Lightning implementation¶

To train a lightning model (more specifically a LitConvNet object), the train_lightning_model() function must be used. It is necessary to pass a DataLoader tuple with the training and validation DataLoaders, as well as a configuration dictionary.

The configuration dictionary contains information about the model itself (“model” entry), as well as the training (“training” entry). It also allow specification of keyword arguments to lightning’s lightning.pytorch.Trainer, (through the “training.trainer_kwargs”). An example of a commented config dictionary is given below:

{
   "seed": 74, # Seed passed to lightning's "seed_everything".
   "comment": "First lightning test!", # A comment to add to the model report
      # (this is done on_fit_start(), through training.TrainingClass.add_training()).
   "model": {
      "model_name": "fourlayer", # The name of the PyTorch model.
         # See LitConvNet's class attribute "model_dict" for possible values.
      "model_args": [], # *args passed to the PyTorch model initialization.
      "model_kwargs": { # **kwargs passed to the PyTorch model initialization.
         "dirname": "FourLayer_lightning_test"
      }
   },
   "training": {
      "loss_fn": "NLLLoss", # Name of the loss function to be used.
         # must be the name of a function defined in the torch.nn module.
         # We'll fetch it using getattr(torch.nn, config["training"]["loss_fn"])
      "params": "all", # See training.NetBase.get_params().
      "learning_rate": 1e-4,
      "trainer_kwargs": {
         # kwargs passed directly to lightning's Trainer class
         "max_epochs": 2,
         "limit_train_batches": 10,
         "limit_val_batches": 10,
         "accelerator": "cpu",
         "deterministic": True

      },
      """
      fit_kwargs: {
         'ckpt_path' can be specified for loading a checkpoint. See train_lightning_model's
         docstring.
         'ckpt_path': path_to_checkpoint
      }
      """
      "finish_training_kwargs": {
         # kwargs passed to training.TrainingClass.finish_training()
         # at the end of training.
         "remove_bool": True,
         "plot_accuracy": True
      }
   }
}

Note that some of the documentation for LitConvNet comes directly from Lightning’s documentation (specifically, for configure_optimizers(), forward(), training_step(), on_train_epoch_end(), validation_step(), on_validation_epoch_end()). Lightning functions which have custom documentation are noted with “[Custom documentation]” at the beginning of its docstring.

Note that, since LitConvNet is just a wrapper to a ConvNetBase, it also contains TrainingClass (dealt with in on_train_epoch_end(), on_validation_epoch_end() and on_fit_end()) and ReportManager classes. These classes can also be directly accessed by the property attributes manager and trainclass.

lightning_objects.train_lightning_model(config, loader_tuple)[source]¶

Function for training a Lightning model (more specifically a LitConvNet model).

There are two possible ways of loading a pre-existing model:

By loading the best_model.pkl: if the user specified the directory of the model’s report with config["training"]["trainer_kwargs"]["default_root_dir"]. Note that this is preferable as it will also load the TrainingClass and ReportManager classes in their latest state, instead of recreating them from scratch.
By loading a lightning checkpoint: if the user specified its path through the config entry config["training"]["fit_kwargs"]["ckpt_path"].

Note

If both options are specified in the config file, both the best model will be loaded and its checkpoints.

Parameters:

config – The config dictionary from which defined training and model hyperparameters.
loader_tuple – A tuple with two torch.utils.data.DataLoader objects: for training and validation respectively.

lightning_objects.get_trainer(**user_trainer_kwargs)[source]¶

Returns lightning’s trainer object according to user-defined kwargs. Some default kwargs are defined in this function.

Keyword Arguments:: **user_trainer_kwargs – Any kwargs that can be passed to Lightning’s lightning.pytorch.Trainer object. Only one default kwarg is specified: deterministic = True.

class lightning_objects.LitConvNet(config)[source]¶

Generic LightningModule wrapping Pytorch models defined in nets.

Class Attributes:

model_dict: dictionary relating names to the pytorch models defined in nets. Must

be updated if new models are added.

# Dictionary of available model names:
model_dict = {
    "twolayer":   TwoLayer,
    "threelayer": ThreeLayer,
    "fourlayer":  FourLayer,
    "tfcnn":      TFCNN,
}

Parameters:: config – a dictionary containing all information required for model initialization and training. See the example at the top of the page.

config¶: the config passed for __init__.

model_config¶: the "model" entry of the config.

training_config¶: the "training" entry of the config.

model¶: the actual PyTorch model (one of the models defined in nets).

loss¶: a torch.nn loss function, used for calculating the loss metrirc.

train_step_loss¶: a list containing each training step’s loss. Cleared at the end of the epoch.

valid_step_loss¶: a list containing each validation step’s loss. Cleared at the end of the epoch.

train_step_accur¶: a list containing each training step’s accuracy. Cleared at the end of the epoch.

valid_step_accur¶: a list containing each validation step’s accuracy. Cleared at the end of the epoch.

configure_optimizers()[source]¶

Choose what optimizers and learning-rate schedulers to use in your optimization. Normally you’d need one. But in the case of GANs or similar you might have multiple. Optimization with multiple optimizers only works in the manual optimization mode.

Returns:

Any of these 6 options.

Single optimizer.
List or Tuple of optimizers.
Two lists - The first list has multiple optimizers, and the second has multiple LR schedulers (or multiple lr_scheduler_config).
Dictionary, with an "optimizer" key, and (optionally) a "lr_scheduler" key whose value is a single LR scheduler or lr_scheduler_config.
None - Fit will run without any optimizer.

The lr_scheduler_config is a dictionary which contains the scheduler and its associated configuration. The default configuration is shown below.

lr_scheduler_config = {
    # REQUIRED: The scheduler instance
    "scheduler": lr_scheduler,
    # The unit of the scheduler's step size, could also be 'step'.
    # 'epoch' updates the scheduler on epoch end whereas 'step'
    # updates it after a optimizer update.
    "interval": "epoch",
    # How many epochs/steps should pass between calls to
    # `scheduler.step()`. 1 corresponds to updating the learning
    # rate after every epoch/step.
    "frequency": 1,
    # Metric to to monitor for schedulers like `ReduceLROnPlateau`
    "monitor": "val_loss",
    # If set to `True`, will enforce that the value specified 'monitor'
    # is available when the scheduler is updated, thus stopping
    # training if not found. If set to `False`, it will only produce a warning
    "strict": True,
    # If using the `LearningRateMonitor` callback to monitor the
    # learning rate progress, this keyword can be used to specify
    # a custom logged name
    "name": None,
}

When there are schedulers in which the .step() method is conditioned on a value, such as the torch.optim.lr_scheduler.ReduceLROnPlateau scheduler, Lightning requires that the lr_scheduler_config contains the keyword "monitor" set to the metric name that the scheduler should be conditioned on.

Metrics can be made available to monitor by simply logging it using self.log('metric_to_track', metric_val) in your LightningModule.

Note

Some things to know:

Lightning calls .backward() and .step() automatically in case of automatic optimization.
If a learning rate scheduler is specified in configure_optimizers() with key "interval" (default “epoch”) in the scheduler configuration, Lightning will call the scheduler’s .step() method automatically in case of automatic optimization.
If you use 16-bit precision (precision=16), Lightning will automatically handle the optimizer.
If you use torch.optim.LBFGS, Lightning handles the closure function automatically for you.
If you use multiple optimizers, you will have to switch to ‘manual optimization’ mode and step them yourself.
If you need to control how often the optimizer steps, override the optimizer_step() hook.

forward(x: Tensor) → Tensor[source]¶

Same as torch.nn.Module.forward().

Parameters:

*args – Whatever you decide to pass into the forward method.
**kwargs – Keyword arguments are also possible.

Returns:

Your model’s output

on_fit_start() → None[source]¶: [Custom documentation] Does some reporting with the model TrainClass add_training() method. If it is the model’s first fit, then a ModelReport.txt is also created.

training_step(batch: tuple[Tensor, Tensor], batch_idx: int) → Tensor[source]¶

Here you compute and return the training loss and some additional metrics for e.g. the progress bar or logger.

Parameters:

batch – The output of your data iterable, normally a DataLoader.
batch_idx – The index of this batch.
dataloader_idx – The index of the dataloader that produced this batch. (only if multiple dataloaders used)

Returns:

Tensor - The loss tensor
dict - A dictionary which can include any keys, but must include the key 'loss' in the case of automatic optimization.
None - In automatic optimization, this will skip to the next batch (but is not supported for multi-GPU, TPU, or DeepSpeed). For manual optimization, this has no special meaning, as returning the loss is not required.

In this step you’d normally do the forward pass and calculate the loss for a batch. You can also do fancier things like multiple forward passes or something model specific.

Example:

def training_step(self, batch, batch_idx):
    x, y, z = batch
    out = self.encoder(x)
    loss = self.loss(out, x)
    return loss

To use multiple optimizers, you can switch to ‘manual optimization’ and control their stepping:

def __init__(self):
    super().__init__()
    self.automatic_optimization = False


# Multiple optimizers (e.g.: GANs)
def training_step(self, batch, batch_idx):
    opt1, opt2 = self.optimizers()

    # do training_step with encoder
    ...
    opt1.step()
    # do training_step with decoder
    ...
    opt2.step()

Note

When accumulate_grad_batches > 1, the loss returned here will be automatically normalized by accumulate_grad_batches internally.

on_train_epoch_end() → None[source]¶

Called in the training loop at the very end of the epoch.

To access all batch outputs at the end of the epoch, you can cache step outputs as an attribute of the LightningModule and access them in this hook:

class MyLightningModule(L.LightningModule):
    def __init__(self):
        super().__init__()
        self.training_step_outputs = []

    def training_step(self):
        loss = ...
        self.training_step_outputs.append(loss)
        return loss

    def on_train_epoch_end(self):
        # do something with all training_step outputs, for example:
        epoch_mean = torch.stack(self.training_step_outputs).mean()
        self.log("training_epoch_mean", epoch_mean)
        # free up the memory
        self.training_step_outputs.clear()

validation_step(batch: tuple[Tensor, Tensor], batch_idx: int) → Tensor[source]¶

Operates on a single batch of data from the validation set. In this step you’d might generate examples or calculate anything of interest like accuracy.

Parameters:

batch – The output of your data iterable, normally a DataLoader.
batch_idx – The index of this batch.
dataloader_idx – The index of the dataloader that produced this batch. (only if multiple dataloaders used)

Returns:

Tensor - The loss tensor
dict - A dictionary. Can include any keys, but must include the key 'loss'.
None - Skip to the next batch.

# if you have one val dataloader:
def validation_step(self, batch, batch_idx): ...


# if you have multiple val dataloaders:
def validation_step(self, batch, batch_idx, dataloader_idx=0): ...

Examples:

# CASE 1: A single validation dataset
def validation_step(self, batch, batch_idx):
    x, y = batch

    # implement your own
    out = self(x)
    loss = self.loss(out, y)

    # log 6 example images
    # or generated text... or whatever
    sample_imgs = x[:6]
    grid = torchvision.utils.make_grid(sample_imgs)
    self.logger.experiment.add_image('example_images', grid, 0)

    # calculate acc
    labels_hat = torch.argmax(out, dim=1)
    val_acc = torch.sum(y == labels_hat).item() / (len(y) * 1.0)

    # log the outputs!
    self.log_dict({'val_loss': loss, 'val_acc': val_acc})

If you pass in multiple val dataloaders, validation_step() will have an additional argument. We recommend setting the default value of 0 so that you can quickly switch between single and multiple dataloaders.

# CASE 2: multiple validation dataloaders
def validation_step(self, batch, batch_idx, dataloader_idx=0):
    # dataloader_idx tells you which dataset this is.
    ...

Note

If you don’t need to validate you don’t need to implement this method.

Note

When the validation_step() is called, the model has been put in eval mode and PyTorch gradients have been disabled. At the end of validation, the model goes back to training mode and gradients are enabled.

on_validation_epoch_end() → None[source]¶: Called in the validation loop at the very end of the epoch.

on_fit_end() → None[source]¶: [Custom documentation] Calls the model TrainClass’ finish_training()

report() → str[source]¶: TODO: Could eventually add more to this. Currently just returns self.model.report()