Switch to using low accuracy mode language detection

This is needed since it consumes way less ram and is also faster and
could cause the process to be killed in some docker configurations.

Author: rolandgeider

wger/core/api/serializers.py

@@ -232,7 +232,10 @@ def validate(self, data):
 
         # Try to detect the language
         detector = (
-            LanguageDetectorBuilder.from_all_languages().with_preloaded_language_models().build()
+            LanguageDetectorBuilder.from_all_languages()
+            .with_low_accuracy_mode()
+            .with_preloaded_language_models()
+            .build()
         )
         input_str = data.get('input')
 

wger/exercises/api/serializers.py

@@ -226,7 +226,10 @@ def create(self, validated_data, **kwargs):
         # -> This is done here instead of in the serializer's validate method
         #    because the language is not available in the serializer's initial_data
         detector = (
-            LanguageDetectorBuilder.from_all_languages().with_preloaded_language_models().build()
+            LanguageDetectorBuilder.from_all_languages()
+            .with_low_accuracy_mode()
+            .with_preloaded_language_models()
+            .build()
         )
         language = translation.language
 
@@ -410,7 +413,10 @@ def validate(self, data):
         provided language.
         """
         detector = (
-            LanguageDetectorBuilder.from_all_languages().with_preloaded_language_models().build()
+            LanguageDetectorBuilder.from_all_languages()
+            .with_low_accuracy_mode()
+            .with_preloaded_language_models()
+            .build()
         )
 
         language = data.get('language')