x)
     15         loss = F.cross_entropy(y_hat, y.long())
---> 16         self.log("train_loss", loss)
     17         return loss
     18 

D:\programs\anaconda3\lib\site-packages\pytorch_lightning\core\lightning.py in log(self, name, value, prog_bar, logger, on_step, on_epoch, reduce_fx, tbptt_reduce_fx, tbptt_pad_token, enable_graph, sync_dist, sync_dist_op, sync_dist_group, add_dataloader_idx, batch_size, metric_attribute, rank_zero_only)
    405         on_epoch = self.__auto_choose_log_on_epoch(on_epoch)
    406 
--> 407         results = self.trainer._results
    408         assert results is not None
    409         assert self._current_fx_name is not None

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

Seems like a bug was introduced in 1.4.2, I have the same issue. I will revert to a previous version and let you know — bluesmonk
– bluesmonk, Commented Aug 23, 2021 at 11:50
I think you must pass a logger object through a Trainer for self.log to work. I was thrown off by this too. Hope someone else can provide a better answer. — bluesmonk
– bluesmonk, Commented Aug 23, 2021 at 15:01

ayandas · Accepted Answer · 2021-08-25 17:45:03Z

2

This is NOT the correct usage of LightningModule class. You can't call a hook (namely .training_step()) manually and expect everything to work fine.

You need to setup a Trainer as suggested by PyTorch Lightning at the very start of its tutorial - it is a requirement. The functions (or hooks) that you define in a LightningModule merely tells Lightning "what to do" in a specific situation (in this case, at each training step). It is the Trainer that actually "orchestrates" the training by instantiating the necessary environment (including Logging functionality) and feeding it into the Lightning Module whenever needed.

So, do it the way Lightning suggests and it will work.

answered Aug 25, 2021 at 17:45

ayandas

2,3181 gold badge15 silver badges29 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Pytorch Lightning Automatic Logging - AttributeError: 'NoneType' object has no attribute '_results'

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related