Updating BasedSettings and other compatibility issues

howl/client/howl_client.py

@@ -109,7 +109,7 @@ def _on_audio(self, in_data, frame_count, time_info, status_flags):
         return data_ok
 
     def _normalize_audio(self, audio_data):
-        return np.frombuffer(audio_data, dtype=np.int16).astype(np.float) / self._audio_float_size
+        return np.frombuffer(audio_data, dtype=np.int16).astype(float) / self._audio_float_size
 
     def start(self):
         """Start the audio stream for inference"""

howl/model/cnn.py

@@ -1,6 +1,6 @@
 from typing import Tuple
 
-from pydantic import BaseSettings
+from pydantic_settings import BaseSettings
 from torchvision.models import MobileNetV2, mobilenet_v2
 import torch
 import torch.nn as nn

howl/model/rnn.py

@@ -3,7 +3,7 @@
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
-from pydantic import BaseSettings
+from pydantic_settings import BaseSettings
 from torch.nn.utils.rnn import pack_padded_sequence, pad_packed_sequence
 
 from .base import RegisteredModel

howl/settings.py

@@ -1,7 +1,7 @@
 import multiprocessing
-from typing import List
+from typing import List, Optional
 
-from pydantic import BaseSettings
+from pydantic_settings import BaseSettings
 
 __all__ = ["AudioSettings", "DatasetSettings", "SETTINGS"]
 
@@ -38,7 +38,7 @@ class AudioTransformSettings(BaseSettings):
 class InferenceEngineSettings(BaseSettings):
     """Base settings for inference engine"""
 
-    inference_weights: List[float] = None
+    inference_weights: Optional[List[float]] = None
     inference_sequence: List[int] = [0]
     inference_window_ms: float = 2000  # look at last of these seconds
     smoothing_window_ms: float = 50  # prediction smoothed
@@ -66,9 +66,9 @@ class TrainingSettings(BaseSettings):
     objective: str = "frame"  # frame or ctc
     # TODO: support phone token_type
     token_type: str = "word"
-    phone_dictionary: str = None
+    phone_dictionary: Optional[str] = None
     use_noise_dataset: bool = False
-    noise_dataset_path: str = None
+    noise_dataset_path: Optional[str] = None
 
 
 class DatasetSettings(BaseSettings):

howl/utils/dataclass.py

@@ -1,6 +1,6 @@
 import json
 
-from pydantic import BaseSettings
+from pydantic_settings import BaseSettings
 from pathlib import PosixPath
 
 

requirements.txt

@@ -8,11 +8,12 @@ praat-textgrids==1.3.1
 protobuf==3.19.4 # Protobuf 4.21.0 contains a breaking change for Python, as explained here: https://developers.google.com/protocol-buffers/docs/news/2022-05-06#python-updates
 pyaudio
 pydantic>=1.7.3
+pydantic_settings
 pytest==6.2.3
 pytest-timeout==2.1.0
 soundfile==0.10.3.post1
 tensorboard==2.8.0
-torch==1.10.1
+torch>=1.10.1
 torchaudio>=0.5.0
 torchvision>=0.6.0
 tqdm