shimmer_ssd.modules.domains.visual

View Source

  1from collections.abc import Mapping
  2from typing import Any
  3
  4import torch
  5from shimmer import LossOutput
  6from shimmer.modules.domain import DomainModule
  7from shimmer.modules.vae import VAE, gaussian_nll, kl_divergence_loss
  8from torch.nn.functional import mse_loss
  9from torch.optim.lr_scheduler import OneCycleLR
 10
 11from shimmer_ssd import LOGGER
 12from shimmer_ssd.modules.vae import RAEDecoder, RAEEncoder
 13
 14
 15class VisualDomainModule(DomainModule):
 16    def __init__(
 17        self,
 18        num_channels: int,
 19        latent_dim: int,
 20        ae_dim: int,
 21        beta: float = 1,
 22        optim_lr: float = 1e-3,
 23        optim_weight_decay: float = 0,
 24        scheduler_args: Mapping[str, Any] | None = None,
 25    ):
 26        """
 27        Visual domain module. This defines shimmer's `DomainModule` for the vision
 28        side with a VAE.
 29
 30        Args:
 31            num_channels (`int`): number of input channels (for RGB image, use 3)
 32            latent_dim (`int`): latent dimension of the vision domain
 33            ae_dim (`int`): internal auto-encoder dimension of the VAE
 34            beta (`float`): beta value if beta-VAE. (Defaults to 1.0)
 35            optim_lr (`float`): training learning rate
 36            optim_weight_decay (`float`): training weight decay
 37            scheduler_args (`Mapping[str, Any] | None`): Args for the scheduler.
 38        """
 39
 40        super().__init__(latent_dim)
 41        self.save_hyperparameters()
 42
 43        vae_encoder = RAEEncoder(num_channels, ae_dim, latent_dim, use_batchnorm=True)
 44        vae_decoder = RAEDecoder(num_channels, latent_dim, ae_dim)
 45        self.vae = VAE(vae_encoder, vae_decoder, beta)
 46        self.optim_lr = optim_lr
 47        self.optim_weight_decay = optim_weight_decay
 48        self.scheduler_args: dict[str, Any] = {
 49            "max_lr": optim_lr,
 50            "total_steps": 1,
 51        }
 52        self.scheduler_args.update(scheduler_args or {})
 53
 54    def compute_loss(
 55        self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any
 56    ) -> LossOutput:
 57        return LossOutput(mse_loss(pred, target, reduction="mean"))
 58
 59    def encode(self, x: torch.Tensor) -> torch.Tensor:
 60        return self.vae.encode(x)
 61
 62    def decode(self, z: torch.Tensor) -> torch.Tensor:
 63        out = self.vae.decode(z)
 64        return out
 65
 66    def forward(self, x: torch.Tensor) -> torch.Tensor:  # type: ignore
 67        return self.decode(self.encode(x))
 68
 69    def generic_step(
 70        self,
 71        x: torch.Tensor,
 72        mode: str = "train",
 73    ) -> torch.Tensor:
 74        (mean, logvar), reconstruction = self.vae(x)
 75
 76        reconstruction_loss = gaussian_nll(reconstruction, torch.tensor(0), x).sum()
 77
 78        kl_loss = kl_divergence_loss(mean, logvar)
 79        total_loss = reconstruction_loss + self.vae.beta * kl_loss
 80
 81        self.log(f"{mode}/reconstruction_loss", reconstruction_loss)
 82        self.log(f"{mode}/kl_loss", kl_loss)
 83        self.log(f"{mode}/loss", total_loss)
 84        return total_loss
 85
 86    def validation_step(  # type: ignore
 87        self,
 88        batch: Mapping[str, torch.Tensor],
 89        batch_idx: int,
 90    ) -> torch.Tensor:
 91        x = batch["v"]
 92        return self.generic_step(x, "val")
 93
 94    def training_step(  # type: ignore
 95        self,
 96        batch: Mapping[frozenset[str], Mapping[str, torch.Tensor]],
 97        batch_idx: int,
 98    ) -> torch.Tensor:
 99        x = batch[frozenset(["v"])]["v"]
100        return self.generic_step(x, "train")
101
102    def configure_optimizers(  # type: ignore
103        self,
104    ) -> dict[str, Any]:
105        optimizer = torch.optim.AdamW(
106            self.parameters(),
107            lr=self.optim_lr,
108            weight_decay=self.optim_weight_decay,
109        )
110        lr_scheduler = OneCycleLR(optimizer, **self.scheduler_args)
111
112        return {
113            "optimizer": optimizer,
114            "lr_scheduler": {
115                "scheduler": lr_scheduler,
116                "interval": "step",
117            },
118        }
119
120
121class VisualLatentDomainModule(DomainModule):
122    def __init__(self, visual_module: VisualDomainModule):
123        super().__init__(visual_module.latent_dim)
124        self.visual_module = visual_module
125
126    def encode(self, x: torch.Tensor) -> torch.Tensor:
127        return x
128
129    def decode(self, z: torch.Tensor) -> torch.Tensor:
130        return z
131
132    def compute_loss(
133        self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any
134    ) -> LossOutput:
135        return LossOutput(mse_loss(pred, target, reduction="mean"))
136
137    def decode_images(self, z: torch.Tensor) -> torch.Tensor:
138        LOGGER.debug(f"VisualLatentDomainModule.decode_images: z.shape = {z.size()}")
139        return self.visual_module.decode(z)
140
141
142class VisualLatentDomainWithUnpairedModule(DomainModule):
143    def __init__(self, visual_module: VisualDomainModule, coef_unpaired: float = 0.5):
144        super().__init__(visual_module.latent_dim + 1)
145
146        if coef_unpaired < 0 or coef_unpaired > 1:
147            raise ValueError("coef_unpaired should be in [0, 1]")
148
149        self.visual_module = visual_module
150        self.paired_dim = self.visual_module.latent_dim
151        self.coef_unpaired = coef_unpaired
152
153    def encode(self, x: torch.Tensor) -> torch.Tensor:
154        return x
155
156    def decode(self, z: torch.Tensor) -> torch.Tensor:
157        return z
158
159    def compute_loss(
160        self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any
161    ) -> LossOutput:
162        paired_loss = mse_loss(pred[:, : self.paired_dim], target[:, : self.paired_dim])
163        unpaired_loss = mse_loss(
164            pred[:, self.paired_dim :], target[:, self.paired_dim :]
165        )
166        total_loss = (
167            self.coef_unpaired * unpaired_loss + (1 - self.coef_unpaired) * paired_loss
168        )
169        return LossOutput(
170            loss=total_loss,
171            metrics={
172                "unpaired": unpaired_loss,
173                "paired": paired_loss,
174            },
175        )
176
177    def decode_images(self, z: torch.Tensor) -> torch.Tensor:
178        LOGGER.debug(f"VisualLatentDomainModule.decode_images: z.shape = {z.size()}")
179        return self.visual_module.decode(z[:, :-1])

class VisualDomainModule(shimmer.modules.domain.DomainModule): View Source

 16class VisualDomainModule(DomainModule):
 17    def __init__(
 18        self,
 19        num_channels: int,
 20        latent_dim: int,
 21        ae_dim: int,
 22        beta: float = 1,
 23        optim_lr: float = 1e-3,
 24        optim_weight_decay: float = 0,
 25        scheduler_args: Mapping[str, Any] | None = None,
 26    ):
 27        """
 28        Visual domain module. This defines shimmer's `DomainModule` for the vision
 29        side with a VAE.
 30
 31        Args:
 32            num_channels (`int`): number of input channels (for RGB image, use 3)
 33            latent_dim (`int`): latent dimension of the vision domain
 34            ae_dim (`int`): internal auto-encoder dimension of the VAE
 35            beta (`float`): beta value if beta-VAE. (Defaults to 1.0)
 36            optim_lr (`float`): training learning rate
 37            optim_weight_decay (`float`): training weight decay
 38            scheduler_args (`Mapping[str, Any] | None`): Args for the scheduler.
 39        """
 40
 41        super().__init__(latent_dim)
 42        self.save_hyperparameters()
 43
 44        vae_encoder = RAEEncoder(num_channels, ae_dim, latent_dim, use_batchnorm=True)
 45        vae_decoder = RAEDecoder(num_channels, latent_dim, ae_dim)
 46        self.vae = VAE(vae_encoder, vae_decoder, beta)
 47        self.optim_lr = optim_lr
 48        self.optim_weight_decay = optim_weight_decay
 49        self.scheduler_args: dict[str, Any] = {
 50            "max_lr": optim_lr,
 51            "total_steps": 1,
 52        }
 53        self.scheduler_args.update(scheduler_args or {})
 54
 55    def compute_loss(
 56        self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any
 57    ) -> LossOutput:
 58        return LossOutput(mse_loss(pred, target, reduction="mean"))
 59
 60    def encode(self, x: torch.Tensor) -> torch.Tensor:
 61        return self.vae.encode(x)
 62
 63    def decode(self, z: torch.Tensor) -> torch.Tensor:
 64        out = self.vae.decode(z)
 65        return out
 66
 67    def forward(self, x: torch.Tensor) -> torch.Tensor:  # type: ignore
 68        return self.decode(self.encode(x))
 69
 70    def generic_step(
 71        self,
 72        x: torch.Tensor,
 73        mode: str = "train",
 74    ) -> torch.Tensor:
 75        (mean, logvar), reconstruction = self.vae(x)
 76
 77        reconstruction_loss = gaussian_nll(reconstruction, torch.tensor(0), x).sum()
 78
 79        kl_loss = kl_divergence_loss(mean, logvar)
 80        total_loss = reconstruction_loss + self.vae.beta * kl_loss
 81
 82        self.log(f"{mode}/reconstruction_loss", reconstruction_loss)
 83        self.log(f"{mode}/kl_loss", kl_loss)
 84        self.log(f"{mode}/loss", total_loss)
 85        return total_loss
 86
 87    def validation_step(  # type: ignore
 88        self,
 89        batch: Mapping[str, torch.Tensor],
 90        batch_idx: int,
 91    ) -> torch.Tensor:
 92        x = batch["v"]
 93        return self.generic_step(x, "val")
 94
 95    def training_step(  # type: ignore
 96        self,
 97        batch: Mapping[frozenset[str], Mapping[str, torch.Tensor]],
 98        batch_idx: int,
 99    ) -> torch.Tensor:
100        x = batch[frozenset(["v"])]["v"]
101        return self.generic_step(x, "train")
102
103    def configure_optimizers(  # type: ignore
104        self,
105    ) -> dict[str, Any]:
106        optimizer = torch.optim.AdamW(
107            self.parameters(),
108            lr=self.optim_lr,
109            weight_decay=self.optim_weight_decay,
110        )
111        lr_scheduler = OneCycleLR(optimizer, **self.scheduler_args)
112
113        return {
114            "optimizer": optimizer,
115            "lr_scheduler": {
116                "scheduler": lr_scheduler,
117                "interval": "step",
118            },
119        }

Base class for a DomainModule that defines domain specific modules of the GW.

VisualDomainModule( num_channels: int, latent_dim: int, ae_dim: int, beta: float = 1, optim_lr: float = 0.001, optim_weight_decay: float = 0, scheduler_args: Mapping[str, typing.Any] | None = None) View Source

17    def __init__(
18        self,
19        num_channels: int,
20        latent_dim: int,
21        ae_dim: int,
22        beta: float = 1,
23        optim_lr: float = 1e-3,
24        optim_weight_decay: float = 0,
25        scheduler_args: Mapping[str, Any] | None = None,
26    ):
27        """
28        Visual domain module. This defines shimmer's `DomainModule` for the vision
29        side with a VAE.
30
31        Args:
32            num_channels (`int`): number of input channels (for RGB image, use 3)
33            latent_dim (`int`): latent dimension of the vision domain
34            ae_dim (`int`): internal auto-encoder dimension of the VAE
35            beta (`float`): beta value if beta-VAE. (Defaults to 1.0)
36            optim_lr (`float`): training learning rate
37            optim_weight_decay (`float`): training weight decay
38            scheduler_args (`Mapping[str, Any] | None`): Args for the scheduler.
39        """
40
41        super().__init__(latent_dim)
42        self.save_hyperparameters()
43
44        vae_encoder = RAEEncoder(num_channels, ae_dim, latent_dim, use_batchnorm=True)
45        vae_decoder = RAEDecoder(num_channels, latent_dim, ae_dim)
46        self.vae = VAE(vae_encoder, vae_decoder, beta)
47        self.optim_lr = optim_lr
48        self.optim_weight_decay = optim_weight_decay
49        self.scheduler_args: dict[str, Any] = {
50            "max_lr": optim_lr,
51            "total_steps": 1,
52        }
53        self.scheduler_args.update(scheduler_args or {})

Visual domain module. This defines shimmer's DomainModule for the vision side with a VAE.

Arguments:

num_channels (int): number of input channels (for RGB image, use 3)
latent_dim (int): latent dimension of the vision domain
ae_dim (int): internal auto-encoder dimension of the VAE
beta (float): beta value if beta-VAE. (Defaults to 1.0)
optim_lr (float): training learning rate
optim_weight_decay (float): training weight decay
scheduler_args (Mapping[str, Any] | None): Args for the scheduler.

vae

optim_lr

optim_weight_decay

scheduler_args: dict[str, typing.Any]

def compute_loss( self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any) -> shimmer.modules.domain.LossOutput: View Source

55    def compute_loss(
56        self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any
57    ) -> LossOutput:
58        return LossOutput(mse_loss(pred, target, reduction="mean"))

Generic loss computation the modality.

Arguments:

pred (torch.Tensor): prediction of the model
target (torch.Tensor): target tensor
raw_target (Any): raw data from the input

Results:

LossOutput | None: LossOuput with training loss and additional metrics. If None is returned, this loss will be ignored and will not participate in the total loss.

def encode(self, x: torch.Tensor) -> torch.Tensor: View Source

60    def encode(self, x: torch.Tensor) -> torch.Tensor:
61        return self.vae.encode(x)

Encode the domain data into a unimodal representation.

Arguments:

x (Any): data of the domain.

Returns:

torch.Tensor: a unimodal representation.

def decode(self, z: torch.Tensor) -> torch.Tensor: View Source

63    def decode(self, z: torch.Tensor) -> torch.Tensor:
64        out = self.vae.decode(z)
65        return out

Decode data from unimodal representation back to the domain data.

Arguments:

z (torch.Tensor): unimodal representation of the domain.

Returns:

Any: the original domain data.

def forward(self, x: torch.Tensor) -> torch.Tensor: View Source

67    def forward(self, x: torch.Tensor) -> torch.Tensor:  # type: ignore
68        return self.decode(self.encode(x))

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

Arguments:

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

Return:

Your model's output

def generic_step(self, x: torch.Tensor, mode: str = 'train') -> torch.Tensor: View Source

70    def generic_step(
71        self,
72        x: torch.Tensor,
73        mode: str = "train",
74    ) -> torch.Tensor:
75        (mean, logvar), reconstruction = self.vae(x)
76
77        reconstruction_loss = gaussian_nll(reconstruction, torch.tensor(0), x).sum()
78
79        kl_loss = kl_divergence_loss(mean, logvar)
80        total_loss = reconstruction_loss + self.vae.beta * kl_loss
81
82        self.log(f"{mode}/reconstruction_loss", reconstruction_loss)
83        self.log(f"{mode}/kl_loss", kl_loss)
84        self.log(f"{mode}/loss", total_loss)
85        return total_loss

def validation_step(self, batch: Mapping[str, torch.Tensor], batch_idx: int) -> torch.Tensor: View Source

87    def validation_step(  # type: ignore
88        self,
89        batch: Mapping[str, torch.Tensor],
90        batch_idx: int,
91    ) -> torch.Tensor:
92        x = batch["v"]
93        return self.generic_step(x, "val")

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.

Arguments:

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

Return:

~torch.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.

def training_step( self, batch: Mapping[frozenset[str], Mapping[str, torch.Tensor]], batch_idx: int) -> torch.Tensor: View Source

 95    def training_step(  # type: ignore
 96        self,
 97        batch: Mapping[frozenset[str], Mapping[str, torch.Tensor]],
 98        batch_idx: int,
 99    ) -> torch.Tensor:
100        x = batch[frozenset(["v"])]["v"]
101        return self.generic_step(x, "train")

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

Arguments:

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

Return:

~torch.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.

def configure_optimizers(self) -> dict[str, typing.Any]: View Source

103    def configure_optimizers(  # type: ignore
104        self,
105    ) -> dict[str, Any]:
106        optimizer = torch.optim.AdamW(
107            self.parameters(),
108            lr=self.optim_lr,
109            weight_decay=self.optim_weight_decay,
110        )
111        lr_scheduler = OneCycleLR(optimizer, **self.scheduler_args)
112
113        return {
114            "optimizer": optimizer,
115            "lr_scheduler": {
116                "scheduler": lr_scheduler,
117                "interval": "step",
118            },
119        }

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.

Return:

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.

.. testcode::

# 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 ~lightning.pytorch.core.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.

class VisualLatentDomainModule(shimmer.modules.domain.DomainModule): View Source

122class VisualLatentDomainModule(DomainModule):
123    def __init__(self, visual_module: VisualDomainModule):
124        super().__init__(visual_module.latent_dim)
125        self.visual_module = visual_module
126
127    def encode(self, x: torch.Tensor) -> torch.Tensor:
128        return x
129
130    def decode(self, z: torch.Tensor) -> torch.Tensor:
131        return z
132
133    def compute_loss(
134        self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any
135    ) -> LossOutput:
136        return LossOutput(mse_loss(pred, target, reduction="mean"))
137
138    def decode_images(self, z: torch.Tensor) -> torch.Tensor:
139        LOGGER.debug(f"VisualLatentDomainModule.decode_images: z.shape = {z.size()}")
140        return self.visual_module.decode(z)

Base class for a DomainModule that defines domain specific modules of the GW.

VisualLatentDomainModule(visual_module: VisualDomainModule) View Source

123    def __init__(self, visual_module: VisualDomainModule):
124        super().__init__(visual_module.latent_dim)
125        self.visual_module = visual_module

Initializes a DomainModule.

Arguments:

latent_dim (int): latent dimension of the unimodal module

visual_module

def encode(self, x: torch.Tensor) -> torch.Tensor: View Source

127    def encode(self, x: torch.Tensor) -> torch.Tensor:
128        return x

Encode the domain data into a unimodal representation.

Arguments:

x (Any): data of the domain.

Returns:

torch.Tensor: a unimodal representation.

def decode(self, z: torch.Tensor) -> torch.Tensor: View Source

130    def decode(self, z: torch.Tensor) -> torch.Tensor:
131        return z

Decode data from unimodal representation back to the domain data.

Arguments:

z (torch.Tensor): unimodal representation of the domain.

Returns:

Any: the original domain data.

def compute_loss( self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any) -> shimmer.modules.domain.LossOutput: View Source

133    def compute_loss(
134        self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any
135    ) -> LossOutput:
136        return LossOutput(mse_loss(pred, target, reduction="mean"))

Generic loss computation the modality.

Arguments:

pred (torch.Tensor): prediction of the model
target (torch.Tensor): target tensor
raw_target (Any): raw data from the input

Results:

LossOutput | None: LossOuput with training loss and additional metrics. If None is returned, this loss will be ignored and will not participate in the total loss.

def decode_images(self, z: torch.Tensor) -> torch.Tensor: View Source

138    def decode_images(self, z: torch.Tensor) -> torch.Tensor:
139        LOGGER.debug(f"VisualLatentDomainModule.decode_images: z.shape = {z.size()}")
140        return self.visual_module.decode(z)

class VisualLatentDomainWithUnpairedModule(shimmer.modules.domain.DomainModule): View Source

143class VisualLatentDomainWithUnpairedModule(DomainModule):
144    def __init__(self, visual_module: VisualDomainModule, coef_unpaired: float = 0.5):
145        super().__init__(visual_module.latent_dim + 1)
146
147        if coef_unpaired < 0 or coef_unpaired > 1:
148            raise ValueError("coef_unpaired should be in [0, 1]")
149
150        self.visual_module = visual_module
151        self.paired_dim = self.visual_module.latent_dim
152        self.coef_unpaired = coef_unpaired
153
154    def encode(self, x: torch.Tensor) -> torch.Tensor:
155        return x
156
157    def decode(self, z: torch.Tensor) -> torch.Tensor:
158        return z
159
160    def compute_loss(
161        self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any
162    ) -> LossOutput:
163        paired_loss = mse_loss(pred[:, : self.paired_dim], target[:, : self.paired_dim])
164        unpaired_loss = mse_loss(
165            pred[:, self.paired_dim :], target[:, self.paired_dim :]
166        )
167        total_loss = (
168            self.coef_unpaired * unpaired_loss + (1 - self.coef_unpaired) * paired_loss
169        )
170        return LossOutput(
171            loss=total_loss,
172            metrics={
173                "unpaired": unpaired_loss,
174                "paired": paired_loss,
175            },
176        )
177
178    def decode_images(self, z: torch.Tensor) -> torch.Tensor:
179        LOGGER.debug(f"VisualLatentDomainModule.decode_images: z.shape = {z.size()}")
180        return self.visual_module.decode(z[:, :-1])

Base class for a DomainModule that defines domain specific modules of the GW.

VisualLatentDomainWithUnpairedModule( visual_module: VisualDomainModule, coef_unpaired: float = 0.5) View Source

144    def __init__(self, visual_module: VisualDomainModule, coef_unpaired: float = 0.5):
145        super().__init__(visual_module.latent_dim + 1)
146
147        if coef_unpaired < 0 or coef_unpaired > 1:
148            raise ValueError("coef_unpaired should be in [0, 1]")
149
150        self.visual_module = visual_module
151        self.paired_dim = self.visual_module.latent_dim
152        self.coef_unpaired = coef_unpaired

Initializes a DomainModule.

Arguments:

latent_dim (int): latent dimension of the unimodal module

visual_module

paired_dim

coef_unpaired

def encode(self, x: torch.Tensor) -> torch.Tensor: View Source

154    def encode(self, x: torch.Tensor) -> torch.Tensor:
155        return x

Encode the domain data into a unimodal representation.

Arguments:

x (Any): data of the domain.

Returns:

torch.Tensor: a unimodal representation.

def decode(self, z: torch.Tensor) -> torch.Tensor: View Source

157    def decode(self, z: torch.Tensor) -> torch.Tensor:
158        return z

Decode data from unimodal representation back to the domain data.

Arguments:

z (torch.Tensor): unimodal representation of the domain.

Returns:

Any: the original domain data.

def compute_loss( self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any) -> shimmer.modules.domain.LossOutput: View Source

160    def compute_loss(
161        self, pred: torch.Tensor, target: torch.Tensor, raw_target: Any
162    ) -> LossOutput:
163        paired_loss = mse_loss(pred[:, : self.paired_dim], target[:, : self.paired_dim])
164        unpaired_loss = mse_loss(
165            pred[:, self.paired_dim :], target[:, self.paired_dim :]
166        )
167        total_loss = (
168            self.coef_unpaired * unpaired_loss + (1 - self.coef_unpaired) * paired_loss
169        )
170        return LossOutput(
171            loss=total_loss,
172            metrics={
173                "unpaired": unpaired_loss,
174                "paired": paired_loss,
175            },
176        )

Generic loss computation the modality.

Arguments:

pred (torch.Tensor): prediction of the model
target (torch.Tensor): target tensor
raw_target (Any): raw data from the input

Results:

LossOutput | None: LossOuput with training loss and additional metrics. If None is returned, this loss will be ignored and will not participate in the total loss.

def decode_images(self, z: torch.Tensor) -> torch.Tensor: View Source

178    def decode_images(self, z: torch.Tensor) -> torch.Tensor:
179        LOGGER.debug(f"VisualLatentDomainModule.decode_images: z.shape = {z.size()}")
180        return self.visual_module.decode(z[:, :-1])