[auto-optimize]: Speed up modisco motifs with byte-identical output

CHANGELOG

@@ -20,6 +20,12 @@ modisco
 	- GitHub Actions runs the full test suite, including the CLI tests, on
 	Python 3.9 through 3.13 for every pull request and push to master.
 
+	- Process scheduling and background compilation respect runtime thread
+	limits set with numba.set_num_threads(). Explicit worker budgets are retained.
+
+	- Require Python 3.9 or newer, matching the pinned Leiden/igraph dependencies
+	and the process APIs used by the optimized pipeline.
+
 
 
 Version 2.5.2

README.md

@@ -32,6 +32,9 @@ Related tools:
 
 ## Installation
 
+TF-MoDISco requires Python 3.9 or newer. Leiden is pinned to version 0.11.0
+to keep clustering results consistent; its igraph dependency requires Python 3.9 or newer.
+
 You can install TF-MoDISco using `pip install modisco`
 
 ## Running TF-MoDISco (test)

modiscolite/__init__.py

@@ -1,3 +1,5 @@
+import importlib
+
 from . import aggregator
 from . import affinitymat
 from . import extract_seqlets
@@ -7,10 +9,21 @@
 from . import util
 from . import tfmodisco
 from . import cluster
-from . import report
 from . import meme_writer
 from . import bed_writer
 from . import fasta_writer
-from . import descriptive_report
 
 __version__ = '2.5.2'
+
+
+# The report modules load matplotlib, pandas, logomaker, memelite and jinja2,
+# which only the report commands use, so they are imported on first access
+# (modiscolite.report, from modiscolite import descriptive_report).
+_LAZY = ('report', 'descriptive_report')
+
+
+def __getattr__(name):
+	if name in _LAZY:
+		return importlib.import_module('.' + name, __name__)
+	raise AttributeError('module {!r} has no attribute {!r}'.format(__name__,
+		name))