feat: use structlog to handle logging, add some logs

aip_trainer/lambdas/lambdaSpeechToScore.py

@@ -60,8 +60,10 @@ def lambda_handler(event, context):
     duration = time.time() - start
     app_logger.info(f'Loaded .ogg file {random_file_name} in {duration}s.')
 
-    result = trainer_SST_lambda[language].processAudioForGivenText(
-        signal, real_text)
+    language_trainer_sst_lambda = trainer_SST_lambda[language]
+    app_logger.info(f'language_trainer_sst_lambda: preparing...')
+    result = language_trainer_sst_lambda.processAudioForGivenText(signal, real_text)
+    app_logger.info(f'language_trainer_sst_lambda: result: {result}...')
 
     start = time.time()
     os.remove(random_file_name)

aip_trainer/models/RuleBasedModels.py

@@ -1,6 +1,7 @@
 import eng_to_ipa
 
 from aip_trainer.models import ModelInterfaces
+from aip_trainer import app_logger
 
 
 class EpitranPhonemConverter(ModelInterfaces.ITextToPhonemModel):
@@ -12,7 +13,10 @@ class EpitranPhonemConverter(ModelInterfaces.ITextToPhonemModel):
         self.epitran_model = epitran_model
 
     def convertToPhonem(self, sentence: str) -> str:
+
+        app_logger.info(f'starting EpitranPhonemConverter.convertToPhonem...')
         phonem_representation = self.epitran_model.transliterate(sentence)
+        app_logger.info(f'EpitranPhonemConverter: got phonem_representation!')
         return phonem_representation
 
 

aip_trainer/pronunciationTrainer.py

@@ -21,15 +21,14 @@ def getTrainer(language: str):
     asr_model = AIModels.NeuralASR(model, decoder)
 
     if language == 'de':
-        phonem_converter = RuleBasedModels.EpitranPhonemConverter(
-            epitran.Epitran('deu-Latn'))
+        epitran_deu_latn = epitran.Epitran('deu-Latn')
+        phonem_converter = RuleBasedModels.EpitranPhonemConverter(epitran_deu_latn)
     elif language == 'en':
         phonem_converter = RuleBasedModels.EngPhonemConverter()
     else:
         raise ValueError('Language not implemented')
 
-    trainer = PronunciationTrainer(
-        asr_model, phonem_converter)
+    trainer = PronunciationTrainer(asr_model, phonem_converter)
 
     return trainer
 
@@ -82,8 +81,8 @@ class PronunciationTrainer:
     def processAudioForGivenText(self, recordedAudio: torch.Tensor = None, real_text=None):
 
         start = time.time()
-        recording_transcript, recording_ipa, word_locations = self.getAudioTranscript(
-            recordedAudio)
+        app_logger.info(f'starting getAudioTranscript...')
+        recording_transcript, recording_ipa, word_locations = self.getAudioTranscript(recordedAudio)
 
         duration = time.time() - start
         app_logger.info(f'Time for NN to transcript audio: {duration}.')
@@ -114,16 +113,19 @@ class PronunciationTrainer:
     def getAudioTranscript(self, recordedAudio: torch.Tensor = None):
         current_recorded_audio = recordedAudio
 
-        current_recorded_audio = self.preprocessAudio(
-            current_recorded_audio)
+        app_logger.info(f'starting preprocessAudio...')
+        current_recorded_audio = self.preprocessAudio(current_recorded_audio)
 
+        app_logger.info(f'starting processAudio...')
         self.asr_model.processAudio(current_recorded_audio)
 
+        app_logger.info(f'starting getTranscriptAndWordsLocations...')
         current_recorded_transcript, current_recorded_word_locations = self.getTranscriptAndWordsLocations(
             current_recorded_audio.shape[1])
-        current_recorded_ipa = self.ipa_converter.convertToPhonem(
-            current_recorded_transcript)
+        app_logger.info(f'starting convertToPhonem...')
+        current_recorded_ipa = self.ipa_converter.convertToPhonem(current_recorded_transcript)
 
+        app_logger.info(f'ok, return audio transcript!')
         return current_recorded_transcript, current_recorded_ipa, current_recorded_word_locations
 
     def getWordLocationsFromRecordInSeconds(self, word_locations, mapped_words_indices) -> list:

tests/__init__.py

@@ -0,0 +1,12 @@
+import os
+
+import structlog
+from aip_trainer import PROJECT_ROOT_FOLDER, LOG_JSON_FORMAT
+from aip_trainer.utils import session_logger
+
+
+TEST_ROOT_FOLDER = PROJECT_ROOT_FOLDER / "tests"
+EVENTS_FOLDER = TEST_ROOT_FOLDER / "events"
+log_level = os.getenv("LOG_LEVEL", "INFO")
+session_logger.setup_logging(json_logs=LOG_JSON_FORMAT, log_level=log_level)
+test_logger = structlog.stdlib.get_logger(__name__)