Update app.py

app.py

@@ -52,6 +52,59 @@ def device_info():
     except subprocess.CalledProcessError as e:
         print(f"Command failed: {e}")
 
+from gpustat import GPUStatCollection
+
+def update_gpu_status():
+    if torch.cuda.is_available() == False:
+        return "No Nvidia Device"
+    try:
+        gpu_stats = GPUStatCollection.new_query()
+        for gpu in gpu_stats:
+            # Assuming you want to monitor the first GPU, index 0
+            gpu_id = gpu.index
+            gpu_name = gpu.name
+            gpu_utilization = gpu.utilization
+            memory_used = gpu.memory_used
+            memory_total = gpu.memory_total
+            memory_utilization = (memory_used / memory_total) * 100
+            gpu_status=(f"> **GPU** {gpu_id}: {gpu_name}, Utilization: {gpu_utilization}%, **Memory Used**: {memory_used}MB, **Memory Total**: {memory_total}MB, **Memory Utilization**: {memory_utilization:.2f}%")
+            return gpu_status
+
+    except Exception as e:
+        print(f"Error getting GPU stats: {e}")
+        return torch_update_gpu_status()
+
+def torch_update_gpu_status():
+    if torch.cuda.is_available():
+        gpu_info = torch.cuda.get_device_name(0)
+        gpu_memory = torch.cuda.mem_get_info(0)
+        total_memory = gpu_memory[1] / (1024 * 1024)
+        free_memory=gpu_memory[0] /(1024 *1024)
+        used_memory = (gpu_memory[1] - gpu_memory[0]) / (1024 * 1024)
+        
+        gpu_status = f"> **GPU**: {gpu_info} **Free Memory**:{free_memory}MB   **Total Memory**: {total_memory:.2f} MB  **Used Memory**: {used_memory:.2f} MB"
+    else:
+        gpu_status = "No GPU available"
+    return gpu_status
+
+def update_cpu_status():
+    import datetime
+    current_time = datetime.datetime.now().time()
+    time_str = current_time.strftime("%H:%M:%S")
+
+    cpu_percent = psutil.cpu_percent()
+    cpu_status = f"> **CPU Usage: {cpu_percent}% {time_str}"
+    return cpu_status
+
+def update_status():
+    gpu_status = update_gpu_status()
+    cpu_status = update_cpu_status()
+    sys_status=gpu_status+"\n"+cpu_status
+    return sys_status
+
+def refresh_status():
+    return update_status()
+    
 @gpu_decorator(duration=SPACES_GPU_DURATION)
 def transcribe(inputs, model, language, batch_size, chunk_length_s, stride_length_s, task, timestamp_mode, progress=gr.Progress(track_tqdm=True)):
     try:
@@ -387,6 +440,10 @@ with demo:
         interface_list=[file_transcribe, video_transcribe, yt_transcribe],
         tab_names=["Audio", "Video", "YouTube"]
     )
-
+    with gr.Group():
+        sys_status_output = gr.Markdown(label="System Status", interactive=False)
+        refresh_button = gr.Button("Refresh System Status")
+        refresh_button.click(refresh_status, None, sys_status_output)
+    
 if __name__ == "__main__":
     demo.queue().launch(ssr_mode=False)