feat: support pytorch and torchaudio, update test, add requirements-dev.txt

.gitignore

@@ -199,6 +199,7 @@ tmp
 nohup.out
 /tests/events.tar
 function_dump_*.json
+*.yml
 
 # onnx models
 *.onnx

aip_trainer/models/models.py

@@ -1,25 +1,14 @@
-from typing import Any
-
-import torch
 import torch.nn as nn
+from silero import silero_stt
+from silero.utils import Decoder
 
 
 # second returned type here is the custom class src.silero.utils.Decoder from snakers4/silero-models
-def getASRModel(language: str) -> tuple[nn.Module, Any]:
-
-
+def getASRModel(language: str) -> tuple[nn.Module, Decoder]:
     if language == 'de':
-
-        model, decoder, utils = torch.hub.load(repo_or_dir='snakers4/silero-models',
-                                               model='silero_stt',
-                                               language='de',
-                                               device=torch.device('cpu'))
-
+        model, decoder, _ = silero_stt(language='de', version="v4", jit_model="jit_large")
     elif language == 'en':
-        model, decoder, utils = torch.hub.load(repo_or_dir='snakers4/silero-models',
-                                               model='silero_stt',
-                                               language='en',
-                                               device=torch.device('cpu'))
+        model, decoder, _ = silero_stt(language='en')
     else:
         raise NotImplementedError("currenty works only for 'de' and 'en' languages, not for '{}'.".format(language))
 

requirements-dev.txt

@@ -0,0 +1,2 @@
+pytest
+pytest-cov

requirements.txt

@@ -7,7 +7,6 @@ flask_cors
 omegaconf
 ortools==9.11.4210
 pandas
-numpy<2.0.0
 pickle-mixin
 python-dotenv
 requests
@@ -15,6 +14,6 @@ sentencepiece
 soundfile==0.12.1
 sqlalchemy
 structlog
-torch==1.13.1
-torchaudio==0.13.1
+torch
+torchaudio
 transformers

tests/test_GetAccuracyFromRecordedAudio.py

@@ -40,7 +40,9 @@ class TestGetAccuracyFromRecordedAudio(unittest.TestCase):
             output["matched_transcripts"] = expected_output["matched_transcripts"]
             output["matched_transcripts_ipa"] = expected_output["matched_transcripts_ipa"]
             output["pronunciation_accuracy"] = expected_output["pronunciation_accuracy"]
+            output["pair_accuracy_category"] = expected_output["pair_accuracy_category"]
             output["ipa_transcript"] = expected_output["ipa_transcript"]
+            output["real_transcript"] = expected_output["real_transcript"]
             output["real_transcripts_ipa"] = expected_output["real_transcripts_ipa"]
             self.assertEqual(expected_output, output)