Merge 8e091ef into 7767df5

bothub/api/serializers/repository.py

@@ -26,6 +26,7 @@ class Meta:
             'name',
             'slug',
             'language',
+            'use_language_model_featurizer',
             'categories',
             'description',
             'is_private',
@@ -65,6 +66,7 @@ class Meta:
             'slug',
             'language',
             'available_languages',
+            'use_language_model_featurizer',
             'categories',
             'categories_list',
             'description',

bothub/common/migrations/0023_repository_use_language_model_featurizer.py

@@ -0,0 +1,18 @@
+# Generated by Django 2.0.6 on 2018-10-02 17:47
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('common', '0022_repositoryupdate_training_log'),
+    ]
+
+    operations = [
+        migrations.AddField(
+            model_name='repository',
+            name='use_language_model_featurizer',
+            field=models.BooleanField(default=True, help_text='You can use language featurizer to get words similarity. You need less examples to create a great bot.', verbose_name='Use language model featurizer'),
+        ),
+    ]

bothub/common/migrations/0024_repositoryupdate_use_language_model_featurizer.py

@@ -0,0 +1,18 @@
+# Generated by Django 2.0.6 on 2018-10-02 17:54
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('common', '0023_repository_use_language_model_featurizer'),
+    ]
+
+    operations = [
+        migrations.AddField(
+            model_name='repositoryupdate',
+            name='use_language_model_featurizer',
+            field=models.BooleanField(default=True),
+        ),
+    ]

bothub/common/models.py

@@ -99,6 +99,12 @@ class Meta:
         validators=[
             languages.validate_language,
         ])
+    use_language_model_featurizer = models.BooleanField(
+        _('Use language model featurizer'),
+        help_text=_('You can use language featurizer to get words ' +
+                    'similarity. You need less examples to create a great ' +
+                    'bot.'),
+        default=True)
     categories = models.ManyToManyField(
         RepositoryCategory,
         help_text=CATEGORIES_HELP_TEXT)
@@ -293,7 +299,8 @@ def current_update(self, language=None):
         language = language or self.language
         repository_update, created = self.updates.get_or_create(
             language=language,
-            training_started_at=None)
+            training_started_at=None,
+            use_language_model_featurizer=self.use_language_model_featurizer)
         return repository_update
 
     def last_trained_update(self, language=None):
@@ -337,6 +344,7 @@ class Meta:
         validators=[
             languages.validate_language,
         ])
+    use_language_model_featurizer = models.BooleanField(default=True)
     created_at = models.DateTimeField(
         _('created at'),
         auto_now_add=True)
@@ -434,6 +442,12 @@ def requirements_to_train(self):
 
     @property
     def ready_for_train(self):
+        last_trained_update = self.repository.last_trained_update(
+            language=self.language)
+        if last_trained_update:
+            if last_trained_update.use_language_model_featurizer is not \
+               self.repository.use_language_model_featurizer:
+                return True
         return len(self.requirements_to_train) is 0
 
     def validate_init_train(self, by=None):
@@ -450,10 +464,13 @@ def start_training(self, by):
         self.validate_init_train(by)
         self.by = by
         self.training_started_at = timezone.now()
+        self.use_language_model_featurizer = self.repository \
+            .use_language_model_featurizer
         self.save(
             update_fields=[
                 'by',
                 'training_started_at',
+                'use_language_model_featurizer',
             ])
 
     def save_training(self, bot_data):

bothub/common/tests.py

@@ -1045,3 +1045,48 @@ def test_set_label_to_none(self):
         name_entity.set_label(None)
 
         self.assertIsNone(name_entity.label)
+
+
+class UseLanguageModelFeaturizerTestCase(TestCase):
+    def setUp(self):
+        self.language = languages.LANGUAGE_EN
+
+        self.owner = User.objects.create_user('owner@user.com', 'user')
+
+        self.repository = Repository.objects.create(
+            owner=self.owner,
+            name='Test',
+            slug='test',
+            language=self.language,
+            use_language_model_featurizer=True)
+
+        RepositoryExample.objects.create(
+            repository_update=self.repository.current_update(),
+            text='my name is Douglas',
+            intent='greet')
+        RepositoryExample.objects.create(
+            repository_update=self.repository.current_update(),
+            text='my name is John',
+            intent='greet')
+
+    def test_change_ready_for_train(self):
+        self.assertTrue(self.repository.ready_for_train)
+        current_update = self.repository.current_update()
+        current_update.start_training(self.owner)
+        current_update.save_training(b'')
+        self.assertFalse(self.repository.ready_for_train)
+        self.repository.use_language_model_featurizer = False
+        self.repository.save()
+        self.assertTrue(self.repository.ready_for_train)
+        self.repository.use_language_model_featurizer = True
+        self.repository.save()
+        self.assertFalse(self.repository.ready_for_train)
+
+    def test_equal_repository_value_after_train(self):
+        current_update = self.repository.current_update()
+        self.repository.use_language_model_featurizer = False
+        self.repository.save()
+        self.assertTrue(current_update.use_language_model_featurizer)
+        current_update.start_training(self.owner)
+        current_update.save_training(b'')
+        self.assertFalse(current_update.use_language_model_featurizer)