diff --git a/src/loaders/prec_inn.py b/src/loaders/prec_inn.py
index 39ef27c..e5886e6 100644
--- a/src/loaders/prec_inn.py
+++ b/src/loaders/prec_inn.py
@@ -36,8 +36,8 @@ def load(params: dict) -> list[DiscriminatorData]:
     multiplicity_fake = np.sum(fake_momenta[:,:,0] != 0., axis=1)
 
     subsets = [
-        {"multiplicity": 3, "label": "$Z+1j$", "suffix": "z1j"},
-        {"multiplicity": 4, "label": "$Z+2j$", "suffix": "z2j"},
+        #{"multiplicity": 3, "label": "$Z+1j$", "suffix": "z1j"},
+        #{"multiplicity": 4, "label": "$Z+2j$", "suffix": "z2j"},
         {"multiplicity": 5, "label": "$Z+3j$", "suffix": "z3j"},
     ]
     datasets = []
@@ -61,12 +61,14 @@ def load(params: dict) -> list[DiscriminatorData]:
             "append_mass": params.get("append_mass", False),
             "append_delta_r": params.get("append_delta_r", False)
         }
-        pp_train_true = compute_preprocessing(train_true, **preproc_kwargs)
+        pp_train_all = compute_preprocessing(
+            np.concatenate((train_true, train_fake), axis=0), **preproc_kwargs
+        )
         datasets.append(DiscriminatorData(
             label = subset["label"],
             suffix = subset["suffix"],
-            dim = pp_train_true.shape[1],
-            train_true = pp_train_true,
+            dim = pp_train_all.shape[1],
+            train_true = compute_preprocessing(train_true, **preproc_kwargs),
             train_fake = compute_preprocessing(train_fake, **preproc_kwargs),
             test_true = compute_preprocessing(test_true, **preproc_kwargs),
             test_fake = compute_preprocessing(test_fake, **preproc_kwargs),
@@ -122,11 +124,12 @@ def compute_preprocessing(
         input_obs.append(mass)
 
     if append_delta_r:
+        drinv = lambda x: np.minimum(1/(x+1e-7), 20)
         if mult > 3:
-            input_obs.append(dr(obs.phi[:,2], obs.phi[:,3], obs.eta[:,2], obs.eta[:,3]))
+            input_obs.append(drinv(dr(obs.phi[:,2], obs.phi[:,3], obs.eta[:,2], obs.eta[:,3])))
         if mult > 4:
-            input_obs.append(dr(obs.phi[:,3], obs.phi[:,4], obs.eta[:,3], obs.eta[:,4]))
-            input_obs.append(dr(obs.phi[:,2], obs.phi[:,4], obs.eta[:,2], obs.eta[:,4]))
+            input_obs.append(drinv(dr(obs.phi[:,3], obs.phi[:,4], obs.eta[:,3], obs.eta[:,4])))
+            input_obs.append(drinv(dr(obs.phi[:,2], obs.phi[:,4], obs.eta[:,2], obs.eta[:,4])))
 
     data_preproc = np.stack(input_obs, axis=1)
     if "means" not in norm:
@@ -181,17 +184,17 @@ def compute_observables(true_data: np.ndarray, fake_data: np.ndarray) -> list[Ob
                 ),
                 unit = "GeV"
             ))
-        observables.append(Observable(
-            true_data = obs_two_true[(0,1)].m,
-            fake_data = obs_two_fake[(0,1)].m,
-            tex_label = r"M_{\mu\mu}",
-            bins = np.linspace(
-                np.quantile(obs_two_true[(0,1)].m, 0.005),
-                np.quantile(obs_two_true[(0,1)].m, 0.995),
-                50
-            ),
-            unit = "GeV"
-        ))
+    observables.append(Observable(
+        true_data = obs_two_true[(0,1)].m,
+        fake_data = obs_two_fake[(0,1)].m,
+        tex_label = r"M_{\mu\mu}",
+        bins = np.linspace(
+            np.quantile(obs_two_true[(0,1)].m, 0.005),
+            np.quantile(obs_two_true[(0,1)].m, 0.995),
+            50
+        ),
+        unit = "GeV"
+    ))
     for i, j in [(2,3), (2,4), (3,4)]:
         if i > mult-1 or j > mult-1:
             continue