Update app.py

app.py

@@ -4,7 +4,7 @@ from peft import PeftModel
 import torch
 
 # -----------------------------
-# Load 4-bit Qwen model locally
+# Model config
 # -----------------------------
 MODEL_NAME = "unsloth/qwen2.5-math-1.5b-bnb-4bit"
 
@@ -12,35 +12,40 @@ print("Loading tokenizer...")
 tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, use_fast=False)
 
 print("Loading model in 4-bit...")
-base_model = AutoModelForCausalLM.from_pretrained(
-    MODEL_NAME,
-    device_map="auto",
-    torch_dtype=torch.float16,
-    low_cpu_mem_usage=True
-)
-
-# Check if LoRA adapter exists
 try:
-    model = PeftModel.from_pretrained(base_model, MODEL_NAME, device_map="auto")
-except:
-    model = base_model
+    base_model = AutoModelForCausalLM.from_pretrained(
+        MODEL_NAME,
+        device_map="auto",
+        dtype=torch.float16,  # newer transformers prefers dtype
+        low_cpu_mem_usage=True
+    )
+    # Load LoRA adapter if exists
+    try:
+        model = PeftModel.from_pretrained(base_model, MODEL_NAME, device_map="auto")
+    except:
+        model = base_model
+except Exception as e:
+    print("Error loading model:", e)
+    raise e
 
 model.eval()
 
 # -----------------------------
-# Respond function
+# Response function
 # -----------------------------
 def respond(message, history, system_message, max_tokens, temperature, top_p):
-    # Build chat prompt
+    # Limit max tokens for safety
+    if max_tokens > 128:
+        max_tokens = 128
+
+    # Build prompt
     prompt = system_message + "\n"
     for h in history:
         prompt += f"User: {h['content']}\n"
     prompt += f"User: {message}\nBot:"
 
-    # Tokenize
     inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
 
-    # Generation config
     gen_config = GenerationConfig(
         max_new_tokens=max_tokens,
         temperature=temperature,
@@ -48,7 +53,6 @@ def respond(message, history, system_message, max_tokens, temperature, top_p):
         do_sample=True
     )
 
-    # Generate output
     with torch.no_grad():
         output_ids = model.generate(**inputs, **gen_config.to_dict())
     output = tokenizer.decode(output_ids[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
@@ -62,8 +66,8 @@ chatbot = gr.ChatInterface(
     type="messages",
     additional_inputs=[
         gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+        gr.Slider(minimum=1, maximum=2048, value=64, step=1, label="Max new tokens"),
+        gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature"),
         gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
     ],
 )
@@ -71,9 +75,5 @@ chatbot = gr.ChatInterface(
 with gr.Blocks() as demo:
     chatbot.render()
 
-# -----------------------------
-# Launch Gradio app
-# -----------------------------
 if __name__ == "__main__":
     demo.launch(server_name="0.0.0.0", server_port=7860)
-