Update app.py

app.py

@@ -32,8 +32,8 @@ from gpuinfo import GPUInfo
 #import csv
 import numpy as np
 import torch
-#import torchaudio
-#import torchaudio.transforms as transforms
+import torchaudio
+import torchaudio.transforms as transforms
 
 from transformers import pipeline, AutoModel
 
@@ -68,19 +68,13 @@ pipe = pipeline("automatic-speech-recognition", model="NbAiLab/nb-whisper-large"
 
 @spaces.GPU()
 def transcribe_audio(audio_file, batch_size=16):
-    
-    # --check if audio_file is tuple
-    if isinstance(audio_file, tuple):
-        audio_file = audio_file[0]
-        
-    # --place audio file in numpy array
-    audio = AudioSegment.from_wav(audio_file)
-    samples = np.array(audio.get_array_of_samples())
-    sample_rate = audio.frame_rate
+    # --audio file to tensor
+    waveform, sample_rate = torchaudio.load(audio_file)
+    samples = waveform.numpy()
 
     start_time = time.time()
 
-    # --transcribe 
+    # --pipe it
     outputs = pipe(samples, sampling_rate=sample_rate, batch_size=batch_size, return_timestamps=False)
     text = outputs["text"]
 
@@ -89,13 +83,13 @@ def transcribe_audio(audio_file, batch_size=16):
     output_time = end_time - start_time
     word_count = len(text.split())
 
-    # --GPU usage
+    # --GPU metrics
     memory = psutil.virtual_memory()
     gpu_utilization, gpu_memory = GPUInfo.gpu_usage()
     gpu_utilization = gpu_utilization[0] if len(gpu_utilization) > 0 else 0
     gpu_memory = gpu_memory[0] if len(gpu_memory) > 0 else 0
 
-    # --CPU usage
+    # --CPU metric
     cpu_usage = psutil.cpu_percent(interval=1)
 
     # --system info string
@@ -110,6 +104,7 @@ def transcribe_audio(audio_file, batch_size=16):
     return text.strip(), system_info
 
 
+
 #              ------------summary section------------