Auto-update via watch script

app.py

@@ -168,7 +168,7 @@ def respond(message, history, system_prompt, max_tokens, temperature, disable_th
             # Set up generator parameters
             params = og.GeneratorParams(model)
             params.set_search_options(
-                max_length=int(max_tokens),
+                max_length=8192,
                 temperature=temperature
             )
             
@@ -183,12 +183,15 @@ def respond(message, history, system_prompt, max_tokens, temperature, disable_th
                 generator.append_tokens(np.array(inputs, dtype=np.int32))
                 
             partial = ""
-            while not generator.is_done():
+            max_gen_tokens = int(max_tokens)
+            tokens_generated = 0
+            while not generator.is_done() and tokens_generated < max_gen_tokens:
                 generator.generate_next_token()
                 new_token = generator.get_next_tokens()[0]
                 decoded = tokenizer.decode([new_token])
                 partial += decoded
                 yield partial
+                tokens_generated += 1
         except Exception as e:
             print(f"[inference] Error during Phi-3 generation: {e}", flush=True)
             yield f"⚠️ **Error during model inference**: {str(e)}\n\n*This was caught gracefully to prevent the Space from crashing. Please try again or rephrase.*"