Attribution Models¶

Some time_interpret attributions use specific models, which are listed here:

Summary¶

`tint.attr.models.BLRRegression`(**kwargs)
`tint.attr.models.BLRRidge`(**kwargs)
`tint.attr.models.ExtremalMaskNet`(forward_func)	Extremal mask model as a Pytorch Lightning model.
`tint.attr.models.JointFeatureGeneratorNet`([...])	Conditional generator model to predict future observations as a Pytorch Lightning module.
`tint.attr.models.MaskNet`(forward_func[, ...])	Mask network as a Pytorch Lightning module.
`tint.attr.models.RetainNet`([dim_emb, ...])	Retain Network as a Pytorch Lightning module.
`tint.attr.models.scale_inputs`(input_ids, ...)	Creates a monotonic path between input_ids and ref_input_ids (the baseline).

Detailed classes and methods¶

class tint.attr.models.BLRRegression(**kwargs)[source]¶

class tint.attr.models.BLRRidge(**kwargs)[source]¶

class tint.attr.models.ExtremalMaskNN(forward_func: Callable, model: Optional[Module] = None, batch_size: int = 32)[source]¶

Extremal Mask NN model.

Parameters:

forward_func¶ (callable) – The forward function of the model or any modification of it.
model¶ (nnn.Module) – A model used to recreate the original predictions, in addition to the mask. Default to None
batch_size¶ (int) – Batch size of the model. Default to 32

References

forward(x: ~torch.Tensor, batch_idx, baselines, target, *additional_forward_args) -> (<class 'torch.Tensor'>, <class 'torch.Tensor'>)[source]¶

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class tint.attr.models.ExtremalMaskNet(forward_func: Callable, preservation_mode: bool = True, model: Optional[Module] = None, batch_size: int = 32, lambda_1: float = 1.0, lambda_2: float = 1.0, loss: Union[str, Callable] = 'mse', optim: str = 'adam', lr: float = 0.001, lr_scheduler: Optional[Union[dict, str]] = None, lr_scheduler_args: Optional[dict] = None, l2: float = 0.0)[source]¶

Extremal mask model as a Pytorch Lightning model.

Parameters:

forward_func¶ (callable) – The forward function of the model or any modification of it.
preservation_mode¶ (bool) – If True, uses the method in preservation mode. Otherwise, uses the deletion mode. Default to True
model¶ (nnn.Module) – A model used to recreate the original predictions, in addition to the mask. Default to None
batch_size¶ (int) – Batch size of the model. Default to 32
lambda_1¶ (float) – Weighting for the mask loss. Default to 1.
lambda_2¶ (float) – Weighting for the model output loss. Default to 1.
loss¶ (str, callable) – Which loss to use. Default to 'mse'
optim¶ (str) – Which optimizer to use. Default to 'adam'
lr¶ (float) – Learning rate. Default to 1e-3
lr_scheduler¶ (dict, str) – Learning rate scheduler. Either a dict (custom scheduler) or a string. Default to None
lr_scheduler_args¶ (dict) – Additional args for the scheduler. Default to None
l2¶ (float) – L2 regularisation. Default to 0.0

Examples

>>> import torch as th
>>> from tint.attr.models import ExtremalMaskNet
>>> from tint.models import MLP

>>> inputs = th.rand(8, 7, 5)
>>> data = th.rand(32, 7, 5)
>>> mlp = MLP([5, 3, 1])

>>> mask = ExtremalMaskNet(
...     forward_func=mlp,
...     optim="adam",
...     lr=0.01,
... )

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.

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.
Tuple of dictionaries as described above, with an optional "frequency" key.
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.

# The ReduceLROnPlateau scheduler requires a monitor
def configure_optimizers(self):
    optimizer = Adam(...)
    return {
        "optimizer": optimizer,
        "lr_scheduler": {
            "scheduler": ReduceLROnPlateau(optimizer, ...),
            "monitor": "metric_to_track",
            "frequency": "indicates how often the metric is updated"
            # If "monitor" references validation metrics, then "frequency" should be set to a
            # multiple of "trainer.check_val_every_n_epoch".
        },
    }


# In the case of two optimizers, only one using the ReduceLROnPlateau scheduler
def configure_optimizers(self):
    optimizer1 = Adam(...)
    optimizer2 = SGD(...)
    scheduler1 = ReduceLROnPlateau(optimizer1, ...)
    scheduler2 = LambdaLR(optimizer2, ...)
    return (
        {
            "optimizer": optimizer1,
            "lr_scheduler": {
                "scheduler": scheduler1,
                "monitor": "metric_to_track",
            },
        },
        {"optimizer": optimizer2, "lr_scheduler": scheduler2},
    )

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

Note

The frequency value specified in a dict along with the optimizer key is an int corresponding to the number of sequential batches optimized with the specific optimizer. It should be given to none or to all of the optimizers. There is a difference between passing multiple optimizers in a list, and passing multiple optimizers in dictionaries with a frequency of 1:

In the former case, all optimizers will operate on the given batch in each optimization step.

In the latter, only one optimizer will operate on the given batch at every step.

This is different from the frequency value specified in the lr_scheduler_config mentioned above.

def configure_optimizers(self):
    optimizer_one = torch.optim.SGD(self.model.parameters(), lr=0.01)
    optimizer_two = torch.optim.SGD(self.model.parameters(), lr=0.01)
    return [
        {"optimizer": optimizer_one, "frequency": 5},
        {"optimizer": optimizer_two, "frequency": 10},
    ]

In this example, the first optimizer will be used for the first 5 steps, the second optimizer for the next 10 steps and that cycle will continue. If an LR scheduler is specified for an optimizer using the lr_scheduler key in the above dict, the scheduler will only be updated when its optimizer is being used.

Examples:

# most cases. no learning rate scheduler
def configure_optimizers(self):
    return Adam(self.parameters(), lr=1e-3)

# multiple optimizer case (e.g.: GAN)
def configure_optimizers(self):
    gen_opt = Adam(self.model_gen.parameters(), lr=0.01)
    dis_opt = Adam(self.model_dis.parameters(), lr=0.02)
    return gen_opt, dis_opt

# example with learning rate schedulers
def configure_optimizers(self):
    gen_opt = Adam(self.model_gen.parameters(), lr=0.01)
    dis_opt = Adam(self.model_dis.parameters(), lr=0.02)
    dis_sch = CosineAnnealing(dis_opt, T_max=10)
    return [gen_opt, dis_opt], [dis_sch]

# example with step-based learning rate schedulers
# each optimizer has its own scheduler
def configure_optimizers(self):
    gen_opt = Adam(self.model_gen.parameters(), lr=0.01)
    dis_opt = Adam(self.model_dis.parameters(), lr=0.02)
    gen_sch = {
        'scheduler': ExponentialLR(gen_opt, 0.99),
        'interval': 'step'  # called after each training step
    }
    dis_sch = CosineAnnealing(dis_opt, T_max=10) # called every epoch
    return [gen_opt, dis_opt], [gen_sch, dis_sch]

# example with optimizer frequencies
# see training procedure in `Improved Training of Wasserstein GANs`, Algorithm 1
# https://arxiv.org/abs/1704.00028
def configure_optimizers(self):
    gen_opt = Adam(self.model_gen.parameters(), lr=0.01)
    dis_opt = Adam(self.model_dis.parameters(), lr=0.02)
    n_critic = 5
    return (
        {'optimizer': dis_opt, 'frequency': n_critic},
        {'optimizer': gen_opt, 'frequency': 1}
    )

Note

Some things to know:

Lightning calls .backward() and .step() on each optimizer as needed.
If 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 optimizers.
If you use multiple optimizers, training_step() will have an additional optimizer_idx parameter.
If you use torch.optim.LBFGS, Lightning handles the closure function automatically for you.
If you use multiple optimizers, gradients will be calculated only for the parameters of current optimizer at each training step.
If you need to control how often those optimizers step or override the default .step() schedule, override the optimizer_step() hook.

forward(*args, **kwargs) → 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

class tint.attr.models.JointFeatureGenerator(rnn_hidden_size: int = 100, dist_hidden_size: int = 10, latent_size: int = 100)[source]¶

Conditional generator model to predict future observations.

Parameters:

rnn_hidden_size¶ (int) – Size of hidden units for the recurrent structure. Default to 100
dist_hidden_size¶ (int) – Size of the distribution hidden units. Default to 10
latent_size¶ – Size of the latent distribution. Default to 100

References

A Recurrent Latent Variable Model for Sequential Data

forward(past: Tensor)[source]¶

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class tint.attr.models.JointFeatureGeneratorNet(rnn_hidden_size: int = 100, dist_hidden_size: int = 10, latent_size: int = 100, optim: str = 'adam', lr: float = 0.001, lr_scheduler: Optional[Union[dict, str]] = None, lr_scheduler_args: Optional[dict] = None, l2: float = 0.0)[source]¶

Conditional generator model to predict future observations as a Pytorch Lightning module.

Parameters:

rnn_hidden_size¶ (int) – Size of hidden units for the recurrent structure. Default to 100
dist_hidden_size¶ (int) – Size of the distribution hidden units. Default to 10
latent_size¶ – Size of the latent distribution. Default to 100
optim¶ (str) – Which optimizer to use. Default to 'adam'
lr¶ (float) – Learning rate. Default to 1e-3
lr_scheduler¶ (dict, str) – Learning rate scheduler. Either a dict (custom scheduler) or a string. Default to None
lr_scheduler_args¶ (dict) – Additional args for the scheduler. Default to None
l2¶ (float) – L2 regularisation. Default to 0.0

References

A Recurrent Latent Variable Model for Sequential Data

Examples

>>> from tint.attr.models import JointFeatureGeneratorNet

>>> generator = JointFeatureGeneratorNet(rnn_hidden_size=6)

test_step(batch, batch_idx)[source]¶

Operates on a single batch of data from the test set. In this step you’d normally generate examples or calculate anything of interest such as accuracy.

# the pseudocode for these calls
test_outs = []
for test_batch in test_data:
    out = test_step(test_batch)
    test_outs.append(out)
test_epoch_end(test_outs)

Parameters:

batch¶ – The output of your DataLoader.
batch_idx¶ – The index of this batch.
dataloader_id¶ – The index of the dataloader that produced this batch. (only if multiple test dataloaders used).

Returns:

Any of.

Any object or value

None - Testing will skip to the next batch

# if you have one test dataloader:
def test_step(self, batch, batch_idx):
    ...


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

Examples:

# CASE 1: A single test dataset
def test_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)
    test_acc = torch.sum(y == labels_hat).item() / (len(y) * 1.0)

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

If you pass in multiple test dataloaders, test_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 test dataloaders
def test_step(self, batch, batch_idx, dataloader_idx=0):
    # dataloader_idx tells you which dataset this is.
    ...

Note

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

Note

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

training_step(batch, batch_idx)[source]¶

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

Parameters:

batch¶ (Tensor | (Tensor, …) | [Tensor, …]) – The output of your DataLoader. A tensor, tuple or list.
batch_idx¶ (int) – Integer displaying index of this batch
optimizer_idx¶ (int) – When using multiple optimizers, this argument will also be present.
hiddens¶ (Any) – Passed in if truncated_bptt_steps > 0.

Returns:

Any of.

Tensor - The loss tensor
dict - A dictionary. Can include any keys, but must include the key 'loss'
None - Training will skip to the next batch. This is only for automatic optimization.
This is not supported for multi-GPU, TPU, IPU, or DeepSpeed.

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

If you define multiple optimizers, this step will be called with an additional optimizer_idx parameter.

# Multiple optimizers (e.g.: GANs)
def training_step(self, batch, batch_idx, optimizer_idx):
    if optimizer_idx == 0:
        # do training_step with encoder
        ...
    if optimizer_idx == 1:
        # do training_step with decoder
        ...

If you add truncated back propagation through time you will also get an additional argument with the hidden states of the previous step.

# Truncated back-propagation through time
def training_step(self, batch, batch_idx, hiddens):
    # hiddens are the hidden states from the previous truncated backprop step
    out, hiddens = self.lstm(data, hiddens)
    loss = ...
    return {"loss": loss, "hiddens": hiddens}

Note

The loss value shown in the progress bar is smoothed (averaged) over the last values, so it differs from the actual loss returned in train/validation step.

Note

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

validation_step(batch, batch_idx)[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.

# the pseudocode for these calls
val_outs = []
for val_batch in val_data:
    out = validation_step(val_batch)
    val_outs.append(out)
validation_epoch_end(val_outs)

Parameters:

batch¶ – The output of your DataLoader.
batch_idx¶ – The index of this batch.
dataloader_idx¶ – The index of the dataloader that produced this batch. (only if multiple val dataloaders used)

Returns:

Any object or value
None - Validation will skip to the next batch

# pseudocode of order
val_outs = []
for val_batch in val_data:
    out = validation_step(val_batch)
    if defined("validation_step_end"):
        out = validation_step_end(out)
    val_outs.append(out)
val_outs = validation_epoch_end(val_outs)

# 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.

class tint.attr.models.Mask(forward_func: Callable, perturbation: str = 'fade_moving_average', batch_size: int = 32, deletion_mode: bool = False, initial_mask_coef: float = 0.5, keep_ratio: Union[float, List[float]] = 0.5, size_reg_factor_init: float = 0.5, size_reg_factor_dilation: float = 100.0, time_reg_factor: float = 0.0, **kwargs)[source]¶

Mask network for DynaMask method.

Parameters:

forward_func¶ (Callable) – The function to get prediction from.
perturbation¶ (str) – Which perturbation to apply. Default to 'fade_moving_average'
deletion_mode¶ (bool) – True if the mask should identify the most impactful deletions. Default to False
initial_mask_coef¶ (float) – Which value to use to initialise the mask. Default to 0.5
keep_ratio¶ (float, list) – Fraction of elements in x that should be kept by the mask. Default to 0.5
size_reg_factor_init¶ (float) – Initial coefficient for the regulator part of the total loss. Default to 0.5
size_reg_factor_dilation¶ (float) – Ratio between the final and the initial size regulation factor. Default to 100
time_reg_factor¶ (float) – Regulation factor for the variation in time. Default to 0.0

References

Explaining Time Series Predictions with Dynamic Masks

forward(x: Tensor, batch_idx, *additional_forward_args) → Tensor[source]¶

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class tint.attr.models.MaskNet(forward_func: Callable, perturbation: str = 'fade_moving_average', batch_size: int = 32, deletion_mode: bool = False, initial_mask_coef: float = 0.5, keep_ratio: Union[float, List[float]] = 0.5, size_reg_factor_init: float = 0.5, size_reg_factor_dilation: float = 100.0, time_reg_factor: float = 0.0, loss: Union[str, Callable] = 'mse', optim: str = 'adam', lr: float = 0.001, lr_scheduler: Optional[Union[dict, str]] = None, lr_scheduler_args: Optional[dict] = None, l2: float = 0.0, **kwargs)[source]¶

Mask network as a Pytorch Lightning module.

Parameters:

forward_func¶ (Callable) – The function to get prediction from.
perturbation¶ (str) – Which perturbation to apply. Default to 'fade_moving_average'
deletion_mode¶ (bool) – True if the mask should identify the most impactful deletions. Default to False
initial_mask_coef¶ (float) – Which value to use to initialise the mask. Default to 0.5
keep_ratio¶ (float, list) – Fraction of elements in x that should be kept by the mask. Default to 0.5
size_reg_factor_init¶ (float) – Initial coefficient for the regulator part of the total loss. Default to 0.5
size_reg_factor_dilation¶ (float) – Ratio between the final and the initial size regulation factor. Default to 100
time_reg_factor¶ (float) – Regulation factor for the variation in time. Default to 0.0
loss¶ (str, callable) – Which loss to use. Default to 'mse'
optim¶ (str) – Which optimizer to use. Default to 'adam'
lr¶ (float) – Learning rate. Default to 1e-3
lr_scheduler¶ (dict, str) – Learning rate scheduler. Either a dict (custom scheduler) or a string. Default to None
lr_scheduler_args¶ (dict) – Additional args for the scheduler. Default to None
l2¶ (float) – L2 regularisation. Default to 0.0

References

Explaining Time Series Predictions with Dynamic Masks

Examples

>>> import numpy as np
>>> import torch as th
>>> from tint.attr.models import MaskNet
>>> from tint.models import MLP

>>> inputs = th.rand(8, 7, 5)
>>> data = th.rand(32, 7, 5)
>>> mlp = MLP([5, 3, 1])

>>> mask = MaskNet(
...     forward_func=mlp,
...     perturbation="gaussian_blur",
...     sigma_max=1,
...     keep_ratio=list(np.arange(0.25, 0.35, 0.01)),
...     size_reg_factor_init=0.1,
...     size_reg_factor_dilation=100,
...     time_reg_factor=1.0,
... )

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.

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.
Tuple of dictionaries as described above, with an optional "frequency" key.
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.

# The ReduceLROnPlateau scheduler requires a monitor
def configure_optimizers(self):
    optimizer = Adam(...)
    return {
        "optimizer": optimizer,
        "lr_scheduler": {
            "scheduler": ReduceLROnPlateau(optimizer, ...),
            "monitor": "metric_to_track",
            "frequency": "indicates how often the metric is updated"
            # If "monitor" references validation metrics, then "frequency" should be set to a
            # multiple of "trainer.check_val_every_n_epoch".
        },
    }


# In the case of two optimizers, only one using the ReduceLROnPlateau scheduler
def configure_optimizers(self):
    optimizer1 = Adam(...)
    optimizer2 = SGD(...)
    scheduler1 = ReduceLROnPlateau(optimizer1, ...)
    scheduler2 = LambdaLR(optimizer2, ...)
    return (
        {
            "optimizer": optimizer1,
            "lr_scheduler": {
                "scheduler": scheduler1,
                "monitor": "metric_to_track",
            },
        },
        {"optimizer": optimizer2, "lr_scheduler": scheduler2},
    )

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

Note

The frequency value specified in a dict along with the optimizer key is an int corresponding to the number of sequential batches optimized with the specific optimizer. It should be given to none or to all of the optimizers. There is a difference between passing multiple optimizers in a list, and passing multiple optimizers in dictionaries with a frequency of 1:

In the former case, all optimizers will operate on the given batch in each optimization step.

In the latter, only one optimizer will operate on the given batch at every step.

This is different from the frequency value specified in the lr_scheduler_config mentioned above.

def configure_optimizers(self):
    optimizer_one = torch.optim.SGD(self.model.parameters(), lr=0.01)
    optimizer_two = torch.optim.SGD(self.model.parameters(), lr=0.01)
    return [
        {"optimizer": optimizer_one, "frequency": 5},
        {"optimizer": optimizer_two, "frequency": 10},
    ]

In this example, the first optimizer will be used for the first 5 steps, the second optimizer for the next 10 steps and that cycle will continue. If an LR scheduler is specified for an optimizer using the lr_scheduler key in the above dict, the scheduler will only be updated when its optimizer is being used.

Examples:

# most cases. no learning rate scheduler
def configure_optimizers(self):
    return Adam(self.parameters(), lr=1e-3)

# multiple optimizer case (e.g.: GAN)
def configure_optimizers(self):
    gen_opt = Adam(self.model_gen.parameters(), lr=0.01)
    dis_opt = Adam(self.model_dis.parameters(), lr=0.02)
    return gen_opt, dis_opt

# example with learning rate schedulers
def configure_optimizers(self):
    gen_opt = Adam(self.model_gen.parameters(), lr=0.01)
    dis_opt = Adam(self.model_dis.parameters(), lr=0.02)
    dis_sch = CosineAnnealing(dis_opt, T_max=10)
    return [gen_opt, dis_opt], [dis_sch]

# example with step-based learning rate schedulers
# each optimizer has its own scheduler
def configure_optimizers(self):
    gen_opt = Adam(self.model_gen.parameters(), lr=0.01)
    dis_opt = Adam(self.model_dis.parameters(), lr=0.02)
    gen_sch = {
        'scheduler': ExponentialLR(gen_opt, 0.99),
        'interval': 'step'  # called after each training step
    }
    dis_sch = CosineAnnealing(dis_opt, T_max=10) # called every epoch
    return [gen_opt, dis_opt], [gen_sch, dis_sch]

# example with optimizer frequencies
# see training procedure in `Improved Training of Wasserstein GANs`, Algorithm 1
# https://arxiv.org/abs/1704.00028
def configure_optimizers(self):
    gen_opt = Adam(self.model_gen.parameters(), lr=0.01)
    dis_opt = Adam(self.model_dis.parameters(), lr=0.02)
    n_critic = 5
    return (
        {'optimizer': dis_opt, 'frequency': n_critic},
        {'optimizer': gen_opt, 'frequency': 1}
    )

Note

Some things to know:

Lightning calls .backward() and .step() on each optimizer as needed.
If 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 optimizers.
If you use multiple optimizers, training_step() will have an additional optimizer_idx parameter.
If you use torch.optim.LBFGS, Lightning handles the closure function automatically for you.
If you use multiple optimizers, gradients will be calculated only for the parameters of current optimizer at each training step.
If you need to control how often those optimizers step or override the default .step() schedule, override the optimizer_step() hook.

forward(*args, **kwargs) → 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_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, either:

Implement training_epoch_end in the LightningModule OR
Cache data across steps on the attribute(s) of the LightningModule and access them in this hook

predict_step(batch, batch_idx, dataloader_idx=0)[source]¶

Step function called during predict(). By default, it calls forward(). Override to add any processing logic.

The predict_step() is used to scale inference on multi-devices.

To prevent an OOM error, it is possible to use BasePredictionWriter callback to write the predictions to disk or database after each batch or on epoch end.

The BasePredictionWriter should be used while using a spawn based accelerator. This happens for Trainer(strategy="ddp_spawn") or training on 8 TPU cores with Trainer(accelerator="tpu", devices=8) as predictions won’t be returned.

Example

class MyModel(LightningModule):

    def predict_step(self, batch, batch_idx, dataloader_idx=0):
        return self(batch)

dm = ...
model = MyModel()
trainer = Trainer(accelerator="gpu", devices=2)
predictions = trainer.predict(model, dm)

Parameters:

batch¶ – Current batch.
batch_idx¶ – Index of current batch.
dataloader_idx¶ – Index of the current dataloader.

Returns:

Predicted output

class tint.attr.models.Retain(dim_emb: int = 128, dropout_input: float = 0.8, dropout_emb: float = 0.5, dim_alpha: int = 128, dim_beta: int = 128, dropout_context: float = 0.5, dim_output: int = 2, temporal_labels: bool = True)[source]¶

RETAIN network.

Parameters:

dim_emb¶ (int) – Dimension of the embedding. Default to 128
dropout_input¶ (float) – Dropout rate for the input. Default to 0.8
dropout_emb¶ (float) – Dropout of the embedding. Default to 0.5
dim_alpha¶ (int) – Hidden size of the alpha rnn. Default to 128
dim_beta¶ (int) – Hidden size of the beta rnn. Default to 128
dropout_context¶ (float) – Dropout rate of the context vector. Default to 0.5
dim_output¶ (int) – Size of the output. Default to 2
temporal_labels¶ (bool) – Whether to use temporal labels or static labels. Default to True

References

RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechanism

forward(x, lengths)[source]¶

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class tint.attr.models.RetainNet(dim_emb: int = 128, dropout_input: float = 0.8, dropout_emb: float = 0.5, dim_alpha: int = 128, dim_beta: int = 128, dropout_context: float = 0.5, dim_output: int = 2, temporal_labels: bool = True, loss: Union[str, Callable] = 'mse', optim: str = 'adam', lr: float = 0.001, lr_scheduler: Optional[Union[dict, str]] = None, lr_scheduler_args: Optional[dict] = None, l2: float = 0.0)[source]¶

Retain Network as a Pytorch Lightning module.

Parameters:

dim_emb¶ (int) – Dimension of the embedding. Default to 128
dropout_input¶ (float) – Dropout rate for the input. Default to 0.8
dropout_emb¶ (float) – Dropout of the embedding. Default to 0.5
dim_alpha¶ (int) – Hidden size of the alpha rnn. Default to 128
dim_beta¶ (int) – Hidden size of the beta rnn. Default to 128
dropout_context¶ (float) – Dropout rate of the context vector. Default to 0.5
dim_output¶ (int) – Size of the output. Default to 2
temporal_labels¶ (bool) – Whether to use temporal labels or static labels. Default to True
loss¶ (str, callable) – Which loss to use. Default to 'mse'
optim¶ (str) – Which optimizer to use. Default to 'adam'
lr¶ (float) – Learning rate. Default to 1e-3
lr_scheduler¶ (dict, str) – Learning rate scheduler. Either a dict (custom scheduler) or a string. Default to None
lr_scheduler_args¶ (dict) – Additional args for the scheduler. Default to None
l2¶ (float) – L2 regularisation. Default to 0.0

References

RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechanism

Examples

>>> from tint.attr.models import RetainNet

>>> retain = RetainNet(
...     dim_emb=128,
...     dropout_emb=0.4,
...     dim_alpha=8,
...     dim_beta=8,
...     dropout_context=0.4,
...     dim_output=2,
...     loss="cross_entropy",
... )

tint.attr.models.scale_inputs(input_ids, ref_input_ids, device, auxiliary_data, steps=30, factor=0, strategy='greedy')[source]¶

Creates a monotonic path between input_ids and ref_input_ids (the baseline). This path is only composed of data points, which have been ‘monotonized’. The strategy used to build the path is either 'greedy' or 'maxcount'.

Parameters:

input_ids¶ – The inputs.
ref_input_ids¶ – The baseline.
device¶ – Which device to use for the path.
auxiliary_data¶ – The knns previously computed.
steps¶ – Number of steps for the path. Default to 30
factor¶ – Up-scaling of the embeddings. Default to 0
strategy¶ – Strategy to build the path. Either 'greedy' or
'greedy' ('maxcount'. _sphinx_paramlinks_tint.attr.models.scale_inputs.Default to) –

Returns:

The monotonic path.

Return type:

torch.Tensor

References