import flash raises ModuleNotFoundError

[rajsingh100]

Hi,
I installed lightning-flash 0.4.0 on my machine using pip install lightning-flash. Please see my environment below

pytorch version 1.8.0
lightning-flash 0.4.0
pytorch lightning version 1.4.4
conda 4.10.3
Python 3.6.13
OS - Windows.

I tried running this example from lightning-flash example https://github.com/PyTorchLightning/lightning-flash/blob/master/flash_examples/image_classification.py on my machine but I am getting this below error :

from flash.core.data.utils import download_data
ModuleNotFoundError: No module named 'flash.core'; 'flash' is not a package

Even using 'import flash' gives the below error

ModuleNotFoundError: No module named 'flash.core'; 'flash' is not a package

Tried searching in bug/fixes and on discussion forums but couldn't find any question related to this error.

Any help would be appreciated.

Thanks

[akihironitta]

It seems like Lightning Flash isn't installed properly.

Could you try reinstalling it with the following command and see if that works?

python -m pip install lightning-flash -U

[rajsingh100]

Hi akihironitta,

Sorry, I was away for a few days and I couldn't reply to your reply. I just saw your reply just know.
I tried using the above command. While installing it gave no errors.

Requirement already satisfied: lightning-flash in f:\anaconda3\envs\kaggle\lib\site-packages (0.4.0)
Requirement already satisfied: torch>=1.7 in f:\anaconda3\envs\kaggle\lib\site-packages (from lightning-flash) (1.9.0)
Requirement already satisfied: pyDeprecate in f:\anaconda3\envs\kaggle\lib\site-packages (from lightning-flash) (0.3.1)

This time I tried another example from examples https://github.com/PyTorchLightning/lightning-flash/blob/master/flash_examples/tabular_classification.py

Still, it's not working. Please note Pytorch is working, pytorch-lightning is also working

from flash.tabular import TabularClassificationData, TabularClassifier
ImportError: cannot import name 'TabularClassificationData'

[Done] exited with code=1 in 114.303 seconds

Thanks for your response. I will try again.

[ethanwharris]

Hi @rajsingh100 could you try uninstalling and reinstalling flash with the following:

pip uninstall lightning-flash
pip install --upgrade --pre lightning-flash

Hope that helps 😃

[rajsingh100]

Hi @ethanwharris ,

Looks like it worked to some extent. After uninstalling and installing with the commands you sent, it got passed the import section and started the downloading part but then it got stuck somewhere.

Please note I have 2 GPUS in my system.

<module>
    model = TabularClassifier.from_data(datamodule)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\flash\tabular\classification\model.py", line 116, in from_data
    model = cls(datamodule.num_features, datamodule.num_classes, datamodule.embedding_sizes, **kwargs)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\flash\core\utilities\imports.py", line 181, in wrapper
    f"Required dependencies not available. Please run: pip install {formatter(module_paths)}"
ModuleNotFoundError: Required dependencies not available. Please run: pip install 'lightning-flash[tabular]'

Now, I tried running pip install lightning-flash[tabular]. This solved the dependency problem, but then it got error in the multiprocessing step. The error is given below.

<module>
    trainer.fit(model, datamodule=datamodule)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\flash\core\trainer.py", line 128, in fit
    return super().fit(model, train_dataloader, val_dataloaders, datamodule)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 552, in fit
    self._run(model)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 917, in _run
    self._dispatch()
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 985, in _dispatch
    self.accelerator.start_training(self)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\accelerators\accelerator.py", line 92, in start_training
    self.training_type_plugin.start_training(trainer)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\plugins\training_type\ddp_spawn.py", line 158, in start_training
    mp.spawn(self.new_process, **self.mp_spawn_kwargs)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\torch\multiprocessing\spawn.py", line 230, in spawn
    return start_processes(fn, args, nprocs, join, daemon, start_method='spawn')
  File "F:\anaconda3\envs\kaggle\lib\site-packages\torch\multiprocessing\spawn.py", line 179, in start_processes
    process.start()
  File "F:\anaconda3\envs\kaggle\lib\multiprocessing\process.py", line 105, in start
    self._popen = self._Popen(self)
  File "F:\anaconda3\envs\kaggle\lib\multiprocessing\context.py", line 322, in _Popen
    return Popen(process_obj)
  File "F:\anaconda3\envs\kaggle\lib\multiprocessing\popen_spawn_win32.py", line 33, in __init__
    prep_data = spawn.get_preparation_data(process_obj._name)
  File "F:\anaconda3\envs\kaggle\lib\multiprocessing\spawn.py", line 143, in get_preparation_data
    _check_not_importing_main()
  File "F:\anaconda3\envs\kaggle\lib\multiprocessing\spawn.py", line 136, in _check_not_importing_main
    is not going to be frozen to produce an executable.''')
RuntimeError: 
        An attempt has been made to start a new process before the
        current process has finished its bootstrapping phase.

        This probably means that you are not using fork to start your
        child processes and you have forgotten to use the proper idiom
        in the main module:

            if __name__ == '__main__':
                freeze_support()
                ...

        The "freeze_support()" line can be omitted if the program
        is not going to be frozen to produce an executable.
    self._dispatch()
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 985, in _dispatch
    self.accelerator.start_training(self)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\accelerators\accelerator.py", line 92, in start_training
    self.training_type_plugin.start_training(trainer)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\plugins\training_type\ddp_spawn.py", line 158, in start_training
    mp.spawn(self.new_process, **self.mp_spawn_kwargs)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\torch\multiprocessing\spawn.py", line 230, in spawn
    return start_processes(fn, args, nprocs, join, daemon, start_method='spawn')
  File "F:\anaconda3\envs\kaggle\lib\site-packages\torch\multiprocessing\spawn.py", line 179, in start_processes
    process.start()
  File "F:\anaconda3\envs\kaggle\lib\multiprocessing\process.py", line 105, in start
    self._popen = self._Popen(self)
  File "F:\anaconda3\envs\kaggle\lib\multiprocessing\context.py", line 322, in _Popen
    return Popen(process_obj)
  File "F:\anaconda3\envs\kaggle\lib\multiprocessing\popen_spawn_win32.py", line 65, in __init__
    reduction.dump(process_obj, to_child)
  File "F:\anaconda3\envs\kaggle\lib\multiprocessing\reduction.py", line 60, in dump
    ForkingPickler(file, protocol).dump(obj)
BrokenPipeError: [Errno 32] Broken pipe

Broken pipe error.

Thanks , the dependency problem is gone. Now am looking at this error.

[ethanwharris]

Hi @rajsingh100 glad your first issue was fixed 😃 the second error can happen with multiprocessing, could you make sure that all of the code that runs anything in your main python file is inside a:

if __name__ == '__main__':

Hope that helps!

[rajsingh100]

Hi @ethanwharris ,

Yes, the Broken pipe error is gone with that change , now I getting the below error

RuntimeError: Distributed package doesn't have NCCL built in

I see that , the Trainer has requested for multi-gpus but have not specified the backend. Whether this backend has to be set by the programmer or in case the programmer misses it whether the framework can set the backend to some default.

I have run other multi gpus pytorch-lightning examples without any issues.
For your reference, below is the whole stack trace, sorry it's a little bit too long.

F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\trainer\connectors\accelerator_connector.py:747: UserWarning: You requested multiple GPUs but did not specify a backend, e.g. `Trainer(accelerator="dp"|"ddp"|"ddp2")`. Setting `accelerator="ddp_spawn"` for you.
  "You requested multiple GPUs but did not specify a backend, e.g."
GPU available: True, used: True
TPU available: False, using: 0 TPU cores
IPU available: False, using: 0 IPUs
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1]
initializing ddp: GLOBAL_RANK: 0, MEMBER: 1/2
initializing ddp: GLOBAL_RANK: 1, MEMBER: 2/2
Traceback (most recent call last):
  File "f:\vscode\python\kaggle\competitions\Google Landmark Recognition\Google-Landmark-Recognition\test.py", line 25, in <module>
    trainer.fit(model, datamodule=datamodule)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\flash\core\trainer.py", line 128, in fit
    return super().fit(model, train_dataloader, val_dataloaders, datamodule)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 552, in fit
    self._run(model)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 917, in _run
    self._dispatch()
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 985, in _dispatch
    self.accelerator.start_training(self)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\accelerators\accelerator.py", line 92, in start_training
    self.training_type_plugin.start_training(trainer)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\plugins\training_type\ddp_spawn.py", line 158, in start_training
    mp.spawn(self.new_process, **self.mp_spawn_kwargs)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\torch\multiprocessing\spawn.py", line 230, in spawn
    return start_processes(fn, args, nprocs, join, daemon, start_method='spawn')
  File "F:\anaconda3\envs\kaggle\lib\site-packages\torch\multiprocessing\spawn.py", line 188, in start_processes
    while not context.join():
  File "F:\anaconda3\envs\kaggle\lib\site-packages\torch\multiprocessing\spawn.py", line 150, in join
    raise ProcessRaisedException(msg, error_index, failed_process.pid)
torch.multiprocessing.spawn.ProcessRaisedException: 

-- Process 0 terminated with the following error:
Traceback (most recent call last):
  File "F:\anaconda3\envs\kaggle\lib\site-packages\torch\multiprocessing\spawn.py", line 59, in _wrap
    fn(i, *args)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\plugins\training_type\ddp_spawn.py", line 181, in new_process
    self.init_ddp_connection(self.global_rank, self.world_size)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\pytorch_lightning\plugins\training_type\ddp_spawn.py", line 264, in init_ddp_connection
    self.torch_distributed_backend, rank=global_rank, world_size=world_size
  File "F:\anaconda3\envs\kaggle\lib\site-packages\torch\distributed\distributed_c10d.py", line 531, in init_process_group
    timeout=timeout)
  File "F:\anaconda3\envs\kaggle\lib\site-packages\torch\distributed\distributed_c10d.py", line 625, in _new_process_group_helper
    raise RuntimeError("Distributed package doesn't have NCCL "
RuntimeError: Distributed package doesn't have NCCL built in

Thanks @ethanwharris @akihironitta