Update app.py

app.py

@@ -1,42 +1,34 @@
+import gradio as gr
+from transformers import AutoTokenizer, AutoModelForCausalLM, AutoConfig
 import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer
 from peft import PeftModel
-import gradio as gr
 
-# --------------------
-# Load Base Model and LoRA Adapter
-# --------------------
+# Function to load the base model and LoRA adapter
 def load_model_and_adapter():
-    base_model_name = "unsloth/Llama-3.2-3B-Instruct"  # Replace with your base model name
-    adapter_repo = "Futuresony/future_ai_12_10_2024"   # Your Hugging Face LoRA repo
-    
-    # Load tokenizer
+    base_model_name = "Futuresony/future_ai_12_10_2024.gguf"  # Replace with your model path or ID
+    adapter_path = "./adapter"  # Adjust this path to your LoRA adapter files location
+
+    # Load configuration and tokenizer
+    config = AutoConfig.from_pretrained(base_model_name)
     tokenizer = AutoTokenizer.from_pretrained(base_model_name)
-    
-    # Fix rope_scaling in model configuration
-    config = AutoModelForCausalLM.config_class.from_pretrained(base_model_name)
-    if hasattr(config, "rope_scaling"):
-        config.rope_scaling = {"type": "dynamic", "factor": 32.0}  # Override with valid keys
-    
-    # Load base model with fixed config
+
+    # Load the base model
     base_model = AutoModelForCausalLM.from_pretrained(
         base_model_name,
-        config=config,
-        torch_dtype=torch.float16,  # Use float16 for efficiency if GPU is available
-        device_map="auto"           # Automatically map to GPU or CPU
+        torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
+        device_map="auto",
     )
-    
+
     # Load LoRA adapter
-    model = PeftModel.from_pretrained(base_model, adapter_repo)
-    model.eval()  # Set to evaluation mode
+    model = PeftModel.from_pretrained(base_model, adapter_path)
+
+    # Ensure the model is ready
+    model.eval()
+
     return tokenizer, model
 
-# Load the model and tokenizer once
-tokenizer, model = load_model_and_adapter()
 
-# --------------------
-# Generate Response Function
-# --------------------
+# Function to handle the conversation
 def respond(
     message,
     history: list[tuple[str, str]],
@@ -45,49 +37,58 @@ def respond(
     temperature,
     top_p,
 ):
+    # Load tokenizer and model
+    global tokenizer, model
+    tokenizer, model = load_model_and_adapter()
+
+    # Build the input prompt
     messages = [{"role": "system", "content": system_message}]
-    
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-    
+    for user_input, assistant_response in history:
+        if user_input:
+            messages.append({"role": "user", "content": user_input})
+        if assistant_response:
+            messages.append({"role": "assistant", "content": assistant_response})
     messages.append({"role": "user", "content": message})
 
-    # Prepare input prompt for generation
-    prompt = "\n".join([f"{m['role']}: {m['content']}" for m in messages])
-    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+    # Tokenize input
+    inputs = tokenizer.apply_chat_template(
+        messages,
+        add_generation_prompt=True,
+        return_tensors="pt",
+    ).to(model.device)
 
     # Generate response
-    outputs = model.generate(
-        **inputs,
-        max_length=max_tokens,
+    output = model.generate(
+        inputs,
+        max_new_tokens=max_tokens,
         temperature=temperature,
         top_p=top_p,
         pad_token_id=tokenizer.eos_token_id,
-        eos_token_id=tokenizer.eos_token_id
     )
-    
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    response = response.split("assistant:")[-1].strip()  # Clean response
+
+    # Decode response
+    response = tokenizer.decode(output[0], skip_special_tokens=True)
+
     return response
 
-# --------------------
-# Gradio Interface
-# --------------------
+
+# Gradio interface
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
-        gr.Textbox(value="You are a helpful assistant.", label="System message"),
+        gr.Textbox(value="You are a helpful AI assistant.", 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=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
+        gr.Slider(
+            minimum=0.1,
+            maximum=1.0,
+            value=0.95,
+            step=0.05,
+            label="Top-p (nucleus sampling)",
+        ),
     ],
 )
 
-# --------------------
-# Launch the Interface
-# --------------------
 if __name__ == "__main__":
     demo.launch()
+