Doc

docs/source/tutorial.rst

@@ -174,11 +174,18 @@ the dataset we generated earlier:
 
     engine.fit(dataset)
 
-Note that the training of the engine is not deterministic by default. This
-means that if you train your NLU twice on the same data you'll end up
-the output of the NLU can be different.
+Note that by default, the training of the engine is non-deterministic: if you
+train your NLU twice on the same data and test it on the same input, you'll get
+different outputs.
 
-To
+If you want to run training in a reproducible way you can pass a random seed to
+your engine:
+
+.. code-block:: python
+
+    seed = 42
+    engine = SnipsNLUEngine(config=CONFIG_EN, random_state=seed)
+    engine.fit(dataset)
 
 
 Parsing

snips_nlu/tests/integration_test.py

@@ -66,6 +66,7 @@ def test_nlu_engine_training_is_deterministic(self):
 
         ref_log_reg, ref_crfs = None, None
         for _ in range(num_runs):
+            # When
             engine = SnipsNLUEngine(random_state=random_state).fit(dataset)
             log_reg = _extract_log_reg(engine)
             crfs = _extract_crfs(engine)
@@ -74,6 +75,7 @@ def test_nlu_engine_training_is_deterministic(self):
                 ref_log_reg = log_reg
                 ref_crfs = crfs
             else:
+                # Then
                 self.assertDictEqual(ref_log_reg, log_reg)
                 self.assertDictEqual(ref_crfs, crfs)