I hope it's final commit..

app.py

@@ -3,8 +3,8 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 import torch
 
 # Use lighter model for CPU
-model_name = "microsoft/phi-2"  # 2.7B - TOO HEAVY
-#model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"  # 1.1B - much lighter
+#model_name = "microsoft/phi-2"  # 2.7B - TOO HEAVY
+model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"  # 1.1B - much lighter
 
 try:
     print(f"Loading {model_name}...")
@@ -43,8 +43,8 @@ def generate_response(message):
             outputs = model.generate(
                 inputs.input_ids,
                 attention_mask=inputs.attention_mask,  # FIX: Add attention mask
-                max_new_tokens=400,  # Reduced for CPU
-                temperature=0.7,
+                max_new_tokens=600,  # Reduced for CPU
+                temperature=0.8,
                 do_sample=True,
                 top_p=0.9,
                 pad_token_id=tokenizer.pad_token_id,
@@ -79,6 +79,6 @@ interface.queue(default_concurrency_limit=1)
 interface.launch(
     server_name="0.0.0.0", 
     server_port=7860, 
-    share=False,
+    share=True,
     debug=False
 )