cryovit.models

Implementations of deep-learning automated segmentation models for CryoViT.

Functions

create_sam_model_from_weights(cfg, sam_dir)

Creates a SAM2 model from pre-trained weights specified in the config.

Classes

BaseModel(input_key, lr, weight_decay, ...)	Base model with configurable loss functions and metrics.
CryoVIT(**kwargs)	CryoVIT model implementation.
UNet3D(**kwargs)	UNet3D model implementation.
SAM2(sam_model, custom_kwargs, **kwargs)	Lightning wrapper over the SAM2 model.

class BaseModel(input_key: str, lr: float, weight_decay: float, losses: dict[str, Callable], metrics: dict[str, Callable], name: str = 'BaseModel', custom_kwargs: dict | None = None, **kwargs)

Bases: LightningModule, ABC

Base model with configurable loss functions and metrics.

__init__(input_key: str, lr: float, weight_decay: float, losses: dict[str, Callable], metrics: dict[str, Callable], name: str = 'BaseModel', custom_kwargs: dict | None = None, **kwargs) → None

Initializes the BaseModel with specified learning rate, weight decay, loss functions, and metrics.

Parameters:

• input_key (str) – Key to access input data in the batch.

• lr (float) – Learning rate for the optimizer.

• weight_decay (float) – Weight decay factor for AdamW optimizer.

• losses (dict[str, Callable]) – Dictionary of loss functions for training, validation, and testing.

• metrics (dict[str, Callable]) – Dictionary of metric functions for training, validation, and testing.

• name (str) – Name of the model.

• custom_kwargs (Optional[dict[str, Any]]) – Additional custom keyword arguments to set as attributes.

configure_optimizers() → Optimizer

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.

# 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

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.

on_before_optimizer_step(optimizer: Optimizer) → None

Logs gradient norms just before the optimizer updates weights.

log_stats(losses: dict[str, Tensor], prefix: Literal['train', 'val', 'test'], batch_size: int) → None

Logs computed loss and metric statistics for each training or validation step.

training_step(batch: BatchedTomogramData, batch_idx: int) → Tensor

Processes one batch during training, returning the total loss.

validation_step(batch: BatchedTomogramData, batch_idx: int) → Tensor

Processes one batch during validation, returning the total loss.

test_step(batch: BatchedTomogramData, batch_idx: int) → BatchedModelResult

Processes one batch during testing, captures predictions, and computes losses and metrics.

Parameters:

• batch (BatchedTomogramData) – The batch of data being processed.

• batch_idx (int) – Index of the batch.

Returns:

Contains test results and metrics, as well as file metadata for this batch.

Return type:

BatchedModelResult

predict_step(batch: BatchedTomogramData, batch_idx: int) → BatchedModelResult

Processes one batch during prediction, capturing model outputs along with file metadata.

Parameters:

• batch (BatchedTomogramData) – The batch of data being processed.

• batch_idx (int) – Index of the batch.

Returns:

Contains test results and metrics, as well as file metadata for this batch.

Return type:

BatchedModelResult

abstract forward()

Should be implemented in subclass.

class CryoVIT(**kwargs)

Bases: BaseModel

CryoVIT model implementation.

__init__(**kwargs) → None

Initializes the CryoVIT model with specific convolutional and synthesis blocks.

forward(batch: BatchedTomogramData) → Tensor

Forward pass for the CryoVIT model.

class UNet3D(**kwargs)

Bases: BaseModel

UNet3D model implementation.

__init__(**kwargs) → None

Initializes the UNet3D model with specific analysis and synthesis blocks.

forward_volume(x: Tensor) → Tensor

Memory optimized forward pass for the UNet3D model.

forward(batch: BatchedTomogramData) → Tensor

Should be implemented in subclass.

class SAM2(sam_model: SAM2Train, custom_kwargs, **kwargs)

Bases: BaseModel

Lightning wrapper over the SAM2 model.

__init__(sam_model: SAM2Train, custom_kwargs, **kwargs) → None

Initializes the SAM2 model with specific convolutional and synthesis blocks.

freeze_parameters()

Freezes all model parameters except for the prompt predictor and mask decoder.

configure_optimizers() → Optimizer

Configures the optimizer with separate learning rates for the prompt predictor and mask decoder.

forward(data: BatchedTomogramData) → dict[str, Tensor]

Should be implemented in subclass.

load_sam_state_dict(state_dict: dict[str, Tensor], strict: bool = False, assign: bool = True) → tuple

Override load_state_dict to handle loading of SAM2 weights.

compile() → None

Compiles the model image encoder for training.

create_sam_model_from_weights(cfg: BaseModel, sam_dir: Path) → SAM2

Creates a SAM2 model from pre-trained weights specified in the config.