normalization for relative jitter

flows/embedding_flow.py

@@ -17,6 +17,7 @@ from prefect.cache_policies import INPUTS, NO_CACHE
 from prefect_ray import RayTaskRunner
 
 import pandas as pd
+from sklearn.preprocessing import StandardScaler
 
 import embedding_utils as E
 from joblib import cpu_count
@@ -44,6 +45,10 @@ def generate_initial_frame_task(
     generator_func = E.dynamic_import(generator_path)
     data, labels = generator_func(**generator_kwargs)
 
+    # Per-feature z-score so jitter_scale has consistent meaning across
+    # generators and reducers see comparably-scaled inputs.
+    data = StandardScaler().fit_transform(data)
+
     df = pd.DataFrame(
         {
             "feature_0": data[:, 0],