cellmil.models.mil

class cellmil.models.mil.LitStandard(model: Module, optimizer: Optimizer, loss: Module = CrossEntropyLoss(), lr_scheduler: Optional[LRScheduler] = None, n_classes: int = 2)

Bases: LitGeneral

forward(x: Tensor)

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 cellmil.models.mil.LitGraphMIL(gnn: GNN, pooling_classifier: GlobalPooling_Classifier, optimizer_cls: type[torch.optim.optimizer.Optimizer], optimizer_kwargs: dict[str, Any], loss_fn: Module = CrossEntropyLoss(), scheduler_cls: Optional[type[torch.optim.lr_scheduler.LRScheduler]] = None, scheduler_kwargs: Optional[dict[str, Any]] = None, use_aem: bool = False, aem_weight_initial: float = 0.0001, aem_weight_final: float = 0.0, aem_annealing_epochs: int = 25, subsampling: float = 1.0, **kwargs: Any)

Bases: LightningModule

Lightning module for Graph-based Multiple Instance Learning.

This model is designed to work with torch_geometric DataLoader and requires: - batch_size=1 for MIL tasks - Data objects with batch.y containing graph labels - GNNMILDataset from cellmil.datamodels.datasets.gnn_mil_dataset

Example usage:

from torch_geometric.loader import DataLoader from cellmil.datamodels.datasets.gnn_mil_dataset import GNNMILDataset

dataset = GNNMILDataset(…) dataloader = DataLoader(dataset, batch_size=1, shuffle=True)

model = LitGraphMIL(gnn=…, pooling_classifier=…, …) trainer.fit(model, dataloader)

__init__(gnn: GNN, pooling_classifier: GlobalPooling_Classifier, optimizer_cls: type[torch.optim.optimizer.Optimizer], optimizer_kwargs: dict[str, Any], loss_fn: Module = CrossEntropyLoss(), scheduler_cls: Optional[type[torch.optim.lr_scheduler.LRScheduler]] = None, scheduler_kwargs: Optional[dict[str, Any]] = None, use_aem: bool = False, aem_weight_initial: float = 0.0001, aem_weight_final: float = 0.0, aem_annealing_epochs: int = 25, subsampling: float = 1.0, **kwargs: Any)

classmethod load_from_checkpoint(checkpoint_path: Union[str, Path, IO[bytes]], map_location: Optional[Union[device, str, int, Callable[[UntypedStorage, str], Optional[UntypedStorage]], dict[torch.device | str | int, torch.device | str | int]]] = None, hparams_file: Optional[Union[str, Path]] = None, strict: Optional[bool] = None, **kwargs: Any) → Self

Load a model from a checkpoint.

Parameters:

• checkpoint_path (str | Path | IO[bytes]) – Path to the checkpoint file or a file-like object.

• map_location (optional) – Device mapping for loading the model.

• hparams_file (optional) – Path to a YAML file containing hyperparameters.

• strict (optional) – Whether to strictly enforce that the keys in state_dict match the keys returned by the model’s state_dict function.

• **kwargs – Additional keyword arguments passed to the model’s constructor

Returns:

An instance of LitGraphMIL.

_subsample_graph(data: Data, subsampling: float) → Data

Sample subgraph using NeighborLoader to preserve local graph structure.

This method uses k-hop neighborhood sampling which preserves the local connectivity around seed nodes, providing better context for GNN message passing compared to random node sampling.

Note: This method is designed to work on CPU before GPU transfer when called from on_before_batch_transfer hook, saving GPU memory and transfer bandwidth.

Parameters:

• data (Data) – Input graph data (typically on CPU).

• subsampling (float) – Fraction of nodes to keep (0 < subsampling < 1.0) or absolute number of nodes (subsampling >= 1.0).

Returns:

Sampled subgraph with k-hop neighborhoods around seed nodes.

Return type:

Data

Note

This method requires either ‘pyg-lib’ or ‘torch-sparse’ to be installed. Install with: pip install pyg-lib torch-sparse -f https://data.pyg.org/whl/torch-{TORCH_VERSION}+{CUDA_VERSION}.html

forward(data: Data, **kwargs: Any)

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_before_batch_transfer(batch: Data, dataloader_idx: int) → Data

Hook called before batch is transferred to GPU. Performs subsampling on CPU to reduce memory usage and transfer overhead.

Parameters:

• batch (Data) – Input graph data on CPU.

• dataloader_idx (int) – Index of the dataloader.

Returns:

Potentially subsampled graph data (still on CPU).

Return type:

Data

training_step(batch: Data, batch_idx: int)

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.

validation_step(batch: Data, batch_idx: int)

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.
    x, y = batch

    # implement your own
    out = self(x)

    if dataloader_idx == 0:
        loss = self.loss0(out, y)
    else:
        loss = self.loss1(out, y)

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

    # log the outputs separately for each dataloader
    self.log_dict({f"val_loss_{dataloader_idx}": loss, f"val_acc_{dataloader_idx}": acc})

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.

test_step(batch: Data, batch_idx: int)

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.

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 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.
    x, y = batch

    # implement your own
    out = self(x)

    if dataloader_idx == 0:
        loss = self.loss0(out, y)
    else:
        loss = self.loss1(out, y)

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

    # log the outputs separately for each dataloader
    self.log_dict({f"test_loss_{dataloader_idx}": loss, f"test_acc_{dataloader_idx}": acc})

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.

on_train_epoch_end() → None

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()

on_validation_epoch_end()

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

on_test_epoch_end()

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

_flatten_and_log_metrics(computed: dict[str, torch.Tensor], prefix: str) → None

Convert metric dictionary produced by torchmetrics into a flat dict of scalar values and log it with self.log_dict.

• Vector/tensor metrics (e.g. per-class accuracy) are expanded into keys like {prefix}/class_{i}_acc.

• Scalar tensors are converted to floats.

• None values are converted to NaN to satisfy loggers that expect numeric scalars.

configure_optimizers()

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 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.

predict_step(batch: Data, batch_idx: int)

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.

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:

Predicted output (optional).

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)

static calculate_error(y_hat: Tensor, y: Tensor)

Classification error = 1 - accuracy.

get_attention_weights(data: Data) → dict[str, torch.Tensor]

Get attention weights from both GNN layers and pooling classifier.

This method delegates to the individual component classes for clean separation of concerns and better maintainability.

Parameters:

data (Data) – Input graph data.

Returns:

Dictionary containing attention weights:

• GNN attention weights (if available): ‘gnn_attention_layer_{i}’

• Pooling attention weights (if available): ‘pooling_attention’

Return type:

dict[str, torch.Tensor]

class cellmil.models.mil.LitCLAM(model: cellmil.models.mil.clam.CLAM_MB | cellmil.models.mil.clam.CLAM_SB, optimizer: Optimizer, loss_slide: Module = CrossEntropyLoss(), weight_loss_slide: float = 0.7, lr_scheduler: Optional[LRScheduler] = None, subsampling: float = 1.0, use_aem: bool = False, aem_weight_initial: float = 0.0001, aem_weight_final: float = 0.0, aem_annealing_epochs: int = 50)

Bases: LightningModule

static _is_gated_attention(model: cellmil.models.mil.clam.CLAM_MB | cellmil.models.mil.clam.CLAM_SB) → bool

Check if model uses gated attention.

static _get_size_args(model: cellmil.models.mil.clam.CLAM_MB | cellmil.models.mil.clam.CLAM_SB) → list[int]

Extract L and D parameters from attention network (size[1] and size[2]).

__init__(model: cellmil.models.mil.clam.CLAM_MB | cellmil.models.mil.clam.CLAM_SB, optimizer: Optimizer, loss_slide: Module = CrossEntropyLoss(), weight_loss_slide: float = 0.7, lr_scheduler: Optional[LRScheduler] = None, subsampling: float = 1.0, use_aem: bool = False, aem_weight_initial: float = 0.0001, aem_weight_final: float = 0.0, aem_annealing_epochs: int = 50)

classmethod load_from_checkpoint(checkpoint_path: Union[str, Path, IO[bytes]], map_location: Optional[Union[device, str, int, Callable[[UntypedStorage, str], Optional[UntypedStorage]], dict[torch.device | str | int, torch.device | str | int]]] = None, hparams_file: Optional[Union[str, Path]] = None, strict: Optional[bool] = None, **kwargs: Any) → Self

Load a LitCLAM model from a checkpoint file.

Parameters:

• checkpoint_path (str | Path | IO[bytes]) – Path to the checkpoint file.

• map_location – Device mapping for loading the model.

• hparams_file (str | Path | None) – Optional path to a YAML file with hyperparameters.

• strict (bool | None) – Whether to strictly enforce that the keys in state_dict match the keys returned by the model’s state_dict function.

• **kwargs – Additional keyword arguments.

Returns:

The loaded LitCLAM model.

Return type:

LitCLAM

forward(x: Tensor, label: Optional[Tensor] = None, instance_eval: bool = True)

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

training_step(batch: tuple[torch.Tensor, torch.Tensor], batch_idx: int)

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.

validation_step(batch: tuple[torch.Tensor, torch.Tensor], batch_idx: int)

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.
    x, y = batch

    # implement your own
    out = self(x)

    if dataloader_idx == 0:
        loss = self.loss0(out, y)
    else:
        loss = self.loss1(out, y)

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

    # log the outputs separately for each dataloader
    self.log_dict({f"val_loss_{dataloader_idx}": loss, f"val_acc_{dataloader_idx}": acc})

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.

test_step(batch: tuple[torch.Tensor, torch.Tensor], batch_idx: int)

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.

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 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.
    x, y = batch

    # implement your own
    out = self(x)

    if dataloader_idx == 0:
        loss = self.loss0(out, y)
    else:
        loss = self.loss1(out, y)

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

    # log the outputs separately for each dataloader
    self.log_dict({f"test_loss_{dataloader_idx}": loss, f"test_acc_{dataloader_idx}": acc})

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.

on_train_epoch_end() → None

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()

on_validation_epoch_end()

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

on_test_epoch_end()

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

_flatten_and_log_metrics(computed: dict[str, torch.Tensor], prefix: str) → None

Convert metric dictionary produced by torchmetrics into a flat dict of scalar values and log it with self.log_dict.

• Vector/tensor metrics (e.g. per-class accuracy) are expanded into keys like {prefix}/class_{i}_acc.

• Scalar tensors are converted to floats.

• None values are converted to NaN to satisfy loggers that expect numeric scalars.

configure_optimizers()

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 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.

predict_step(batch: tuple[torch.Tensor, torch.Tensor], batch_idx: int) → Any

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.

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:

Predicted output (optional).

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)

static calculate_error(Y_hat: Tensor, Y: Tensor)

Classification error = 1 - accuracy.

get_attention_weights(x: Tensor) → Tensor

Get attention weights for a bag of instances.

Parameters:

x (torch.Tensor) – Input tensor of shape [n_instances, feat_dim].

Returns:

Attention weights of shape [n_classes, n_instances].

Return type:

torch.Tensor

class cellmil.models.mil.LitAttentionDeepMIL(model: Module, optimizer: Optimizer, loss: Module = CrossEntropyLoss(), lr_scheduler: Optional[LRScheduler] = None, subsampling: float = 1.0, use_aem: bool = False, aem_weight_initial: float = 0.0001, aem_weight_final: float = 0.0, aem_annealing_epochs: int = 25)

Bases: LitGeneral

Lightning wrapper for AttentionDeepMIL model .

This class extends the base LitGeneral class to provide Lightning-specific functionality for the AttentionDeepMIL model..

Parameters:

• model (nn.Module) – The AttentionDeepMIL model instance.

• optimizer (torch.optim.Optimizer) – Optimizer for training.

• loss (nn.Module, optional) – Loss function. Defaults to nn.CrossEntropyLoss().

• lr_scheduler (LRScheduler | None, optional) – Learning rate scheduler. Defaults to None.

• subsampling (float, optional) – Fraction of instances to use during training (between 0 and 1). Defaults to 1.0 (no subsampling).

• use_aem (bool, optional) – Whether to use AEM regularization. Defaults to False.

• aem_weight_initial (float, optional) – Initial weight for AEM loss. Defaults to 0.001.

• aem_weight_final (float, optional) – Final weight for AEM loss after annealing. Defaults to 0.0.

• aem_annealing_epochs (int, optional) – Number of epochs to anneal AEM weight. Defaults to 25.

__init__(model: Module, optimizer: Optimizer, loss: Module = CrossEntropyLoss(), lr_scheduler: Optional[LRScheduler] = None, subsampling: float = 1.0, use_aem: bool = False, aem_weight_initial: float = 0.0001, aem_weight_final: float = 0.0, aem_annealing_epochs: int = 25)

classmethod load_from_checkpoint(checkpoint_path: Union[str, Path, IO[bytes]], map_location: Optional[Union[device, str, int, Callable[[UntypedStorage, str], Optional[UntypedStorage]], dict[torch.device | str | int, torch.device | str | int]]] = None, hparams_file: Optional[Union[str, Path]] = None, strict: Optional[bool] = None, **kwargs: Any) → Self

Load a model from a checkpoint.

Parameters:

• checkpoint_path (str | Path | IO[bytes]) – Path to the checkpoint file or a file-like object.

• map_location (optional) – Device mapping for loading the model.

• hparams_file (optional) – Path to a YAML file containing hyperparameters.

• strict (optional) – Whether to strictly enforce that the keys in state_dict match the keys returned by the model’s state_dict function.

• **kwargs – Additional keyword arguments passed to the model’s constructor.

Returns:

An instance of LitAttentionDeepMIL.

forward(x: Tensor) → Tensor

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

get_attention_weights(x: Tensor) → Tensor

Get attention weights for the input instances.

Parameters:

x (torch.Tensor) – Input tensor of shape [n_instances, feat_dim].

Returns:

Attention weights of shape [attention_branches, n_instances].

Return type:

torch.Tensor

Modules

cellmil.models.mil.attentiondeepmil
cellmil.models.mil.cellconv
cellmil.models.mil.clam
cellmil.models.mil.graphmil
cellmil.models.mil.head4type
cellmil.models.mil.histobistro
cellmil.models.mil.multifocus
cellmil.models.mil.standard
cellmil.models.mil.transmil
cellmil.models.mil.utils