Update app.py

app.py

@@ -26,11 +26,7 @@ import tempfile
 nltk.download("punkt")
 nltk.download("punkt_tab")
 
-# Automatically select device based on hardware availability
-DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
-print(f"✅ Using {DEVICE.upper()} for TTS and Summarization")
-
-kokoro_tts = KPipeline(lang_code='a', device=DEVICE)  
+kokoro_tts = KPipeline(lang_code='a')  
 
 # Supported TTS Languages
 SUPPORTED_TTS_LANGUAGES = {
@@ -53,7 +49,7 @@ model_name = "facebook/bart-large-cnn"
 
 try:
     tokenizer = BartTokenizer.from_pretrained(model_name, cache_dir=os.path.join(os.getcwd(), ".cache"))
-    model = BartForConditionalGeneration.from_pretrained(model_name, cache_dir=os.path.join(os.getcwd(), ".cache")).to(DEVICE)
+    model = BartForConditionalGeneration.from_pretrained(model_name, cache_dir=os.path.join(os.getcwd(), ".cache"))
 
 except Exception as e:
     raise RuntimeError(f"Error loading BART model: {e}")
@@ -106,7 +102,7 @@ def fetch_and_display_content(url):
     # Add detected language to metadata
     metadata["Detected Language"] = detected_lang.upper()
 
-    return (
+	return (
         cleaned_text, 
         metadata, 
         detected_lang,
@@ -283,10 +279,11 @@ def summarize_text(text, max_input_tokens=1024, max_output_tokens=200):
     inputs = tokenizer.encode("summarize: " + text, return_tensors="pt", max_length=max_input_tokens, truncation=True)
     summary_ids = model.generate(inputs, max_length=max_output_tokens, min_length=50, length_penalty=2.0, num_beams=4, early_stopping=True)
     return tokenizer.decode(summary_ids[0], skip_special_tokens=True)
-	
+
+@spaces.GPU(duration=1000)	
 def hierarchical_summarization(text):
     """Performs hierarchical summarization by chunking content first."""
-    print(f"✅ Summarization will run on: {DEVICE.upper()}")
+    #print(f"✅ Summarization will run on: {DEVICE.upper()}")
 	
     if len(text) > 10000:
         print("⚠️ Warning: Large input text detected. Summarization may take longer than usual.")
@@ -299,7 +296,7 @@ def hierarchical_summarization(text):
         padding=True,
         truncation=True,
         max_length=1024
-    ).to(DEVICE)
+    )
 	
 	#Generate the summary
     summary_ids = model.generate(
@@ -403,7 +400,9 @@ with gr.Blocks() as demo:
     summary_output = gr.Textbox(label="Summary", visible=True, interactive=False)
     full_audio_output = gr.Audio(label="Generated Audio", visible=True)
     ner_output = gr.Textbox(label="Extracted Entities", visible=True, interactive=False)  
-   
+    
+	
+	
     default_entity_types = gr.Textbox(label="Default Entity Types", value="PERSON, Organization, location, Date, PRODUCT, EVENT", interactive=True)
     custom_entity_types = gr.Textbox(label="Custom Entity Types", placeholder="Enter additional entity types (comma-separated)", interactive=True)
 
@@ -454,7 +453,7 @@ with gr.Blocks() as demo:
          
 	    extract_entities_with_gliner,
          
-        inputs=[extracted_text, default_entity_types, custom_entity_types],
+		inputs=[extracted_text, default_entity_types, custom_entity_types],
         outputs=[ner_output]
     )