Update app.py

app.py

@@ -539,12 +539,9 @@ Additional Context: {context}
         # Generation parameters
         generation_kwargs = dict(
             **inputs,
-            max_new_tokens=150,      # REDUCED from 250 to 150
-            temperature=0.1,         # LOWERED from 0.2 to 0.1 (more deterministic)
-            do_sample=False,         # CHANGED to False (greedy decoding)
-            top_p=0.95,
-            top_k=40,
-            repetition_penalty=1.5,  # INCREASED to strongly discourage verbosity
+            max_new_tokens=150,      # SLIGHTLY LOWER to prevent cutoff
+            do_sample=False,         # Greedy decoding (no temperature needed)
+            repetition_penalty=1.5,  # Keep this - prevents repetition
             pad_token_id=phi_tokenizer.eos_token_id,
             eos_token_id=phi_tokenizer.eos_token_id,
             streamer=request_streamer,