7. Using Custom Trained Model

Custom trained models can be used with custom_model_path parameter and specifiying the path to the custom trained model.

Example

capfinder predict-cap-types \
    --bam_filepath /path/to/sorted.bam \
    --pod5_dir /path/to/pod5_dir \
    --output_dir /path/to/output_dir \
    --n_cpus 100 \
    --dtype float16 \
    --batch_size 256 \
    --custom_model_path /path/to/.keras/file/for/custom/trained/model \
    --no_plot_signal \
    --no-debug \
    --no-refresh-cache

Note

Before running the above command, please ensure that the new cap type is present in the cap map. Please read more about it in Extending cap mapping section of the documentation.