Update app.py

app.py

@@ -1,9 +1,12 @@
+# Import spaces first to avoid CUDA initialization issues
+import spaces
+
+# Then import other libraries
 import torch
 import librosa
 from transformers import pipeline, WhisperProcessor, WhisperForConditionalGeneration, AutoModelForCausalLM, AutoProcessor
 from gtts import gTTS
 import gradio as gr
-import spaces
 from PIL import Image
 import os
 import io
@@ -37,6 +40,26 @@ def load_whisper():
         print(f"Error loading Whisper model: {e}")
         return None, None
 
+# Load sarvam-2b for text generation within a GPU-decorated function
+@spaces.GPU
+def load_sarvam():
+    return load_pipeline('sarvamai/sarvam-2b-v0.5')
+
+# Load Phi-3.5-vision-instruct model
+@spaces.GPU
+def load_vision_model():
+    try:
+        model_id = "microsoft/Phi-3.5-vision-instruct"
+        model = AutoModelForCausalLM.from_pretrained(
+            model_id, trust_remote_code=True, torch_dtype=torch.float16, use_flash_attention_2=False
+        )
+        processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True, num_crops=16)
+        return model, processor
+    except Exception as e:
+        print(f"Error loading vision model: {e}")
+        return None, None
+
+
 # Load sarvam-2b for text generation within a GPU-decorated function
 @spaces.GPU
 def load_sarvam():
@@ -269,7 +292,6 @@ custom_suggestions = """
     </div>
 </div>
 """
-
 # Create Gradio interface
 with gr.Blocks(css=custom_css, theme=gr.themes.Base().set(
     body_background_fill="#0b0f19",