See troubleshooting
and task-output
.
There are several ways to do this:
- Soft weight constraints:
--segtransition-weight-scale
will increase the strength of the soft length prior for longer segments.--prior-strength
also does this but empirically seems to have less of an effect than--segtransition-weight-scale
- Hard weight constraints
- The
--seg-table
option will allow you to specify a hard minimum segment length, as describedhere <hard-length-constraints>
. - Downsampling resolution
- The
See segment-duration-model
for model related methods.
- Train on smaller portion of the genome. Use the
--include-coords
option and supply a BED file. - Splitting up into smaller subsequences by reducing --split-sequences can also help.