Update app.py

app.py

@@ -5,7 +5,7 @@ from transformers import pipeline
 import time, io
 
 device = 0 if torch.cuda.is_available() else -1
-if device == -1: print("⚠️ No GPU detected. Expect ~10–20s for 300,000 chars on CPU.")
+if device == -1: raise RuntimeError("GPU required for 5–10s target")
 
 summarizer = pipeline("summarization", model="google/pegasus-xsum", device=device, torch_dtype=torch.int8)
 
@@ -25,14 +25,13 @@ async def summarize_file(file_bytes):
     chunks = [text[i:i+15000] for i in range(0, len(text), 15000)]
     if not chunks: return "❌ No chunks to summarize"
     summaries = []
-    batch_size = 2 if device == -1 else 10  # Smaller batch for CPU
-    for i in range(0, len(chunks), batch_size):
-        if time.time() - start > 9: 
+    for i in range(0, len(chunks), 10):
+        if time.time() - start > 7: 
             summaries.append("⚠️ Stopped early")
             break
-        batch = chunks[i:i+batch_size]
+        batch = chunks[i:i+10]
         try:
-            batch_summaries = summarizer(batch, max_length=40, min_length=10, do_sample=False, batch_size=batch_size)
+            batch_summaries = summarizer(batch, max_length=40, min_length=10, do_sample=False, batch_size=10)
             summaries.extend(f"**Chunk {i+j+1}**:\n{s['summary_text']}" for j, s in enumerate(batch_summaries))
         except: summaries.append(f"**Chunk {i+1}**: ❌ Error")
     return f"**Chars**: {len(text)}\n**Time**: {time.time()-start:.2f}s\n\n" + "\n\n".join(summaries)
@@ -40,7 +39,7 @@ async def summarize_file(file_bytes):
 demo = gr.Interface(
     fn=summarize_file, inputs=gr.File(label="📄 PDF/TXT Notes"), 
     outputs=gr.Textbox(label="📝 Summary"), 
-    title="Fast Summarizer", description="300,000+ chars in ~5–10s (GPU) or ~10–20s (CPU)"
+    title="Fast Summarizer", description="300,000+ chars in ~5s (GPU)"
 )
 
 if __name__ == "__main__":