decima.model package

Submodules

decima.model.decima_model module

class decima.model.decima_model.DecimaModel(n_tasks, replicate=0, mask=True, init_borzoi=True)

Bases: BaseModel

__annotations__ = {}

__init__(n_tasks, replicate=0, mask=True, init_borzoi=True)

Initialize internal Module state, shared by both nn.Module and ScriptModule.

decima.model.lightning module

The LightningModel class.

class decima.model.lightning.LightningModel(model_params, train_params={}, data_params={})

Bases: LightningModule

Wrapper for predictive sequence models

Parameters:

• model_params (dict) – Dictionary of parameters specifying model architecture

• train_params (dict) – Dictionary specifying training parameters

• data_params (dict) – Dictionary specifying parameters of the training data. This is empty by default and will be filled at the time of training.

__annotations__ = {}

__init__(model_params, train_params={}, data_params={})

add_transform(prediction_transform)

Add a prediction transform

Return type:

None

configure_optimizers()

Configure oprimizer for training

Return type:

None

count_params()

Number of gradient enabled parameters in the model

Return type:

int

format_input(x)

Extract the one-hot encoded sequence from the input

Return type:

Tensor

forward(x, logits=False)

Forward pass

Return type:

Tensor

get_task_idxs(tasks, key='name', invert=False)

Given a task name or metadata entry, get the task index If integers are provided, return them unchanged

Parameters:

• tasks (Union[int, str, List[int], List[str]]) – A string corresponding to a task name or metadata entry, or an integer indicating the index of a task, or a list of strings/integers

• key (str) – key to model.data_params[“tasks”] in which the relevant task data is stored. “name” will be used by default.

• invert (bool) – Get indices for all tasks except those listed in tasks

Return type:

Union[int, List[int]]

Returns:

The index or indices of the corresponding task(s) in the model’s output.

make_predict_loader(dataset, batch_size=None, num_workers=None, **kwargs)

Make dataloader for prediction

Return type:

Callable

make_test_loader(dataset, batch_size=None, num_workers=None)

Make dataloader for validation and testing

Return type:

Callable

make_train_loader(dataset, batch_size=None, num_workers=None)

Make dataloader for training

Return type:

Callable

on_save_checkpoint(checkpoint)

Called by Lightning when saving a checkpoint to give you a chance to store anything else you might want to save.

Parameters:

checkpoint (dict) – The full checkpoint dictionary before it gets dumped to a file. Implementations of this hook can insert additional data into this dictionary.

Return type:

None

Example:

def on_save_checkpoint(self, checkpoint):
    # 99% of use cases you don't need to implement this method
    checkpoint['something_cool_i_want_to_save'] = my_cool_pickable_object

Note

Lightning saves all aspects of training (epoch, global step, etc…) including amp scaling. There is no need for you to store anything about training.

on_test_epoch_end()

Calculate metrics for entire test set

Return type:

None

on_validation_epoch_end()

Calculate metrics for entire validation set

parse_logger()

Parses the name of the logger supplied in train_params.

Return type:

str

predict_on_dataset(dataset, devices=None, num_workers=1, batch_size=6, augment_aggfunc='mean', compare_func=None)

Predict for a dataset of sequences or variants

Parameters:

• dataset (Callable) – Dataset object that yields one-hot encoded sequences

• devices (Optional[int]) – Number of devices to use, e.g. machine has 4 gpu’s but only want to use 2 for predictions

• num_workers (int) – Number of workers for data loader

• batch_size (int) – Batch size for data loader

Returns:

Model predictions as a numpy array or dataframe

predict_step(batch, batch_idx, dataloader_idx=0)

Predict for a single batch of sequences or variants

Parameters:

• batch – Batch of sequences or variants

• batch_idx – Index of the batch

• dataloader_idx – Index of the dataloader

Return type:

Union[dict, Tensor]

Returns:

Dictionary containing predictions and warnings or a tensor of predictions

reset_transform()

Remove a prediction transform

Return type:

None

test_step(batch, batch_idx)

Calculate metrics after a single test step

Return type:

Tensor

train_on_dataset(train_dataset, val_dataset, checkpoint_path=None)

Train model and optionally log metrics to wandb.

Parameters:

• train_dataset (Dataset) – Dataset object that yields training examples

• val_dataset (Dataset) – Dataset object that yields training examples

• checkpoint_path (str) – Path to model checkpoint from which to resume training. The optimizer will be set to its checkpointed state.

Returns:

PyTorch Lightning Trainer

training_step(batch, batch_idx)

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

Parameters:

• batch (Tensor) – The output of your data iterable, normally a DataLoader.

• batch_idx (int) – The index of this batch.

• dataloader_idx – The index of the dataloader that produced this batch. (only if multiple dataloaders used)

Return type:

Tensor

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, batch_idx)

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 (Tensor) – The output of your data iterable, normally a DataLoader.

• batch_idx (int) – The index of this batch.

• dataloader_idx – The index of the dataloader that produced this batch. (only if multiple dataloaders used)

Return type:

Tensor

Returns:

• Tensor - The loss tensor

• dict - A dictionary. Can include any keys, but must include the key 'loss'.

• None - Skip to the next batch.

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


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

Examples:

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

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

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

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

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

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

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

Note

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

Note

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

decima.model.loss module

class decima.model.loss.TaskWisePoissonMultinomialLoss(total_weight=1, eps=1e-07, debug=False)

Bases: Module

__annotations__ = {}

__init__(total_weight=1, eps=1e-07, debug=False)

Initialize internal Module state, shared by both nn.Module and ScriptModule.

forward(input, target)

Define 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.

Return type:

Tensor

decima.model.metrics module

class decima.model.metrics.DiseaseLfcMSE(pairs, average=True)

Bases: Metric

__abstractmethods__ = frozenset({})

__annotations__ = {'__call__': 'Callable[..., Any]', '__jit_ignored_attributes__': 'ClassVar[list[str]]', '__jit_unused_properties__': 'ClassVar[list[str]]', '_backward_hooks': 'Dict[int, Callable]', '_backward_pre_hooks': 'Dict[int, Callable]', '_buffers': 'Dict[str, Optional[Tensor]]', '_cache': 'Optional[dict[str, Union[List[Tensor], Tensor]]]', '_compiled_call_impl': 'Optional[Callable]', '_defaults': 'dict[str, Union[list, Tensor]]', '_forward_hooks': 'Dict[int, Callable]', '_forward_hooks_always_called': 'Dict[int, bool]', '_forward_hooks_with_kwargs': 'Dict[int, bool]', '_forward_pre_hooks': 'Dict[int, Callable]', '_forward_pre_hooks_with_kwargs': 'Dict[int, bool]', '_is_full_backward_hook': 'Optional[bool]', '_load_state_dict_post_hooks': 'Dict[int, Callable]', '_load_state_dict_pre_hooks': 'Dict[int, Callable]', '_modules': "Dict[str, Optional['Module']]", '_non_persistent_buffers_set': 'Set[str]', '_parameters': 'Dict[str, Optional[Parameter]]', '_persistent': 'dict[str, bool]', '_reductions': 'dict[str, Union[str, Callable[..., Any], None]]', '_state_dict_hooks': 'Dict[int, Callable]', '_state_dict_pre_hooks': 'Dict[int, Callable]', '_version': 'int', 'call_super_init': 'bool', 'compute': 'Callable', 'dump_patches': 'bool', 'forward': 'Callable[..., Any]', 'full_state_update': 'Optional[bool]', 'higher_is_better': 'Optional[bool]', 'is_differentiable': 'Optional[bool]', 'plot_legend_name': 'Optional[str]', 'plot_lower_bound': 'Optional[float]', 'plot_upper_bound': 'Optional[float]', 'training': 'bool', 'update': 'Callable'}

__init__(pairs, average=True)

Initialize internal Module state, shared by both nn.Module and ScriptModule.

compute()

Override this method to compute the final metric value.

This method will automatically synchronize state variables when running in distributed backend.

Return type:

Tensor

reset()

Reset metric state variables to their default value.

Return type:

None

update(preds, target)

Override this method to update the state variables of your metric class.

Return type:

None

class decima.model.metrics.WarningCounter(warning_types=None, **kwargs)

Bases: Metric

A TorchMetric to count occurrences of different warning types, including a dedicated category for ‘unknown’ warnings.

__abstractmethods__ = frozenset({})

__annotations__ = {'__call__': 'Callable[..., Any]', '__jit_ignored_attributes__': 'ClassVar[list[str]]', '__jit_unused_properties__': 'ClassVar[list[str]]', '_backward_hooks': 'Dict[int, Callable]', '_backward_pre_hooks': 'Dict[int, Callable]', '_buffers': 'Dict[str, Optional[Tensor]]', '_cache': 'Optional[dict[str, Union[List[Tensor], Tensor]]]', '_compiled_call_impl': 'Optional[Callable]', '_defaults': 'dict[str, Union[list, Tensor]]', '_forward_hooks': 'Dict[int, Callable]', '_forward_hooks_always_called': 'Dict[int, bool]', '_forward_hooks_with_kwargs': 'Dict[int, bool]', '_forward_pre_hooks': 'Dict[int, Callable]', '_forward_pre_hooks_with_kwargs': 'Dict[int, bool]', '_is_full_backward_hook': 'Optional[bool]', '_load_state_dict_post_hooks': 'Dict[int, Callable]', '_load_state_dict_pre_hooks': 'Dict[int, Callable]', '_modules': "Dict[str, Optional['Module']]", '_non_persistent_buffers_set': 'Set[str]', '_parameters': 'Dict[str, Optional[Parameter]]', '_persistent': 'dict[str, bool]', '_reductions': 'dict[str, Union[str, Callable[..., Any], None]]', '_state_dict_hooks': 'Dict[int, Callable]', '_state_dict_pre_hooks': 'Dict[int, Callable]', '_version': 'int', 'call_super_init': 'bool', 'compute': 'Callable', 'dump_patches': 'bool', 'forward': 'Callable[..., Any]', 'full_state_update': 'Optional[bool]', 'higher_is_better': <class 'bool'>, 'is_differentiable': <class 'bool'>, 'plot_legend_name': 'Optional[str]', 'plot_lower_bound': 'Optional[float]', 'plot_upper_bound': 'Optional[float]', 'training': 'bool', 'update': 'Callable'}

__init__(warning_types=None, **kwargs)

Initialize internal Module state, shared by both nn.Module and ScriptModule.

compute()

Compute the final counts of all warning types.

Return type:

Counter

higher_is_better: bool = False

is_differentiable: bool = False

update(warnings)

Update the internal state with new warnings.

Parameters:

warnings (List[WarningType]) – A list of warning strings from a batch.

class decima.model.metrics.WarningType(value)

Bases: Enum

ALLELE_MISMATCH_WITH_REFERENCE_GENOME = 'allele_mismatch_with_reference_genome'

UNKNOWN = 'unknown'

Module contents