Update app.py

app.py

@@ -22,14 +22,16 @@ def generate_response(shloka, transliteration):
             "content": f"Shloka: {shloka} Transliteration: {transliteration}"
         }
     ]
+    
     inputs = tokenizer.apply_chat_template(
         input_message,
         tokenize=True,
-        add_generation_prompt=True,  # Enable for generation
+        add_generation_prompt=True, 
         return_tensors="pt"
-    ).to("cuda")  # Assuming the model is running on GPU
+    ).to("cpu")
+
+model = model.to("cpu")
 
-    # Generate response
     text_streamer = TextStreamer(tokenizer, skip_prompt=True)
     generated_tokens = model.generate(
         input_ids=inputs,
@@ -42,8 +44,7 @@ def generate_response(shloka, transliteration):
 
     raw_response = tokenizer.decode(generated_tokens[0], skip_special_tokens=True)
 
-    # Format the response
-    # Assuming raw_response contains English Meaning, Hindi Meaning, and Word Meaning in sequence
+
     try:
         sections = raw_response.split("Hindi Meaning:")
         english_meaning = sections[0].strip()
@@ -58,12 +59,10 @@ def generate_response(shloka, transliteration):
             f"Word Meaning:\n{word_meaning}"
         )
     except IndexError:
-        # In case the response format is not as expected
         formatted_response = raw_response
 
     return formatted_response
 
-# Gradio interface
 interface = gr.Interface(
     fn=generate_response,
     inputs=[
@@ -75,7 +74,6 @@ interface = gr.Interface(
     description="Input a Shloka with its transliteration, and this model will provide meanings in English and Hindi along with word meanings."
 )
 
-# Launch the interface
 if __name__ == "__main__":
     interface.launch()