feat: add preloader for effiecient in time

app.py

@@ -1,9 +1,9 @@
 from src import infer, utils
 import gradio as gr
 
-infer.download_all_available_model()
+infer.model_preloader_downloader()
 
-audio_examples = [
+AUDIO_EXAMPLE = [
     [None, "assets/audio/male-indonesian.wav", None],
     [None, "assets/audio/female-indonesian.wav", None],
     [None, "assets/audio/male-english.wav", None],
@@ -19,7 +19,10 @@ demo = gr.Interface(
     inputs=[
         gr.Dropdown(
             label="Model",
-            choices=["tiny","small","base","medium","large","large-v2"],
+            choices=[
+                "tiny", "base", "small", "medium", 
+                "large", "large-v1", "large-v2"
+            ],
             value="base"),
         gr.Radio(label="Language", 
                  choices=["indonesian","english"], 

src/infer.py

@@ -2,15 +2,23 @@
 from typing import *
 from src import utils
 import whisper
+  
 
+MODEL_NAMES = ["tiny", "base", "small", "medium", "large", "large-v1","large-v2"]
+MODEL_BASE = whisper.load_model("base")
 
-def download_all_available_model():
-    list_names = ["tiny", "base", "small", "medium", "large", "large-v1","large-v2"]
-    for mname in list_names:
+def model_preloader_downloader():
+    for mname in MODEL_NAMES:
         mdl = whisper.load_model(mname)
         del mdl
 
-def predict(model_name, language, mic_audio=None, audio=None):
+def model_loader(name: str):
+    if name=="base":
+        return MODEL_BASE
+    else:
+        return whisper.load_model(name)
+
+def predict(model_name: str, language: str, mic_audio=None, audio=None):
     if mic_audio is not None:
         voice = mic_audio
     elif audio is not None:
@@ -20,6 +28,6 @@ def predict(model_name, language, mic_audio=None, audio=None):
     
     voice = utils.preprocess_audio(voice)
     
-    model = whisper.load_model(model_name)
+    model = model_loader(model_name)
     result = model.transcribe(voice, language=language)
     return result["text"]

src/utils.py

@@ -1,11 +1,10 @@
 from pathlib import Path
 import whisper
+from typing import Any
 
-sample_rate: int = 16000
-float_factor: float = 32678.0
 
 
-def preprocess_audio(filepath: str):
+def preprocess_audio(filepath: str) -> Any:
     # load audio and pad/trim it to fit 30 seconds
     audio = whisper.load_audio(filepath)
     audio = whisper.pad_or_trim(audio)
@@ -18,4 +17,5 @@ def parsing_text(filepath: str):
         raise ValueError("Invalid file type. Only '.txt' and '.md' files are supported.")
 
     return path.read_text()
-    
+    
+