Update xtts.py

xtts.py

@@ -1,4 +1,4 @@
-import re, io, os, stat
+import re, io, os, stat, logging
 import tempfile, subprocess
 import requests
 import torch
@@ -12,6 +12,8 @@ from flask import Flask, Blueprint,  request, jsonify, send_file
 import torch
 import torchaudio
 
+logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
+
 app = Flask(__name__)
 # def upload_bytes(bytes, ext=".wav"):
 #     return bytes
@@ -52,13 +54,13 @@ def predict():
         if tts is None:
             TTS=import_module("TTS.api").TTS
             model_name="tts_models/multilingual/multi-dataset/xtts_v2"
-            print(f"loading model {model_name} ...")
+            logging.info(f"loading model {model_name} ...")
             tts = TTS(model_name=model_name, progress_bar=False)
             model=tts.synthesizer.tts_model
             #hack to use cache
             model.__get_conditioning_latents=model.get_conditioning_latents
             model.get_conditioning_latents=get_conditioning_latents
-            print("model is ready")
+            logging.info("model is ready")
 
         wav = tts.tts(
             text, 
@@ -76,7 +78,7 @@ def predict():
             scipy.io.wavfile.write(wav_buffer, tts.synthesizer.output_sample_rate, wav_norm)
             wav_bytes = wav_buffer.getvalue()
             url= upload_bytes(wav_bytes, ext=".wav")
-            print(f'wav is at {url}')
+            logging.debug(f'wav is at {url}')
             return url
     except Exception as e:
         traceback.print_exc()
@@ -105,14 +107,14 @@ def get_conditioning_latents(audio_path, **others):
                 gpt_cond_latent,
                 speaker_embedding,
             ) = torch.load(pt_file)
-            print(f'sample wav info loaded from {pt_file}')
+            logging.debug(f'sample wav info loaded from {pt_file}')
     except:
         (
             gpt_cond_latent,
             speaker_embedding,
         ) = model.__get_conditioning_latents(audio_path=speaker_wav, **others)
         torch.save((gpt_cond_latent,speaker_embedding), pt_file)
-        print(f'sample wav info saved to {pt_file}')
+        logging.debug(f'sample wav info saved to {pt_file}')
     return gpt_cond_latent,speaker_embedding
 
 def download(url):
@@ -147,5 +149,5 @@ def trim_sample_audio(speaker_wav):
         return speaker_wav
 
 for key, value in os.environ.items():
-    print(f"{key}: {value}")
-print("xtts is ready")
+    logging.info(f"{key}: {value}")
+loggin.info("xtts is ready")