Update app.py

app.py

@@ -3,11 +3,12 @@ from transformers import AutoTokenizer, AutoModelForCausalLM
 from peft import PeftModel
 import gradio as gr
 
-# Load model and tokenizer
+# Model loading
 base_model_name = "unsloth/gemma-3-12b-it-unsloth-bnb-4bit"
 adapter_name = "adarsh3601/my_gemma3_pt"
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
+# Load base model in 4-bit with float16
 base_model = AutoModelForCausalLM.from_pretrained(
     base_model_name,
     device_map="auto",
@@ -15,44 +16,36 @@ base_model = AutoModelForCausalLM.from_pretrained(
     load_in_4bit=True
 )
 
+# Load tokenizer
 tokenizer = AutoTokenizer.from_pretrained(base_model_name)
+
+# Load fine-tuned adapter
 model = PeftModel.from_pretrained(base_model, adapter_name)
 model.to(device)
 
-# Chat function with debug/error handling
+# Chat function
 def chat(message):
-    if not message or not message.strip():
-        return "Please enter a message."
-
     try:
-        # Tokenize
-        inputs = tokenizer(message, return_tensors="pt")
-        inputs = {k: v.to(device) for k, v in inputs.items()}
-
-        # Cast to float16 only if model is on CUDA
-        if device == "cuda":
-            inputs = {k: v.half() for k, v in inputs.items()}
-
-        # Generate
+        # Tokenize input (do NOT convert to .half())
+        inputs = tokenizer(message, return_tensors="pt").to(device)
+        
+        # Generate output
         outputs = model.generate(
             **inputs,
             max_new_tokens=150,
             do_sample=True,
             temperature=0.7,
-            top_k=50,
             top_p=0.95
         )
 
+        # Decode output
         response = tokenizer.decode(outputs[0], skip_special_tokens=True)
         return response
 
-    except RuntimeError as e:
-        if "CUDA error" in str(e):
-            return "⚠️ CUDA error during generation. Try restarting or changing your input."
-        return f"Unexpected error: {e}"
     except Exception as e:
-        return f"Error: {e}"
+        print("Unexpected error:", e)
+        return "An error occurred during generation."
 
-# Gradio UI
+# Launch Gradio interface
 iface = gr.Interface(fn=chat, inputs="text", outputs="text", title="Gemma Chatbot")
 iface.launch()