API Reference¶

Auto-generated reference for leech's Python API. For command-line usage, see the CLI Reference.

Modules¶

Data handling¶

• Data Preparation -- BAM/POD5 reading, chunk extraction, splitting
• Features -- Move table parsing, dwell time and signal feature computation
• Dataset -- PyTorch Dataset classes for training

Models and training¶

• Models -- Neural network architectures (ConvLSTMDwell, TransformerDwell, etc.)
• CRF -- CTC-CRF sequence models: encoder, loss, and Viterbi decode
• Training -- Trainer class and training loop
• Evaluation -- Model evaluation and metrics
• Inference -- Inference engine and bundle inference
• Grid Search -- Hyperparameter optimization

Utilities¶

• Utilities -- Model loading, checkpoint handling, bundle creation

Quick examples¶

Extract dwell times from a BAM record¶

from leech.features import extract_move_table, compute_dwell_times

move_table = extract_move_table(alignment)
dwell_times = compute_dwell_times(move_table)

Load and run a model¶

from leech.model_loading import load_model_from_checkpoint

model, config = load_model_from_checkpoint("model_best.pt", device="cuda")

Load a model from a bundle¶

from leech.bundling import load_model_from_bundle

model, config = load_model_from_bundle("bundle.pt", pair="Ala_vs_Gly", device="cuda")