Data Preparation and Inference Pipeline

1. Generate k-mer profiles

   Use FastK or Draven to produce per-read, per-position k-mer profiles.
   The output should be .TSV files containing the following columns:

   • k-mer multiplicity
   • k-mer hash

2. Assign ground truth labels

   Run scripts/data_preprocess.py to assign ground-truth labels to each position in the generated profiles.

3. Create training batches

   Use scripts/create_batches.py to convert the labeled data into batched datasets suitable for model training or inference.

4. Run inference

   Finally, run batched_inference.py to evaluate the model and measure its performance on the prepared batches.