Update app.py

app.py

@@ -18,32 +18,22 @@ model.enable_cpu_offload()
 
 import numpy as np
 from scipy.io.wavfile import write as write_wav
-#from pydub import AudioSegment
-#import soundfile as sf
+
 
 def infer(text_prompt):
     text_prompt = text_prompt
     inputs = processor(text_prompt).to(device)
     
     with torch.inference_mode():
-        speech_output = model.generate(**inputs, do_sample = True, fine_temperature = 0.4, coarse_temperature = 0.8)
-        #audio_out = speech_output[0].cpu().numpy()
-        audio_array = speech_output[0].cpu().numpy().squeeze()
+        speech_output = model.generate(**inputs)
+    
+    audio_array = speech_output[0].cpu().numpy().squeeze()
         
-        # Assuming audio_out contains audio data and the sampling rate
-        sampling_rate = model.generation_config.sample_rate
-        write_wav("output.wav", sampling_rate, audio_array)
-        #wavfile.write("output.wav", sampling_rate, audio_out)
+    # Assuming audio_out contains audio data and the sampling rate
+    sampling_rate = model.generation_config.sample_rate
+    write_wav("output.wav", sampling_rate, audio_array)
         
         
-
-    
-
-    #wav_path = "output.wav"
-    #mp3_path = "output.mp3"
-    
-    #audio = AudioSegment.from_wav(wav_path)
-    #audio.export(mp3_path, format="mp3")
     
     return "output.wav"