[plotting] adding capability for hl vars

plotting.py

@@ -17,8 +17,11 @@ class Multiplot:
     def __init__(self, n):
         try:
             self.cols, self.rows = [i for i in n]
+        except ValueError:
+            n = n[0]
+            self.cols = np.ceil(np.sqrt(n)).astype(int)
+            self.rows = np.ceil(n / self.cols).astype(int)
         except TypeError:
-            self.n = n
             self.cols = np.ceil(np.sqrt(n)).astype(int)
             self.rows = np.ceil(n / self.cols).astype(int)
 
@@ -99,12 +102,10 @@ def figure_activations(activations, class_names=None):
 
 
 def figure_history(history_csv_path):
-    # self.input()["training"]["history"].path
     pd.read_csv(history_csv_path).plot(subplots=True, figsize=(30, 30), layout=(7, 6))
     fig = plt.figure()
     fig.tight_layout()
     return fig
-    # plt.savefig(self.output()["plots"].path + "/history.pdf")
 
 
 def plot_histories(history_csv_path, path, cut=None, roll=1):
@@ -128,6 +129,14 @@ def plot_histories(history_csv_path, path, cut=None, roll=1):
         plt.close("all")
 
 
+def plot_dict(importance, path="tmp.pdf"):
+    fig = plt.figure(figsize=(8, len(importance) / 3))
+    plt.barh(*zip(*importance.items()))
+    fig.tight_layout()
+    plt.savefig(path)
+    plt.close("all")
+
+
 def figure_weights(w, y, class_names=None, relative=False):
     n_p = y.shape[1]
     bins = range(n_p)
@@ -238,7 +247,14 @@ def figure_roc_curve(
 
 
 def figure_inputs(
-    inps, truth, sample_weight=None, columns=None, class_names=None, signalvsbkg=False, bins=20
+    inps,
+    truth,
+    sample_weight=None,
+    columns=None,
+    class_names=None,
+    signalvsbkg=False,
+    bins=20,
+    overflow=1000,
 ):
     multiplot = Multiplot(inps.shape[1:][::-1])
     rows, cols = multiplot.lenghts()
@@ -246,6 +262,7 @@ def figure_inputs(
     fig, ax = plt.subplots(rows, cols, figsize=(size, size * rows / cols))
 
     inps = inps.reshape(inps.shape[0], -1)
+    inps = np.clip(inps, -overflow, overflow)
     order = np.argsort(-(sample_weight[:, None] * truth).sum(axis=0))
     class_names = np.array(class_names)[order]
     for feat, name in enumerate(columns):