Save the generated embeddings to a file. Args: embeddings (torch.Tensor): The embeddings to save output_path

Sets a custom loss function for the model. The loss function should be compatible with the model's output format. Args:

Set the token-id-to-one-hot projection matrix. Parameters ---------- weight : torch.Tensor Shape ``(vocab_size, 4

Calculate structure validity loss

Test the model on the test dataset. Returns: Dict[str, Any]: Dictionary containing test metrics

Compare model loading with different methods. This test documents the differences between git-lfs and HF Hub API.

Converts a sequence into a list of tokens. Must be implemented by subclasses. Args: sequence (str): The input sequence.

Converts a sequence into a list of individual nucleotide tokens. This method tokenizes genomic sequences by treating each nucleotide

Tokenize a sequence using the base BPE tokenizer. Args: sequence (str): Input sequence to tokenize **kwargs:

Print and return information about the current training setup. Returns: str: A string containing training setup informat

Updates the cache file on disk. This method saves the in-memory cache to disk. It only saves when the queue_num reaches 100

Validate RNA sequence

Creates a visual representation of an explanation. This method is optional and should be implemented by subclasses that can produce a

Creates an interactive 2D scatter plot of the embeddings. This method uses Plotly Express to generate a rich, interactive visualization

Visualizes the epistatic interaction matrix as an interactive heatmap. This method creates a detailed heatmap where each cell represents the

Visualizes an attribution matrix as an interactive heatmap. This method uses the `plotly` library to create a heatmap where the colo

Visualizes an attribution matrix as a sequence logo. This method uses the `logomaker` library to create a sequence logo. The

Provides a convenient shortcut to explain and visualize in one step. This allows the explainer instance to be called like a function, which

The main forward pass of the model, suitable for training loops. This method is the primary interface for model forward passes durin

Tokenizes inputs using the base tokenizer. This method provides a convenient interface for tokenization with sensible defaul

Allow calling the model directly.

Call the pipeline for inference. This method provides a convenient interface for running inference through the pipeline. It

A convenience method to explain and visualize in one step. Args: sequence (str): The sequence to explain. save_path (

Return an in silico MAVE based on the given sequence. Args: sequence: the sequence to mutate, string num_sim

Return an in silico MAVE based on the given sequence.

Return an in silico MAVE based on the given sequence. Args: sequence: the sequence to mutate, string num_sim

Generates and visualizes an explanation for a sequence. This is a convenience method that combines the `explain` and visualization

A convenience method to generate and visualize the explanation in one step. This method chains the `explain` and `visualize` calls, providing

Tokenize a sequence or list of sequences into tokenized inputs. This method processes the input sequence(s) by first converting them

Tokenizes sequences using single nucleotide tokenization. This method converts genomic sequences into tokenized inputs suitable

Tokenize a sequence using BPE tokenization. This method processes the input sequence using BPE tokenization, handles sequenc

Check if the given item is in the parameter dict. :param item: The item to check. :return: True if the item is in the paramet

Delete a key-value pair from the parameter dict and check if the updated configuration is valid. :param key: The key to delete from t

Check if the parameter dict is equal to another object. :param other: The other object to compare with the parameter dict. :r

Dynamically create ranking metric computation methods. This method intercepts attribute access and creates wrapper functions

Custom attribute getter that falls back to the base tokenizer if an attribute is not found on the wrapper. This method provi

Dynamically create metric computation methods. This method intercepts attribute access and creates wrapper functions for sci

Custom attribute getter that provides dynamic access to scikit-learn metrics. This method provides transparent access to all scikit-

Dynamically create regression metric computation methods. This method intercepts attribute access and creates wrapper functions

Returns a single data sample at the given index. Args: idx (int): The index of the sample. Returns:

Gets a cached structure prediction.

Get the value of a key from the parameter dict. :param item: The key to look for in the parameter dict. :return: The value of

Initializes the explainer with a model. Args: model (Any): The machine learning model (e.g., a PyTorch or TensorFlow model)

Initializes the genomic dataset with flexible input sources and preprocessing options. This method handles dataset loading from vari

Initializes the genomic foundation model with flexible input types. This method handles three initialization patterns: 1. *

Initializes the metric. Args: metric_func (callable, optional): A callable metric function from

Initializes the tokenizer wrapper with genomic-specific preprocessing options. Args: base_tokenizer: Underlying tokenize

Initializes the token classification dataset. Args: dataset_name_or_path (str or list): Path(s) to the dataset file(s).

Initialize the dataset for sequence classification. Args: dataset_name_or_path: Path to the data file or a list of paths

Initialize the dataset for sequence regression. Args: dataset_name_or_path: Path to the data file or a list of paths.

Initialize the dataset for multi-label classification. Args: dataset_name_or_path: Path to the data file or a list of pa

Initialize the LoRA-adapted model.

Initialize the VoteEnsemblePredictor. Args: predictors (List or dict): A list of checkpoints, or a dictionary of initial

Initialize the PipelineHub. Args: *args: Variable length argument list (currently unused). **kwargs: Arbitra

Initialize a Pipeline instance. Args: name (str): Name identifier for the pipeline. config_or_model (Union[s

Initialize the InteractingAttention module. Args: embed_size (int): Size of the embedding dimension num_head

Build the RNN backbone with LSTM.

Initialize convolutional stack and one-hot projection buffer.

Initialize the augmentation model. Args: config_or_model (str): Path or model name for loading the pre-trained model

Initialize the MLM model. Args: config_or_model: Model configuration or pre-trained model tokenizer: Tokeniz

Initialize the sequence-to-sequence model. Args: config_or_model: Model configuration or pre-trained model t

Initialize the sequence regression model. Args: config_or_model: Model configuration or pre-trained model to

Initialize the structural imputation model. Args: config_or_model: Model configuration or pre-trained model

Initialize the 2D structure-aware token regression model. Args: config_or_model: Model configuration or pre-trained mode

Initialize the 2D structure-aware sequence regression model. Args: config_or_model: Model configuration or pre-trained m

Initialize the matrix regression model. Args: config_or_model: Model configuration or pre-trained model toke

Initialize the matrix classification model. Args: config_or_model: Model configuration or pre-trained model

Initialize the BasicBlock. Args: inplanes (int): Number of input channels planes (int): Number of output cha

Initialize the Bottleneck block. Args: inplanes (int): Number of input channels planes (int): Number of outp

Initialize the embedding model. Args: config_or_model (str): Name or path of the pre-trained model to load t

Initializes the sequence classification model. Args: config_or_model: Model configuration, pre-trained model path, or mo