readme

README.md

@@ -9,4 +9,22 @@ app_file: app.py
 pinned: false
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+
+How inference:
+1) [huggingface](https://huggingface.co/spaces/deepkotix/denoise_and_diarization)
+2) [telegram bot](https://t.me/diarizarion_bot)
+3) run local inference:
+   1) GUI:
+   `python app.py`
+   2) Inference local:
+   `python main_pipeline.py --audio-path dialog.mp3`
+
+
+|                       | inference time for file dialog.mp3 |
+|-----------------------|:----------------------------------:|
+| cpu 2v CPU huggingece |              600 s/it              |
+| gpu tesla v100        |             8.23 s/it              |
+
+
+
+

app.py

@@ -3,8 +3,11 @@ import gradio as gr
 from main_pipeline import main_pipeline
 from scipy.io.wavfile import write
 
-title = "audio_denoise and speakser diarization. Fast inference https://t.me/diarizarion_bot"
+title = "audio_denoise and speakser diarization. Faster inference [tg_bot](https://t.me/diarizarion_bot)"
 
+example_list = [
+    ["dialog.mp3"]
+]
 
 def app_pipeline(audio):
     audio_path = 'test.wav'
@@ -14,9 +17,6 @@ def app_pipeline(audio):
     return result_diarization + [None] * (10 - len(result_diarization))
 
 
-example_list = [
-    ["dialog.mp3"]
-]
 gr.Interface(
     app_pipeline,
     gr.Audio(type="numpy", label="Input"),

main_pipeline.py

@@ -1,6 +1,9 @@
+import argparse
 import librosa
 import torch
 import os
+from tqdm import tqdm
+
 from utils.denoise_pipeline import denoise
 from utils.diarization_pipeline import diarization
 import numpy as np
@@ -41,4 +44,8 @@ def main_pipeline(audio_path):
 
 
 if __name__ == '__main__':
-    main_pipeline('dialog.mp3')
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--audio-path', default='dialog.mp3', help='Path to audio')
+    opt = parser.parse_args()
+    for _ in tqdm(range(10)):
+        main_pipeline(audio_path=opt.audio_path)