AttributeError: module 'logging' has no attribute 'TensorBoardLogger'

[colllin]

🐛 Bug

Following the docs, I tried:

import pytorch_lightning as pl
logger = pl.logging.TensorBoardLogger(...)

But I receive an error:
AttributeError: module 'logging' has no attribute 'TensorBoardLogger'

To Reproduce

ubuntu@ip-172-31-41-72:~$ mkdir pltest

ubuntu@ip-172-31-41-72:~$ cd pltest/

ubuntu@ip-172-31-41-72:~/pltest$ pipenv --python 3.7
Creating a virtualenv for this project…
Pipfile: /home/ubuntu/pltest/Pipfile
Using /usr/bin/python3.7 (3.7.3) to create virtualenv…
⠋ Creating virtual environment...Already using interpreter /usr/bin/python3.7
Using base prefix '/usr'
New python executable in /home/ubuntu/.local/share/virtualenvs/pltest-rVdlPDKy/bin/python3.7
Also creating executable in /home/ubuntu/.local/share/virtualenvs/pltest-rVdlPDKy/bin/python
Installing setuptools, pip, wheel...
done.
✔ Successfully created virtual environment! 
Virtualenv location: /home/ubuntu/.local/share/virtualenvs/pltest-rVdlPDKy
Creating a Pipfile for this project…

ubuntu@ip-172-31-41-72:~/pltest$ pipenv install pytorch-lightning
Installing pytorch-lightning…
✔ Installation Succeeded 
Pipfile.lock not found, creating…
Locking [dev-packages] dependencies…
Locking [packages] dependencies…
✔ Success! 
Updated Pipfile.lock (d6238a)!
Installing dependencies from Pipfile.lock (d6238a)…
  🐍   ▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉ 34/34 — 00:00:39
To activate this project's virtualenv, run pipenv shell.
Alternatively, run a command inside the virtualenv with pipenv run.

ubuntu@ip-172-31-41-72:~/pltest$ pipenv graph | grep lightning
pytorch-lightning==0.6.0

ubuntu@ip-172-31-41-72:~/pltest$ pipenv run python
Python 3.7.3 (default, Mar 26 2019, 01:59:45) 
[GCC 5.4.0 20160609] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pytorch_lightning as pl
>>> pl.logging.TestTubeLogger
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: module 'logging' has no attribute 'TestTubeLogger'
>>> pl.logging.TensorBoardLogger
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: module 'logging' has no attribute 'TensorBoardLogger'
>>> pl.logging.__dict__.keys()
dict_keys(['__name__', '__doc__', '__package__', '__loader__', '__spec__', '__path__', '__file__', '__cached__', '__builtins__', 'sys', 'os', 'time', 'io', 'traceback', 'warnings', 'weakref', 'collections', 'Template', '__all__', 'threading', '__author__', '__status__', '__version__', '__date__', '_startTime', 'raiseExceptions', 'logThreads', 'logMultiprocessing', 'logProcesses', 'CRITICAL', 'FATAL', 'ERROR', 'WARNING', 'WARN', 'INFO', 'DEBUG', 'NOTSET', '_levelToName', '_nameToLevel', 'getLevelName', 'addLevelName', 'currentframe', '_srcfile', '_checkLevel', '_lock', '_acquireLock', '_releaseLock', '_at_fork_acquire_release_weakset', '_register_at_fork_acquire_release', '_at_fork_weak_calls', '_before_at_fork_weak_calls', '_after_at_fork_weak_calls', 'LogRecord', '_logRecordFactory', 'setLogRecordFactory', 'getLogRecordFactory', 'makeLogRecord', 'PercentStyle', 'StrFormatStyle', 'StringTemplateStyle', 'BASIC_FORMAT', '_STYLES', 'Formatter', '_defaultFormatter', 'BufferingFormatter', 'Filter', 'Filterer', '_handlers', '_handlerList', '_removeHandlerRef', '_addHandlerRef', 'Handler', 'StreamHandler', 'FileHandler', '_StderrHandler', '_defaultLastResort', 'lastResort', 'PlaceHolder', 'setLoggerClass', 'getLoggerClass', 'Manager', 'Logger', 'RootLogger', '_loggerClass', 'LoggerAdapter', 'root', 'basicConfig', 'getLogger', 'critical', 'fatal', 'error', 'exception', 'warning', 'warn', 'info', 'debug', 'log', 'disable', 'shutdown', 'atexit', 'NullHandler', '_warnings_showwarning', '_showwarning', 'captureWarnings'])
>>> from pytorch_lightning.logging.tensorboard import TensorBoardLogger
>>> TensorBoardLogger
<class 'pytorch_lightning.logging.tensorboard.TensorBoardLogger'>
>>> 

Code sample

Expected behavior

Environment

Collecting environment information...
PyTorch version: 1.3.1
Is debug build: No
CUDA used to build PyTorch: 10.1.243

OS: Ubuntu 16.04.6 LTS
GCC version: (Ubuntu 5.4.0-6ubuntu1~16.04.12) 5.4.0 20160609
CMake version: Could not collect

Python version: 3.7
Is CUDA available: No
CUDA runtime version: 10.0.130
GPU models and configuration: GPU 0: Tesla V100-SXM2-16GB
Nvidia driver version: 410.104
cuDNN version: Could not collect

Versions of relevant libraries:
[pip3] numpy==1.18.1
[pip3] pytorch-lightning==0.6.0
[pip3] torch==1.3.1
[pip3] torchvision==0.4.2
[conda] Could not collect

Additional context

[chutaklee]

Having the same issue but from pytorch_lightning.logging.tensorboard import TensorBoardLogger works for me.

[Borda]

sounds like an issue how the TensorBoardLogger is imported in pytorch_lightning.logging
see https://realpython.com/python-modules-packages/#python-packages

[juliengrenier]

Yeah there is something odd with the logging submodule. I am getting this:

In [17]: pytorch_lightning.__path__                              
Out[17]: ['/Users/j.grenier/.virtualenvs/pytorch_mnist/lib/python3.7/site-packages/pytorch_lightning']

In [18]: pytorch_lightning.logging.__path__                      
Out[18]: ['/usr/local/Cellar/python/3.7.6_1/Frameworks/Python.framework/Versions/3.7/lib/python3.7/logging']

Notice that pytorch_lighting.logging is pointing the standard python3.7 logging module.

[Borda]

wow, good point... 👍

[sudarshan85]

I'm having a similar problems like many others with the loggers. When I run my code:

trainer = pl.Trainer(train_percent_check=0.1)
trainer.fit(model)

I get the following error:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-10-1aee9f118f5f> in <module>
      1 trainer = pl.Trainer(train_percent_check=0.1)
----> 2 trainer.fit(model)

~/miniconda3/envs/adv/lib/python3.7/site-packages/pytorch_lightning/trainer/trainer.py in fit(self, model)
    755             self.optimizers, self.lr_schedulers = self.init_optimizers(model.configure_optimizers())
    756 
--> 757             self.run_pretrain_routine(model)
    758 
    759         # return 1 when finished

~/miniconda3/envs/adv/lib/python3.7/site-packages/pytorch_lightning/trainer/trainer.py in run_pretrain_routine(self, model)
    804             # save exp to get started
    805             if hasattr(ref_model, "hparams"):
--> 806                 self.logger.log_hyperparams(ref_model.hparams)
    807 
    808             self.logger.save()

~/miniconda3/envs/adv/lib/python3.7/site-packages/pytorch_lightning/logging/base.py in wrapped_fn(self, *args, **kwargs)
     12     def wrapped_fn(self, *args, **kwargs):
     13         if self.rank == 0:
---> 14             fn(self, *args, **kwargs)
     15 
     16     return wrapped_fn

~/miniconda3/envs/adv/lib/python3.7/site-packages/pytorch_lightning/logging/tensorboard.py in log_hyperparams(self, params)
     86         else:
     87             # `add_hparams` requires both - hparams and metric
---> 88             self.experiment.add_hparams(hparam_dict=params, metric_dict={})
     89         # some alternative should be added
     90         self.tags.update(params)

~/miniconda3/envs/adv/lib/python3.7/site-packages/torch/utils/tensorboard/writer.py in add_hparams(self, hparam_dict, metric_dict)
    298         if type(hparam_dict) is not dict or type(metric_dict) is not dict:
    299             raise TypeError('hparam_dict and metric_dict should be dictionary.')
--> 300         exp, ssi, sei = hparams(hparam_dict, metric_dict)
    301 
    302         logdir = os.path.join(

~/miniconda3/envs/adv/lib/python3.7/site-packages/torch/utils/tensorboard/summary.py in hparams(hparam_dict, metric_dict)
    154             ssi.hparams[k].number_value = v
    155             continue
--> 156         raise ValueError('value should be one of int, float, str, bool, or torch.Tensor')
    157 
    158     content = HParamsPluginData(session_start_info=ssi,

ValueError: value should be one of int, float, str, bool, or torch.Tensor

I thought I could create a test-tube logger, but the logging module does not have that. I have the latest release installed through

python -m pip install git+https://github.com/williamFalcon/pytorch-lightning.git@master --upgrade

I can't train my models. What are my options here?

[Borda]

We are working on a fix, check following PR #767

[sudarshan85]

Thank you @Borda. Since I'm not able to train my model without a logger, are there any other options other than waiting for the fix to come in?

[colllin]

Train it with a logger using advice above for alternative import statement?

…

On Thu, Jan 30, 2020 at 2:41 PM sudarshan ***@***.***> wrote: Thank you @Borda <https://github.com/Borda>. Since I'm not able to train my model without a logger, are there any other options other than waiting for the fix to come in? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#731>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABNMQ3EELDSTG6OO4LILPLRAMUPBANCNFSM4KKNN3TQ> .

[sudarshan85]

Train it with a logger using advice above for alternative import statement?
…
On Thu, Jan 30, 2020 at 2:41 PM sudarshan @.***> wrote: Thank you @Borda https://github.com/Borda. Since I'm not able to train my model without a logger, are there any other options other than waiting for the fix to come in? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#731?email_source=notifications&email_token=AABNMQ3B4PMPHUQVNACP4XTRAMUPBA5CNFSM4KKNN3T2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEKMIJ7A#issuecomment-580420860>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AABNMQ3EELDSTG6OO4LILPLRAMUPBANCNFSM4KKNN3TQ .

I'm assuming you are referring to @chutaklee reply about using from pytorch_lightning.logging.tensorboard import TensorBoardLogger. Unfortunately, this throws an error about tensorboard not being found. This is because as stated by @Borda earlier, the pl.logging module currently points to Python's logging module.

[colllin]

Really? Check out my initial description, under To Reproduce: I install pytorch-lightning (nothing else), and then am able to access TensorBoardLogger via from pytorch_lightning.logging.tensorboard import TensorBoardLogger. Maybe you're using an older version of pytorch which doesn't include tensorboard?

[Borda]

Well we have a short partial simple fix in #768

[david-waterworth]

The work around from @colllin didn't work for me, but from pytorch_lightning.logging import TensorBoardLogger does.

[rrezino]

Still getting the same error when logging.root.manager

[XYZ-916]

Having the same issue but from pytorch_lightning.logging.tensorboard import TensorBoardLogger works for me.

AttributeError: module 'logging' has no attribute 'tensorboard'