Initializes the multi-label sequence classification model. Args: config_or_model: Model configuration, pre-trained model

Initialize the RNA design model. Args: model (str): Model name or path for the pre-trained MLM model device:

Initialize the Metric class. Args: metric_func (callable, optional): Custom metric function to use ignore_y

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

Initialize the RegressionMetric class. Args: metric_func (callable, optional): Custom metric function to use

Initialize the RankingMetric class. Args: *args: Additional positional arguments passed to parent class **kw

Initialize the RNA structure cache. Args: cache_file (str, optional): Path to the cache file. If None, uses

Initialize the base trainer. Args: model (torch.nn.Module): The model to be trained train_dataset (Optional[

Initialize the HuggingFace training arguments wrapper. Args: *args: Positional arguments passed to the parent TrainingAr

Initialize the accelerate trainer. Args: model (torch.nn.Module): The model to be trained **kwargs: Addition

Initialize the native trainer. Args: model (torch.nn.Module): The model to be trained device (Optional[Union

Initializes the EpistasisExplainer. Args: model (Any): The model to explain, which must be compatible with the

Initialize the LIMEExplainer. Args: model: The trained model (classification or regression) to explain.

Initializes the TSNEExplainer. Args: model: A model object capable of generating embeddings, which should have

Initialize the ISMExplainer. Args: model: The trained model (classification or regression) to explain.

Initialize the SQUIDCombinatorialMutagenesis class. Args: max_order: the maximum order of mutations to generate. If -1,

Initializes the SequenceLogoExplainer. Args: model: The pre-trained model to be explained. This should be a model

Initializes the Visualization2DExplainer. Args: model (OmniModelForEmbedding): The model to explain. It must be an instance of

Initialize the OmniKmersTokenizer. Args: base_tokenizer: The base tokenizer for converting k-mers to token IDs

Initializes the single nucleotide tokenizer. Args: base_tokenizer: The underlying Hugging Face tokenizer. **

Initialize the OmniBPETokenizer. Args: base_tokenizer: The base BPE tokenizer **kwargs: Additional keyword a

Initializes the BenchHub.

Initialize the AutoTrain instance. Args: dataset (str): The name or path of the dataset to use for training.

Initializes the AutoBench instance. Args: benchmark (str): The name or path of the benchmark to use.

Initializes the AutoConfig. :param args: A dictionary or `argparse.Namespace` of parameters. :param kwargs:

Initialize dataset with BPP computation capability.

Initialize fusion module.

Initialize model with BPP fusion.

Initialize model with BPP fusion capability. Args: config_or_model: Path to pre-trained model tokeni

Returns an iterator over the dataset. Returns: iterator: An iterator over the dataset.

Return an iterator over the keys of the parameter dict.

Returns the number of samples in the dataset. Returns: int: The number of samples in the dataset.

Return the number of items in the parameter dict.

Check if the parameter dict is not equal to another object. :param other: The other object to compare with the parameter dict.

Return a detailed string representation of the configuration, including all parameters and the frequency of their access.

Sets a structure prediction in the cache.

Set the value of a key in the parameter dict. Also set the call frequency of the key to 0 and check if the updated configuration is v

String representation of the cache.

Return the same formatted representation as __repr__.

Get a string representation of the parameter dict. :return: A string representation of the parameter dict.

Compute metrics for HuggingFace trainer. This method adapts OmniGenome metrics to work with HuggingFace trainer. Args:

Downloads and extracts datasets from OmniGenBench Hub powered by HuggingFace. .. deprecated:: 0.3.0 Use ``omnigenbench.s

Load model state dictionary from temporary file.

Compute the longest base-pair distance using a stack over dot-bracket notation.

Returns the maximum length of labels in the dataset. Returns: int: The maximum length of labels.

Generate predictions for a batch using the model. Args: batch: Input batch Returns: Dict[str, torch

Generate predictions for a batch using the model. Args: batch: Input batch Returns: Dict[str, torch

Prepare a batch for model input. For HuggingFace trainer, batch preparation is handled internally. Args: batch:

Prepare a batch for model input. For accelerate trainer, the batch is already prepared by accelerator, so we just return it

Save model state dictionary to temporary file.

Set up HuggingFace training-specific components. This method initializes the HuggingFace Trainer with the model, datasets, a

Set up accelerate training-specific components. This method initializes the HuggingFace Accelerator with appropriate mixed p

Set up native training-specific components. This method initializes the device, mixed precision settings, and gradient scale

Align config fields with the actual checkpoint tensor shapes. Some training scripts mutate the transformer architecture (e.g., GLU width)

Train the model for one epoch. This method is not used in HuggingFace trainer as training is handled by the HuggingFace Trai

Train the model for one epoch using native PyTorch. Args: epoch (int): Current epoch number Returns:

Generate multiple augmented instances for a single sequence. Args: seq (str): Input genomic sequence k (int,

Get information about available models in the hub. This method queries the OmniGenome hub to retrieve information about avai

Encode sequences to token-level embeddings (last_hidden_state). Args: sequences (List[str]): Input DNA/RNA sequences for

Encode sequences to token-level embeddings (last_hidden_state). Args: sequences (List[str]): Input DNA/RNA sequences for

Extract attention scores from multiple genomic sequences in batches. This method provides efficient batch processing for attention extraction

Predicts on a batch of texts using the ensemble of predictors. This method processes multiple texts efficiently by combining predict

Available benchmark datasets

Calculate accuracy between two structures

Calculate precision, recall, F1 for base pairs

Checks the version compatibility between local and remote OmniGenome. Args: repo (str, optional): The repository URL to check. If No

Computes the metric. This method must be implemented by subclasses. Args: y_true: Ground truth labels. y_pre

Compute Base Pairing Probability (BPP) matrix using ViennaRNA. Args: sequence: RNA sequence string

This function traverses through nested dictionaries and lists, converting numerical values to PyTorch tensors while preserving the structure.

Factory for building baselines via OmniGenericBaseline. Example: model = create_baseline( tokenizer, backbone_typ

Decode input IDs using the base tokenizer. Args: input_ids: Input IDs to decode **kwargs: Additional keyword

Converts a list of token IDs back into a sequence. This method decodes token IDs back into genomic sequences using the under

Decode a sequence using the base BPE tokenizer. Args: sequence: Input sequence to decode (can be token IDs or tokens)

Download all benchmarks before running tests. This ensures benchmark data is available for testing.

Download a single file from HuggingFace Hub. Args: repo_id (str): HuggingFace repository identifier filename (str): Name of