u

app.py

@@ -1,22 +1,8 @@
+import spaces
 import gradio as gr
-from transformers import Wav2Vec2ForCTC, Wav2Vec2ProcessorWithLM
-import torch
-import torchaudio
-import numpy as np
-
-# Device and dtype configuration
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
-
-# Load model and processor with LM
-processor = Wav2Vec2ProcessorWithLM.from_pretrained("alakxender/wav2vec2-large-mms-1b-dv-syn-md")
-model = Wav2Vec2ForCTC.from_pretrained(
-    "alakxender/wav2vec2-large-mms-1b-dv-syn-md",
-    torch_dtype=torch_dtype
-).to(device)
-
-MAX_LENGTH = 120 # 2 minutes
-MIN_LENGTH = 1 # 1 second
+import subprocess
+import sys
+import os
 
 def transcribe(audio_file):
     try:
@@ -84,6 +70,61 @@ iface = gr.Interface(
     description="Upload an audio file to transcribe Dhivehi speech to text using language model enhanced decoding."
 )
 
+
+def install_requirements():
+    requirements_path = 'requirements.txt'
+    
+    # Check if requirements.txt exists
+    if not os.path.exists(requirements_path):
+        print("Error: requirements.txt not found")
+        return False
+        
+    try:
+        print("Installing requirements...")
+        # Using --no-cache-dir to avoid memory issues
+        subprocess.check_call([
+            sys.executable, 
+            "-m", 
+            "pip", 
+            "install", 
+            "-r", 
+            requirements_path,
+            "--no-cache-dir"
+        ])
+        print("Successfully installed all requirements")
+        return True
+    except subprocess.CalledProcessError as e:
+        print(f"Error installing requirements: {e}")
+        return False
+    except Exception as e:
+        print(f"Unexpected error: {e}")
+        return False
+
 # Launch the interface
 if __name__ == "__main__":
-    iface.launch()
+    success = install_requirements()
+    if success:
+        print("All requirements installed successfully")
+      
+        from transformers import Wav2Vec2ForCTC, Wav2Vec2ProcessorWithLM
+        import torch
+        import torchaudio
+        import numpy as np
+
+        # Device and dtype configuration
+        device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
+
+        # Load model and processor with LM
+        processor = Wav2Vec2ProcessorWithLM.from_pretrained("alakxender/wav2vec2-large-mms-1b-dv-syn-md")
+        model = Wav2Vec2ForCTC.from_pretrained(
+            "alakxender/wav2vec2-large-mms-1b-dv-syn-md",
+            torch_dtype=torch_dtype
+        ).to(device)
+
+        MAX_LENGTH = 120 # 2 minutes
+        MIN_LENGTH = 1 # 1 second
+
+        iface.launch()
+    else:
+        print("Failed to install some requirements")